Google Cloud has recorded its best quarterly growth to date, with earnings reports showing revenue surging 82% to $24.8bn.

Across the wider company, Google-owner Alphabet saw revenue climb 24% to $119.8bn, with income up 30%.

Despite such gains, Google was still on the back foot, with share prices falling 3% after revealing a big boost in capital expenditure and being forced to defend delays to its top-end AI model.

"Q2 was an amazing quarter, with Alphabet revenues growing 24% year-over-year and Google Cloud revenues accelerating to 82% growth, driven by demand for AI infrastructure and AI solutions," said CEO Sundar Pichai in a statement. "It’s great to see wide adoption of Gemini Enterprise, with nearly 90% of the Fortune 100 using it."

"Our AI investments are redefining what’s possible across every part of our business."

Continued capex concerns

After the results were announced, Google admitted in a conference call it will increase its capital expenditure this year, which is largely focused on AI infrastructure spending such as data centers.

Last quarter, that was bumped up by $5bn to an already massive $180bn to $190bn, and it has now been increased to $195bn to $205bn.

Over the second quarter alone, Google recorded capex of $45bn, with most used to boost "technical infrastructure to support our investments in AI," said Anat Ashkenazi, senior VP and CFO, according to a transcript of the conference call.

"The increase in the range is primarily due to an acceleration in the delivery of capacity to meet growing demand," said Ashkenazi, adding that spending would "increase significantly" in 2027.

Google isn’t alone in dealing with surging capex rates. Amazon Web Services (AWS) is projected to spend $200bn and Meta plans as much as $145bn.

These huge outlays have sparked concerns about the cost of AI over the last 18 months, but industry leaders insist it's all part of the process. Pichai compared this stage of AI to the early days of the shift to the cloud.

"I do think it feels like we are in very early innings of what feels like a secular shift across multiple areas in our core information businesses just the possibilities when I see what all you can do with the absolute frontier capabilities, there's still a lot of work ahead to translate all that into experiences for our consumer users," he said, in response to an analyst question.

Ashkenazi added that "while we have increased our capacity quite significantly over the past 3 years, the demand still outpaces that investment."

AI delays

Google Cloud might be flying high, but concerns over AI model-related delays are rising among investors. The firm’s flagship model was once again delayed this week.

The company unveiled Gemini 3.6 Flash and a cyber-focused model, Gemini 3.5 Flash Cyber this week, following the lead of competitors such as Anthropic in the space.

However, Gemini 3.5 Pro wasn’t included in the launch after initially being set for release in June. Google said it would be made available "as soon as it's ready".

"Gemini 3.5 Pro is currently in testing and our team is already building the next generation of models," Pichai said on the conference call. "We have started our most ambitious pretraining run yet for Gemini 4 and are excited by the progress we are seeing at the frontier."

That didn't convince one analyst, Douglas Anmuth from JPMorgan Chase & Company, who asked about Pichai's confidence in keeping Gemini models at the frontier — the front of the pack, the leading edge of AI development.

Pichai defended Google's AI efforts, saying the company was "very committed and very confident of being at the frontier, and, model releases aside, remains a leader on "many attributes" — though he admitted work was needed on coding and agentic coding.

"I am very excited by the progress I'm seeing internally on Gemini 4, and I'm confident that people will be pleased when we are putting it outside," he added.

FOLLOW US ON SOCIAL MEDIA

Google could be facing a fine from the EU in the high hundreds of millions of euros, according to reports, following an investigation by the European regulator into its search.

The fine relates to a European Commission (EC) investigation into whether Google breached the Digital Markets Act by favouring its own search engine. Launched last year, the EC sent preliminary findings to the company, accusing it of preferential treatment of Google's own services in its search results, including shopping, hotels, finance, and more. Google was then given a chance to respond and suggest changes to address the problems.

A report in the German newspaper Handelsblatt reported that the EC is considering fines in the hundreds of millions of euros after failing to come to an agreed solution.

Next steps

However, an EC spokesperson told Reuters that the regulator was keen to get Google in line rather than issue a punitive fine, suggesting that the EC was more interested in securing compliance.

"Even with ‌our ⁠negotiations on future solutions, we will not hesitate to move to the next steps as soon as possible," spokesperson Thomas Regnier told Reuters.

The EC earlier this month gave Google more time to develop a proposal after its previous efforts to avoid a fine failed to win over the regulator. But Google said in response to Reuters that it had already made too many changes to its search tool. ITPro contacted Google for comment, but has yet to hear back at the time of publishing.

Last year, Google-owner Alphabet's profit topped $132 billion with revenue over $400bn. While this fine would be the largest under the 2022 Digital Markets Act, the company has been hit by much bigger fines than that proposed by the EC this time around. In 2017, it was fined €2.4 billion for antitrust complaints related to Google Shopping, and the next year was hit by a €4.3bn over Android and €1.4bn for a case relating to online advertising. Last year, the EC leveled a €‎2.95bn fine following an investigation into adtech, and started a fresh investigation into Google over its use of online content to train its AI.

Appeal in the US

The reports come days after Google filed an appeal in US courts against a decision in 2024 that saw the company deemed a "monopolist" for its search deals, which included paying billions of dollars annually to rivals, including Apple, to use Google as their default for search. Google was ordered to share some search data with rivals, among other remedies.

Such an appeal is no surprise, but Google is arguing that the judge in the case made legal errors and that its search engine dominated because it was technically "superior."

"Competition produces winners and losers," the company said in its filing. "And sometimes a firm – by innovating better, investing more, or just working harder – will leave its rivals behind."

Last week, Google unveiled significant changes to its search tool, heavily embedding AI into the financially lucrative service.

Google has teased a new line-up of laptops coming later in the year called Googlebooks.

In what was a brief announcement at the Android Show, Googlebooks appear to be a confirmation of a long-running rumor: a new operating system based on a fusion of ChromeOS and Android, and essentially a Gemini-native successor to Chromebooks.

“We’re bringing together the best of Android, which comes with powerful apps on Google Play and a modern OS that’s designed for Intelligence, and ChromeOS, which comes with the world’s most popular browser,” Google’s senior director for laptops and tablets, Alex Kuscher, wrote in a blog post.

“The result is Googlebook: a new category of laptops built with Gemini’s helpfulness at its core, designed to work seamlessly with the devices in your life and powered by premium hardware. We’re sharing a sneak peek into the Googlebook experience today and will have a lot more to share later this year.”

What to expect with Googlebooks

Googlebooks will be the first laptops designed “from the ground up” for Gemini Intelligence, the company said in its announcement. The first Googlebooks are expected in the fall from the usual vendors such Acer, Asus, Dell, HP, and Lenovo.

Gemini and its AI capabilities will assist you the moment you open a Googlebook laptop, the company said. One of the first new features is the “Magic Pointer” which adds extra functions as you hover the cursor over various apps.

The cursor will also come alive with “just a wiggle”, according to Google. The wiggle brings up quick and contextual suggestions that depend on where it is on the screen.

Pointing at a date in Gmail, for example, will suggest setting up a meeting. Or selecting two images can create a visualization; the presentation showed a couch being placed in a living room.

Many of the features appeared to be more about everyday life, rather than business use cases. Though it isn’t hard to imagine where capabilities like custom widgets or remote mobile access can help at work.

The latter will allow you to access your mobile from the laptop.

The announcement does raise a few questions, particularly for the future of Chromebooks, but we expect more details to come over the summer.

FOLLOW US ON SOCIAL MEDIA

Cyber criminals have been observed using AI to build a working zero-day exploit in a case Google researchers say is the first of its kind.

According to new research from Google Threat Intelligence Group (GTIG), a threat actor planned to deploy the zero-day in a “mass exploitation event” but was thwarted.

The company said that its “proactive counter discovery” may have prevented its use.

The zero-day in question bears all the hallmarks of an AI-generated exploit, researchers noted, largely due to the fact the script contained an “abundance of educational docstrings" as well as a hallucinated CVSS score.

Other tell-tale signs, such as a “textbook” Python format which is characteristic of LLM training data, also gave the game away for the threat actor.

Google noted that it does not believe its Gemini model was used to develop the zero-day exploit, but said it has “high confidence” another AI model was used.

John Hultquist, chief analyst at GTIG, said the discovery marks a significant moment in the use of AI for nefarious purposes.

“There’s a misconception that the AI vulnerability race is imminent,” he said. “The reality is that it’s already begun.”

Under the hood of the zero-day

According to GTIG, the vulnerability is classified as a two-factor authentication (2FA) bypass. However, researchers noted it requires valid user credentials.

What made the flaw stand out is that it stems from a “high-level semantic logic flaw where the developer hardcoded a trust assumption”.

Simply put, this particular flaw is the type that could be easily identified by large language models, and researchers deduced that the model’s reasoning capabilities allowed it to read the developer’s intent during development.

“While fuzzers and static analysis tools are optimized to detect sinks and crashes, frontier LLMs excel at identifying these types of high-level flaws and hardcoded static anomalies,” researchers noted.

“Though frontier LLMs struggle to navigate complex enterprise authorization logic, they have an increasing ability to perform contextual reasoning, effectively reading the developer's intent to correlate the 2FA enforcement logic with the contradictions of its hardcoded exceptions.”

Nation-state activity accelerates

The discovery by GTIG highlights the growing appeal of AI-based tools for cyber criminal groups and nation state-backed threat actors.

“Threat actors are leveraging AI to augment various phases of the attack lifecycle,” researchers said.

“This includes supporting the development of vulnerability exploits and malware, facilitating autonomous execution of commands, enabling more targeted and well-researched reconnaissance, and improving the efficacy of social engineering and information operations.”

Indeed, the Google threat intelligence wing said groups in China, North Korea, and Russia in particular are flocking to AI for vulnerability research and exploit development.

“These actors have leveraged sophisticated approaches toward AI-augmented vulnerability discovery and exploitation, beginning with persona-driven jailbreaking attempts and the integration of specialized, high-fidelity security datasets to augment their vulnerability discovery and exploitation workflows,” researchers said.

One threat group, tracked as UNC2814, was observed using expert persona prompting in Gemini to research remote code execution flaws in TP-Link router firmware and Odette File Transfer Protocol (OFTP) implementations.

Another group, APT45, was observed sending thousands of repetitive prompts to analyze different CVEs and validate PoC exploits, according to researchers.

This, GTIG added, is helping them to create a “more robust arsenal of exploit capabilities” that would be difficult to tackle for defenders.

FOLLOW US ON SOCIAL MEDIA

Microsoft, Google, and xAI have agreed to hand over their AI tools to the US Center for AI Standards and Innovation (CAISI) and the UK's AI Security Institute (AISI) for pre-deployment testing.

They will evaluate the firms' frontier models, assess safeguards, and help mitigate national security and large-scale public safety risks, Microsoft said.

"Well-constructed tests help us understand whether our systems are working as intended and delivering the benefits they are designed to provide. Testing also helps us stay ahead of risks, such as AI-driven cyber attacks and other criminal misuses of AI systems, that can emerge once advanced AI systems are deployed in the world," said Natasha Crampton, Microsoft’s chief responsible AI officer.

"While Microsoft regularly undertakes many types of AI testing on its own, testing for national security and large-scale public safety risks necessarily must be a collaborative endeavor with governments. This type of testing depends on deep technical, scientific, and national security expertise that is uniquely held by institutions like CAISI in the US and AISI in the UK and the government agencies they work with."

In the US, Microsoft and the National Institute of Standards and Technology (NIST) will collaborate with CAISI on improving methodologies for adversarial assessments.

It will mean testing AI systems by examining unexpected behaviors, misuse pathways, and failure modes. This includes developing more systematic and reproducible approaches to evaluation, including shared frameworks, datasets, and workflows for assessing safety, security, and robustness risks in advanced AI systems.

“Independent, rigorous measurement science is essential to understanding frontier AI and its national security implications,” said CAISI director Chris Fall. “These expanded industry collaborations help us scale our work in the public interest at a critical moment.”

Microsoft strikes agreement with UK researchers

In the UK, Microsoft will collaborate with AISI on research related to frontier safety and security, including ways of evaluating high-risk capabilities and the effectiveness of the safeguards used to address them

"The partnership will also include research into societal resilience, examining how conversational AI systems interact with users insensitive contexts," said AISI.

"As AI systems become increasingly capable, sustained two-way collaboration between government and companies developing and deploying frontier AI is essential to advance our joint understanding of large-scale risks to public safety and national security."

Microsoft said future plans include collaborating with other AI institutes around the world, sharing priorities and methodologies for testing through the International Network for AI Measurement, Evaluation and Science.

The company is also working with Frontier Model Forum (FMF), an initiative dedicated to advancing the science and practice of frontier AI safety and security, to support independent research and promote transparency around risk mitigation strategies.

It is also contributing to MLCommons, a multistakeholder non-profit that develops and operationalizes testing tools such as AILuminate, a family of safety and security benchmarks.

"As AI capabilities advance, so too must the rigor of the testing and safeguards that underpin them. We will apply what we learn from these partnerships directly into how we design, test, and deploy AI systems, ensuring that progress in evaluation science translates into safer, more secure products for our customers," said Crampton.

"As these partnerships progress, we will share what we learn and look for opportunities to apply insights and best practices to AI testing more broadly."

FOLLOW US ON SOCIAL MEDIA

Google is reportedly building its own OpenClaw alternative as the tech giant continues its rapid-fire agentic AI push.

According to reports from Business Insider, the OpenClaw-style agent, dubbed “Remy”, is being tested internally at the firm through a staff-only version of the Gemini app.

Sources told the publication the agent will also be integrated with a host of Google services.

"Remy is your 24/7 personal agent for work, school, and daily life, powered by Gemini," internal documents state.

"It elevates the Gemini app into a true assistant that can take actions on your behalf — not just answer questions or generate content."

Development of the agent comes amid sharp interest in OpenClaw since the beginning of the year. The open source agent, which is designed to run locally on devices, allows users to create their own custom agents which can carry out tasks and take actions on their behalf.

Google has sharpened its focus on agents and orchestration in recent months. As ITPro reported at Google Cloud Next last month, the firm expanded its Gemini Enterprise offering, with an eye on enabling customers to manage agents more efficiently.

That expansion saw the launch of Agent Runtime, a platform designed specifically for agent orchestration.

At present, Google doesn’t have an outright alternative to OpenClaw, and materials shared with Business Insider suggest Remy could fill in this gap.

“Deeply integrated across Google, Remy can monitor for things that matter to you, handle complex tasks proactively, and learn your preferences over time,” an internal description states.

Exact details on a release timeline are yet to be revealed, and sources told the publication that Remy is currently in a “dogfooding” stage. This refers to a practice whereby tech companies use their own products internally to test and fine-tune them before public release.

Google is currently preparing for its annual I/O conference, which is set to take place between 19 and 20 May.

Here, the tech giant could offer customers a preview of the Remy product, and as ITPro reported last week, speculation has been mounting over a raft of potential new product launches - with Gemini 4 chief among them.

Remy-ready models

A Google alternative to OpenClaw would be a complete no-brainer for the hyperscaler, which is slowly succeeding in raising public awareness of its Gemini offerings to the same level as ChatGPT.

OpenClaw became a hit earlier this year, with AI enthusiasts drawn in by the appeal of an on-device, personalized AI assistant.

Google has been working especially hard in the past two years to make inference cost-effective. This is the bread and butter of enterprise agentic AI, as it only makes sense to run an agent around the clock if the cost of doing so isn’t ruinously high.

Google Cloud has already provided a lot of the innovation needed to keep costs low for cloud agents via TPU 8i, its new inference-optimized tensor processing unit (TPU).

For on-device processing, Google DeepMind is focused on deployment of Gemma 4, the latest iteration in its open source Gemma model family. If I had to guess, this will be the default model for Remy as and when it arrives, to deliver the same open source, on-device functionality that originally made OpenClaw famous.

In just the past 24 hours, Google DeepMind announced a major latency improvement for Gemma 4, via the introduction of multi-token prediction (MTP) drafters. These are tiny, specialized LLMs that help Gemma 4 to ‘predict’ multiple output tokens at once to reduce response times.

Google DeepMind claimed that on an Apple M4, this improved Gemma 4 31B token-per-second up to 2.5x. It’s significant that, in addition to pixel TPUs and Nvidia GPUs, Google DeepMind specifically draws focus to Apple silicon, given that the M4 Mac Mini is one of the most popular devices to run OpenClaw.

For me, this is a clear signal that something like Remy is in the works.

OpenClaw also allows users to connect cloud models to their messaging platform of choice via API calls, and to this end I’d expect Remy to also connect to Gemini 3.1 Flash – or Gemini 4 models, as the case may be – for more ‘powerful’ use cases.

Here, TPU 8i could come in handy, particularly if Google Cloud begins to offer Remy functionality at the enterprise scale.

If Google pulls this off, it will be another major corner of the AI market that it’s managed to corner.

FOLLOW US ON SOCIAL MEDIA

Hyperscaler cloud revenues are still surging, according to recent earnings reports, but questions still remain about how much longer they can continue investing massive sums in AI infrastructure.

Alphabet, AWS, Meta, and Microsoft have all posted positive results this quarter with overall revenue in the hundreds of billions of dollars, helped in part by booming demand for data center capacity amid the AI roll-out.

However, analysts have questioned whether spending on AI is sustainable and will deliver a return on investment (ROI).

The current estimate is for tech's leading companies to invest $600 billion in infrastructure to power AI this year alone, with some tech giants already upping their predicted spends as part of their quarterly results.

"The public cloud platform earnings numbers are big as usual — but the capital investment to achieve them are getting big faster," said Forrester’s principal analyst Lee Sustar. "That's why questions will persist about the sustainability of hyperscaler AI data center buildouts."

"Together, the public cloud revenue numbers and the capital expense figures will raise questions about the sustainability of the AI boom," he added. "But with the potential payoff of AI leadership seemingly so high, the companies continue to make those bets, forcing investors and customers alike to assess how their interests are impacted."

According to LSEG data shared by Reuters, Amazon stock is up 13%, Alphabet's is up just shy of 12%, and Meta is up 3% – but Microsoft has slid 12%.

Alphabet's cloud success

Google-owner Alphabet saw overall revenue leap 22% to $109.9 billion for its first quarter, topping estimates, with income climbing to $39.7 billion.

Google Cloud climbed 63% to $20 billion, marking the best growth rate for the cloud company in more than six years, though still below Alphabet's other revenue sources, including Services at $89.6 billion and Search at $60.3 billion.

Google Cloud income climbed from $2.2 billion in the first quarter in 2025 to $6.6 billion this year, underlining significant growth over the last 12 months. Sustar said the figures show the cloud division "can significantly contribute to the wider Alphabet portfolio after years of big operating losses".

In an investor call, Sundar Pichai said that capacity constraints prevented even higher growth, with Google Cloud's backlog doubling to $460 billion.

"Our enterprise AI solutions have become our primary growth driver for cloud for the first time," he said.

In a statement, Pichai added: "Our first-party models, like Gemini, are now processing more than 16 billion tokens per minute via direct API use by our customers, up 60% from last quarter. It’s really exciting to see how our AI investments are delivering value for our users, customers and business."

CFO Anat Ashkenazi said Google would increase its capital expenditure forecast by $5 billion to somewhere between $180 billion and 190 billion. Last year, the tech giant’s capital expenditure reached $91.45 billion, with much of this focused on AI-related spending.

That huge cloud backlog and leap in revenue suggests that Google's capex spending isn't wildly out of place, Thomas Monteiro, a senior analyst at Investing.com, told Reuters.

"Perhaps even more importantly than Alphabet's massive cloud growth pace is the broader justification that the $180 billion capex plan — that surprised the market last quarter — is well within the company’s spending power, considering the durability and quality of the revenue curve shown today," he said.

But concerns remain, according to Sustar, who pointed to the quarterly figures. He noted that "the capital expenditure of $35.7 billion — which Alphabet attributes largely to AI — is much bigger" than the $20 billion in quarterly revenue from the cloud unit.

AWS leaps ahead, but not quite as much

Amazon's overall revenue climbed 17% to $181.5 billion in the first quarter, while revenue at Amazon Web Services (AWS) climbed 28% to $37.6 billion in the first quarter. AWS operating income climbed to $14.2 billion from $11.5 billion in the same quarter last year.

While AWS’ revenue growth rate wasn't as high as Google Cloud's, it's still significant coming from the market leader. Andy Jassy, CEO of Amazon, said in a statement that it was AWS' "fastest growth in 15 quarters."

Jesse Cohen, senior analyst at Investing.com, said that the "reacceleration" in AWS sales was a positive sign, adding that the hyperscaler’s customer base is "fully embracing new workloads, especially in AI."

Amazon has projected $200 billion in capex this year, with $44.2 billion spent in this quarter.

"AWS bagged $37.59 billion in revenue in the first quarter but data center buildouts accounted for a big part of the $44.2 billion capital investment in the same period, even though AWS had an operating income of $14.2 billion," said Sustar.

"That puts pressure on AWS to keep up its historic role of keeping Amazon as a whole in the black from quarter to quarter.

Social media props up Meta

Facebook owner Meta posted revenue of $56.3 billion, up 33% from $42.3 billion the same quarter last year, with income of $22.8 billion, up about 30%.

Meta raised its expected capital expenditure forecast from between $115 billion-$135 billion to $125 billion-$145 billion, which sent stocks sliding 6%.

In a statement, Meta said: "This reflects our expectations for higher component pricing this year and, to a lesser extent, additional data center costs to support future year capacity."

In a conference call, Meta's chief financial officer Susan Li said that such spending was necessary to catch up, as Meta has previously "underestimated our compute needs".

CEO Mark Zuckerberg said it was unclear how the increased spending would pay off.

"I don't think we have a very precise plan for exactly how each product is going to scale or anything like that," he said, per reports from the BBC.

"But I think we have a sense of the shape of where these things should be... and I'm quite comfortable that the lab we're building is on track to be a leading lab in the world."

In the meantime, social media continues to pay Meta's bills.

“The irony is that Meta’s future‑facing AI ambitions are being underwritten almost entirely by the company’s legacy business: advertising inside social media apps," said Forrester’s VP research director Mike Proulx. "There’s no material AI revenue yet. Reality Labs remains a loss. Muse Spark has momentum, but not monetization."

“The question is whether Meta’s core can continue to act as a cash cow while the company reduces headcount and diverts focus toward AI," he added.

"If Meta’s ad engine slows, the market’s margin for patience shrinks fast. Meta’s slight dip in daily active users is already beginning to raise eyebrows. Q2 will tell us if it’s really just a blip or the start of a trend."

Microsoft says margins are strong

Microsoft said its quarterly revenue was up 18% to $82.9 billion, with operating income up 20% to $31.8 billion, though AI spending helped cash flow slip by $6 billion to $15.8 billion.

"We are focused on delivering cloud and AI infrastructure and solutions that empower every business to eval-max their outcomes in the agentic computing era," said Satya Nadella, Microsoft CEO.

"Our AI business surpassed an annual revenue run rate of $37 billion, up 123% year-over-year."

Microsoft Cloud revenue climbed 29% to $54.5 billion, with Azure and other cloud services up 40%.

"For its part, Microsoft claims an upswing of 40% growth in Azure and $54.5 billion in revenue for the wider Microsoft Cloud category," said Forrester's Sustar.

"The company does not break out Azure operating revenue, but with its AI-focused capital investments of $31.9 billion, the cloud revenue is clearly expensive to obtain."

Microsoft's spending on AI has raised concerns, but the company pushed back saying income remains strong. "The margins have been and remain better in our AI business than when we were in our cloud transition," said CFO Amy Hood.

FOLLOW US ON SOCIAL MEDIA

The AI race may have Google and Anthropic pitted as rivals, but that hasn't stopped the tech giant from investing up to $40 billion in the company.

That investment includes $10bn in cash now at Anthropic's previous valuation of $350bn, while the $30bn will arrive if the Claude developer hits performance milestones.

At time of writing, Anthropic has not explained what these performance milestones look like in detail. Neither Anthropic nor Google responded to a request for comment from ITPro by time of publication.

Google has previously invested in the AI developer, coughing up $2 billion back in 2023 and a further $1 billion in 2025, followed by a cloud deal worth tens of billions for up to a million of Google's TPUs.

The deal follows an agreement Anthropic struck with Broadcom and Google earlier in April to bring “multiple gigawatts of next-generation TPU capacity” online from 2027 onwards.

A likely reference to Google’s 8t and 8i TPUs, the majority of these next-generation TPUs will be built in the United States and form what Anthropic said is a diverse platform of training hardware based on AWS, Google, and Nvidia partnerships.

Anthropic ramps up industry collaboration

Anthropic has made a number of high-profile agreements in recent months as it looks to secure massive amounts of compute to train its frontier models. Last week, Amazon and Anthropic announced a deal worth up to $25 billion, following on from a previous investment of $8 billion.

The latest investment will see Anthropic get an additional 5GW of capacity from the cloud giant, and like the Google funding, splits the investment; Anthropic will get $5 billion now with the extra $20 billion tied to "certain commercial milestones."

"Our users tell us Claude is increasingly essential to how they work, and we need to build the infrastructure to keep pace with rapidly growing demand,” said Dario Amodei, CEO and co-founder of Anthropic, following the Amazon investment announcement.

These latest funding announcements follow an investment round in February that raised $30bn and valued the American AI startup at $380bn, a significant leap since its $183bn valuation in September.

At the time, the company said it would help it expand infrastructure and product development, as well as support future frontier model research.

Anthropic has also recently signed major infrastructure deals with cloud provider CoreWeave.

AI funding race

The AI arms race has led to massive investment in infrastructure. Amazon said back in February it expects to spend $200bn to expand compute capacity, with Google reportedly set to invest up to $185bn, with more than half on AI infrastructure.

Across Microsoft, Google-owner Alphabet, Amazon, and Meta, more than $650bn in AI capital spend is expected across 2026 alone.

Anthropic, in particular, says it plans to invest $50 billion in data centers for training and model use in the US this year.

That spending has raised concerns of an AI bubble, particularly with regard to whether the technology can ever pay off after such high spending, with many companies failing to see promised productivity gains or any return on investment as of yet.

Despite that, two-thirds of respondents to a recent survey said they would continue to invest in AI even if they didn't see tangible returns.

Anthropic's annualized revenue is now predicted to stand at $30 billion, versus $9 billion last year, according to Reuters, and is projected to break even in 2028 ahead of rival OpenAI.

FOLLOW US ON SOCIAL MEDIA

Good morning from Las Vegas, where Google Cloud is kicking off its annual event. ITPro's Rory Bathgate is live on the ground, and ready to cover all the announcements and customer stories from the opening keynote. Titled 'The agentic cloud', expect to hear from CEO Thomas Kurian and other executives as they explain how Google Cloud is preparing itself for the coming wave of AI agents.

Google Cloud already made a number of announcements this morning, which we've covered here:

• Google Cloud announces eighth-generation TPUs, boasting AI training and inference leaps
• Google expands Gemini Enterprise, consolidates Vertex AI services to simplify agent deployment

Good morning from Las Vegas! I'm here on the ground at Google Cloud Next, where the opening keynote is just minutes away from kicking off.

And we're off, with what appears to be an AI-generated opening animation including Google Cloud products, customers, and achievements.

Thomas Kurian has now taken to the stage. He's kicked off the keynote by acknowledging the 75% of Google Cloud customers who use the firm's AI tools.

Google's advantage is its full-stack approach, he says. To tell us more, we're now hearing from Sundar Pichai, CEO at Google (via a pre-recorded video address).

Pichai says that in 2022 Google invested $32 billion in CapEx, and now expects to spend $175-185 billion in CapEx.

Today, Pichai says, Google is using agents not only to produce code, but migrate codebases.

In a recent example, Pichai adds, Google developers found that they could migrate code 6x faster with AI agents. This approach is now informing the entire company's approach to development.

Security teams are also benefiting from AI agents, he says, with security operations center agents triaging thousands of unstructured threat reports each month.

Pichai says that businesses are focused on how to manage thousands of agents. For this purpose, Google Cloud has revamped Gemini Enterprise as a "mission control" application. Read more in our write up here.

Kurian is now back on stage, to talk more about Gemini Enterprise.

"In bringing Google Al to every employee and every workflow, Gemini Enterprise is now the end-to-end system for the agentic era," he says.

"The connective tissue between your data, your people and your goals, it transforms disconnected processes into a single intelligent flow. This is our blueprint for the agentic enterprise."

Zooming in on specific features, Kurian summarizes some of the key benefits of the newly-unified Gemini Enterprise Agent Platform. This new offering connects existing Vertex AI features with wider agent deployment and management tools, to help equip businesses with a comprehensive dashboard for their AI agents.

Google Cloud is leaning into its partnerships, and Kurian proudly announces that models such as Claude Opus 4.7 will be available on the platform as default.

In a tease, he also acknowledges Google's partnership with Apple, which will see Gemini underpin AI personalization features within the iOS ecosystem.

"Earlier this year, we announced a monumental partnership with one of the world's most iconic brands, that'll bring the power of our technology to users everywhere around the world," he said.

"We're collaborating with Apple as their preferred cloud provider to develop the next generation of Apple foundation models, based on Gemini technology. These models will help power future Apple Intelligence features, including a more personalised Siri coming later this year."

Kurian also proudly announces that Gemini Enterprise agents were used by NASA for key elements of the Artemis II moon mission.

Managing agents can be a major task for IT teams, especially if agentic actions can't be easily tracked or observed. Features such as Agent Observability help streamline these processes, Kurian says, as covered in more detail in our write-up linked above.

Google Cloud really wants workers to collaborate with agents and this is what Projects in Gemini Enterprise is all about. It's a collaboration platform that combines human and AI-generated work, with built-in functionality for sharing docs to Microsoft 365.

We're now hearing from Virgin Voyages about how it's using Gemini Enterprise to streamline its customer and crew experience.

The firm has reduced its production timeline by 60% and raised month on month sales by 28%.

Using Google Distributed Cloud Edge, Virgin Voyages also runs Gemini models securely on its cruise ships.

It's time for a demo to put this all in context. Here to show us how it all works is Erica Chuong, manager, Applied Al Forward Deployed Engineering at Google Cloud.

With one prompt, Chuong can research interior design trends in her demo identity as an employee at a furniture retailer.

Based on the results, Chuong can also ask Gemini Enterprise to generate landing page videos for her website using Veo 3. She can then ask Gemini to automatically inform the relevant developer about the changes needed for the website.

With that, we're back to Kurian, who is running us through some customer examples.

Walmart is using Gemini Enterprise to improve its in-store operations. It's providing each worker with a Pixel Fold, with Gemini on device helping them to answer customer queries and solve supply chain problems.

Google Cloud also worked closely with Team USA in the 2026 Winter Olympics.

Shaun White, Olympic gold medallist for snowboarding, has just appeared in a blast of fog and accompanied by arena-wide fake snow to tlel us more.

Jason Davenport, arena tech lead for Developer Experience and AI at Google, is on stage alongside White to run us through some of the innovations that Google has made for winter sports.

Using a tool built by Google DeepMind, Google Cloud can analyze snowboarding footage, mapping a wireframe over White and tracking stats such as rotational velocity, flight dynamics, and tuck compression.

Moving on, we're now hearing from Amin Vahdat, SVP and chief technologist, AI and Infrastructure at Google.

Vahdat starts on the subject of Google Cloud's AI Hypercomputer, its end-to-end energy, software, and infrastructure stack that powers its data centers.

But Vahdat is really here to formally announce TPU 8t and 8i, Google Cloud's newest custom chips for AI workloads. We've covered these in more detail here.

The TL;DR for these chips is that Google Cloud has now split its TPUs into two distinct offerings, one optimized for training and the other for inference.

This is crucial as the hyperscaler looks to support massive AI agent demand, with huge leaps achieved for reducing latency and improving its networking technologies to loop ever more chips into distributed clusters.

These chips are underpinned by Google Cloud Axion, its in-house CPU that Vahdat explains offers 2x price performance vs x86 instances.

Google Cloud doesn't solely rely on TPUs, nor does it expect its customers to go all-in on the specialized chips. Vahdat announces that Google Cloud will be one of the first cloud providers to gain access to Nvidia Vera Rubin NVL72, which provides many exaflops of inference per data center rack.

In practice, TPUs are already speeding up workloads. Citadel Securities is already using Google's previous generation TPU, Ironwood, to speed up its workloads many times over, reducing tasks that used to take months or weeks to just minutes.

These systems can't be managed by humans alone, Vahdat says:

"You cannot have humans managing and troubleshooting configurations: you need a cloud that drives itself. We have used some model context protocol to turn every Google Cloud service into a tool that agents can orchestrate directly by integrating Gemini reasoning into our own telemetry. The system now performs autonomous root cause analysis identifying and fixing this configurations before you even realize."

All of this is underpinned by Google Cloud's vast data fabric and pipeline, with Gemini built-in as standard.

Here to explain more is Karthik Narain chief product and business officer at Google Cloud.

Today, Google Cloud is announcing Knowledge Catalog, a universal context engine for enterprise AI, he says.

Knowledge Catalog continuously enriches enterprise data by analyzing logs and profiling data in the background.

"Now, the second that image or PDF hits Google Cloud Storage, they are instantly packed, enriched, and made agent-ready," Narain explains. "Zero manual data engineering, powered natively by Gemini, the Knowledge Catalog goes even deeper, reading files, autonomously extracting entities, mapping relationships and learning about your unique business, semantics."

Another major announcement today is a new corss-cloud lakehouse, that aims to eliminate data silos by connecting data across all of an enterprise's cloud platforms.

Combined with Knowledge Catalog, this means Google Cloud is working harder than ever to enrich enterprise data with context as standard. It also applies across partner data platforms, including Palantir, Salesforce, and Workday.

In a demo, Yasmeen Ahmad, MD, Product Management, Data and AI Cloud at Google Cloud, shows attendees how Google Cloud's new data breakthroughs are streamlining operations.

In the example, Ahmad collates data from thousands of ingredient PDFs to review a recipe for a new flavour of froyo that is suitable for soy-sensitive customers.

Ahmad can ask for Gemini to collate the data, cite sources to confirm ingredient allergens, research market size, and produce a revenue projection. Gemini is able to identify soy in an ingredient that may have been missed by human workers, then call the relevant research agents and models to pull together the business plan Ahmad asks for.

We're moving quickly here, straight onto AI powered security. Here to tell us more is Francis deSouza, COO and president, Security Products at Google Cloud.

Machine-speed security is an imperative, deSouza says, as the time between initial access by hackers and handoff to secondary threat groups has dropped from eight hours to 22 seconds.

To tackle this issue, Google is leaning on telemetry from Mandiant, VirusTotal, and Chrome and deploying triage agents to cut response times from 30 minutes to 60 seconds.

Google's acquisition of Wiz expands this protection even further. deSouza is now introducing Yinon Costica, co-founder of Wiz.

"Our mission from day one has been to help customers protect everything. They've been running. We began by unifying code, cloud, and runtime context to move at developer speed. But Al has fundamentally changed the environment," Costica says.

Wiz has launched red team, blue team, and green team agents to automatically identify, investigate, and fix critical risks faster than human security teams could.

"In the AI era, the best defense is to continuously use AI against ourselves in order to give the defenders the first mover advantage over the attackers," Costica says.

"And this is exactly what the Wiz Red Agent does. It validates every single exposure that we identified, and you can think of it as a friendly hacker that continuously scans your outside like an elite red team."

Costica is now showing us an example of a red agent having found an authentication bypass vulnerability within an enterprise AI agent.

"It's not a potential risk anymore, this is a validated risk," he says, adding that the Wiz AI Application Protection Platform (AI-APP) allows researchers to address these risks directly within the platform.

To achieve this, a user can hand the vulnerability data directly to a green agent, which can produce code to fix the risk as quickly as possible.

Back onstage, deSouza says Google Cloud is helping the Los Angeles Department of Water and Power secure critical infrastructure ahead of the 2028 Olympics. Morgan Stanley has also chosen Wiz within Google Cloud as a key component of its cloud security strategy, he adds.

Moving quickly on, Carrie Tharp, GTM COO & VP, Customer Experience at Google Cloud, is here to give examples of how the firm's agents are helping improve customer experience.

For example, Papa Johns is using Food Ordering Agent to streamline the pizza ordering process for customers.

Best Buy is using Gemini Enterprise for Customer Experience to autonomously guide customers through complex questions about product specs.

Home Depot is also leaning heavily on Gemini Enterprise, having been a Google partner for years, to build its 'Magic Apron' shopping assistant.

It's already seeing a 10% increase in sales conversions compared to traditional interactions, and is aiming to provide even more agentic experiences to customers via its website.

In India, Reliance is using Gemini to provide customers with product recommendations based on natural language requests such as "plan a birthday party".

Google is also acting as its own internal supplier, providing agents for offerings such as YouTube TV. In a demo, Patrick Marlow, senior product manager at Google Cloud, shows us how a telephony agent can answer questions about his YouTube TV plan including switching to Spanish mid-conversation.

All of these interactions are handled via CX Agent Studio, which allows IT workers to view the workflow of agents involved in customer interactions. They can also test the agents directly within the interface.

With a drop-down menu, Marlow can easily add a new sub-agent dedicated to a new promotion offer.

YouTube TV created this multilingual telephony agent in just six weeks, which Marlow says shows how Google Cloud allows leaders to "iterate at the speed of your business".

Moving on to Workspace, we're now hearing from Yulie Kwon Kim, VP of Product, Google Workspace at Google Cloud.

"We all spend half our day finding information, and the other half figuring out what to do with it," she says. To solve this issue, Google Cloud is introducing Workspace Intelligence, which embeds agents throughout productivity tasks.

In a demo, Kim shows us how a worker at a regional furniture brand can interact with Gemini directly within Google Chat.

Kim asks Gemini to find a specific merchandising playbook from last quarter, with a prompt describing it as "the one with the chart showing regional sales".

Kim can then ask Gemini to use a skill called Regional Campaign, to create a deck with a plan for Organic Living. This prompts Gemini to pull together organizational data to produce a deck inside Google Slides.

With that, it's back to Kurian to cover how Unilever is using Google Cloud at scale.

The consumer giant is using Google Cloud to speed up its AI deployment, with a goal to connect data more intelligently and strengthen its relationship with retailers.

Unilever has already deployed an agent via Gemini Enterprise Agent Platform, to speed up procurement analysis and decision-making.

Finally, Kurian pledges that Google Cloud is committed to letting its customers "choose your own destiny", with cross-platform support and by directly assisting independent software vendors and SMBs to adopt agents via Gemini Enterprise.

And with that, the opening keynote is over. ITPro will continue to bring you the latest from Google Cloud Next 2026, so be sure to check back regularly for news and analysis.

Google Cloud has expanded its Gemini Enterprise offering and consolidated existing services in a move intended to help developers manage agents, improve cross-team collaboration, and help enterprises deploy third-party agents.

Gemini Enterprise is now the central platform for all AI work within the organization, with a new ‘Inbox’ within the platform providing a single source for managing AI agents at enterprise scale.

The new Gemini Enterprise Agent Platform subsumes the features of Vertex AI, including model and agent building capabilities, and combines them with DevOps, agent orchestration, and security functionalities.

For building AI agents within the platform, users get access to the low-code Agent Studio, the open source code-first framework Agent Development Kit (ADK), and access to over 200 frontier models via Model Garden.

Beyond agent creation, the platform includes Agent Runtime, which handles agent orchestration and retains persistent context for long-form tasks, as well as Agent Identity, Agent Gateway, and Agent Registry, which together help IT teams to track agent identities and enforce AI guardrails.

Comcast used ADK and Agent Runtime to rebuild its Xfinity Assistant, which is now capable of performing personalized troubleshooting for customers.

“Agent Runtime has been a massive accelerator, allowing us to deploy a sophisticated multi-agent architecture that increases digital containment while ensuring secure, grounded interactions via Gemini,” said Rick Rioboli, CTO, Connectivity & Platforms at Comcast.

“We aren't just reducing repeat interactions by solving customers’ issues the first time; we're redefining the customer experience at scale.”

Post-deployment, admins can track and score agent behavior within the platform via Agent Evaluation, Agent Observability, and Agent Simulation. These assess agent activity in the wild, provide insights into multi-step agent reasoning to help debug errors, and test agents against synthetic user prompts.

Employees can now build and deploy agents using natural language and reusable skills within the Gemini Enterprise app, chat directly as well as manage all agent activity via a central dashboard called Inbox in Gemini Enterprise.

In a demo, Erica Chuong, Applied AI, forward deployed engineering manager NorthAM at Google Cloud, showed how an agent within Gemini Enterprise could help an employee at a furniture retailer relaunch a defunct product line.

Chuong set off a series of agents to analyze interior design trends, identify unsold stock, and come up with plans for a relaunch campaign.

“With that one prompt, multiple agents complete a series of tasks in minutes instead of hours,” said Chuong.

“It knows exactly which agents to call to come up with a plan. My market research agent, powered by Deep Research, analyzes the latest Google search information alongside our own sales and CSAT data.

“My data insights agent is connecting to our global product catalog, regardless of location or format. It knows that dead stock refers to stale inventory, and it uses our agentic data cloud to identify the right data sets. My product strategy agent pulls everything together.”

Collaboration with agents

A core aspect of Google Cloud’s new end-to-end Gemini Enterprise is encouraging workers to share AI interactions with their co-workers.

Via a new offering called Projects, employees can share sources with an agent including their enterprise Google Workspace, OneDrive, or chats in a third-party collaboration software, to help come up with new ideas and brainstorm and produce data-driven business plans.

Another new feature, Canvas, helps workers to co-create Google Docs and Slides with one another directly within Gemini Enterprise. These documents can then be shared in Microsoft 365 formats if needed, for external collaboration and feedback.

In the name of furthering collaboration and AI-driven results within the enterprise, Google Cloud has also announced interoperability with third-party agents within the Gemini Enterprise app. This allows users to deploy third-party agents directly within the Agent Gallery, for easy access to validated agents powered by partner developers.

As an example at time of the announcement, the firm pointed to agents from organizations including Adobe, Salesforce, ServiceNow, and Workday.

Google Cloud has announced a generational overhaul of its AI infrastructure, spearheaded by two new tensor processing units (TPUs) designed to meet demands for frontier model training and inference.

TPU 8t and TPU 8i represent a significant leap in performance and memory optimization over preceding hardware, underpinned by larger high bandwidth memory, faster storage access, and a redesigned machine learning architecture.

The training-focused TPU 8t is approximately 2.8x more powerful than the seventh generation TPU Ironwood, announced at last year’s event, and is capable of 121 FP4 exaflops of AI compute.

Google Cloud said that TPU 8t will dramatically reduce the development lifecycle for frontier models, allowing for new models to be trained in weeks not months.

Each 8t pod can be configured with up to 9,600 individual chips, with each capable of 19.2 Tbits/sec scale-up bandwidth and 400 Gbits/sec scale-out

Google Cloud eyes agent orchestration, inference gains

Compute for inference has fast become a bottleneck, as AI agents need consistent data center capacity to proactively complete tasks and run in enterprise environments around the clock.

To meet this demand, TPU 8i has been built from the ground up to run AI agent ‘swarms’ with minimal latency and no chips sitting idle. Each 8i TPU comes with 384MB of onboard SRAM, allowing an agent’s short term memory to be retained entirely on-chip.

Google Cloud added that each 8i server contains double the CPU hosts, in the form of its in-house Axion CPUs based on Arm designs. These orchestrate tasks such as data input/output within the AI stack, with Google Cloud claiming they improve overall inference performance.

TPU 8i also benefits from Boardfly, a brand new network topology that Google Cloud announced today at Google Cloud Next 2026. This replaces the ‘3D torus’ topology used in previous generations and in 8t, which connects thousands of chips for throughput for training workloads.

Instead, Boardfly focuses on a new chip connection with increased ports, leading to a lower overall network diameter – the longest efficient distance between two nodes within a network. After implementing Boardfly for TPU 8i, Google Cloud recorded up to a 50% latency improvement.

“One of the things that we've had to realize is that the way that we were connecting the chips together, the network topology… our default way of connecting them together didn't support latency, it supported throughput, it supported bandwidth,” explained Amin Vahdat, SVP and chief technologist, AI and Infrastructure at Google.

“It was really good at getting large amounts of data through. But what you really care about in the age of agents is the latency, the minimum time it takes to get the data.”

Vahdat added that Google has been working on TPU 8t and 8i internally for two years, before AI agents were even widely adopted or understood, based on internal discussions with DeepMind about the direction AI was headed and where future bottlenecks could lie.

The move to split its TPU line-up puts Google in a similar position to the likes of AWS, which also segments its application-specific integrated circuits (ASICS), Trainium and Inferentia.

“For us, it’s a natural evolution,” said Google Cloud CEO Thomas Kurian in a briefing for the assembled press.

“We've been working on these chips and systems for multiple years now. And so when AI came along with particularly generative AI became, you know, widespread, we felt that people would want systems that were more optimized for training, and separately, systems that were more optimized for inference.”

AI hypercomputer innovations

In addition to its first-party hardware, Google Cloud relies on its extensive in-house infrastructure and cloud architecture to serve customers with enterprise AI capabilities.

It calls this the ‘AI hypercomputer’, which encompasses Google Cloud’s specialized compute, storage, networking, and software.

Last year, Google Cloud announced it would give customers access to its machine learning runtime Pathways. This is a Google DeepMind system which connects TPU pods and splits inference tasks across them in real time to efficiently act as a unified compute cluster for multimodal AI.

This year, Google Cloud has unveiled Virgo Network, its new AI data center fabric optimized for both high-capacity data access and massive, low latency throughput across TPU pods. Virgo Network is a flat,

Within a single data center, Virgo Network also allows up to 134,000 TPU 8t chips to work as a single fabric, a feat that is notable for organizations with data residency requirements for training models on sensitive data.

Combining Virgo Network, Pathways and Google DeepMind’s open source Python library JAX, Google Cloud is now able to connect more than one million TPU 8t chips across multiple data centers in a single training cluster for frontier model training.

For TPU 8i, Google Cloud said Virgo Network can lower latency on a network without other traffic by 40% compared to Ironwood.

On top of this, Virgo Network was designed to give network engineers constant observability over compute clusters, enabled by sub-millisecond telemetry and automated identification of nodes that are having a detrimental impact on AI training and inference jobs.

Energy optimization and reliability

Alongside their performance improvements, TPU 8t and 8i are Google Cloud’s most efficient and reliable chips to date.

Google Cloud stated that through chip-level innovations and new approaches such as its fourth-generation liquid cooling, it achieves double the performance per watt with the new TPU family compared to Ironwood.

This could be a key metric as customers look to eat the costs of running an increasing number of AI agents in their cloud environment, as it means more tasks can be achieved with the same energy cost.

Microsoft has predicted that 1.3 billion agents will be in operation by 2028, which will require not just a jump in infrastructure for inference, but also successive jumps in energy efficiency to make these workloads economical to run.

TPU energy improvements aren’t just good for AI workloads, either. Vahdat said that Citadel Securities, one of the world’s largest market making firms, has used TPUs to reduce the costs of its trading systems by 30%.

Kurian stated that Google Cloud has dedicated years to optimizing its TPUs, including early focuses on areas that weren’t seen as a market priority until recently.

“For example, we designed these systems to be super efficient in terms of how much power they use, because we felt that power efficiency would become a constraint as people continue to scale both training and inference.

“And the response from customers to both at and AI has been very rewarding to see, and it validates that people increasingly are specializing how they’re deploying AI infrastructure, whether it’s for training or inference.”

One week after announcing it’s putting its UK Stargate plans on ice, another facility that was expected to rent capacity to OpenAI has instead found itself an alternative customer.

According to an announcement from neocloud provider Nscale, which owns both the data centers in question, it will add more than 30,000 Nvidia Rubin GPUs to its 230MW campus in Narvik, northern Norway. These will then be rented by Microsoft, starting in 2027.

This is a far cry from the original plans for the data center, however.

In July 2025, Nscale, OpenAI, and Aker announced they would be creating the first AI gigafactory in Europe and the first European facility in the OpenAI for Countries program, in the form of Stargate Norway.

It was reported at the time that there would be an initial 100,000 Nvidia Rubin GPUs installed by the end of 2026 for the project, with an initial 230MW of capacity. This was anticipated to scale to 520MW over time with the numbers of GPUs also increasing.

OpenAI CEO Sam Altman said: “I've always said we'd love to bring Stargate to Europe if the conditions are right, and we think we've found that in Narvik with clean, affordable energy, ideal climate, and great partners in Nscale and Aker. Stargate Norway will help provide the compute power to drive the next wave of AI breakthroughs and economic progress for Europe, in Europe.”

A statement from OpenAI to ITPro today, however, indicated that OpenAI’s involvement is rather less direct than it appeared in July.

An OpenAI spokesperson told this publication: “We are moving ahead with our plans in Norway. Microsoft is an important partner in our network and we will work with them to access compute in Norway just as we already do in other parts of the world.”

Microsoft’s involvement wasn’t mentioned in the original Stargate Norway announcement, however, and there have been other changes in the set-up, too. In March 2026 it was announced Aker and Nscale would be rolling up the joint venture that established the Narvik facility fully into the neocloud provider. At the same time, Aker became the largest shareholder in Nscale.

The news that OpenAI won’t be directly renting capacity from Nscale in Norway – with the Stargate Norway branding seemingly being dropped at the same time – comes only one week after a similar move in the UK.

On 9 April, a spokesperson told ITPro that it was pausing its Stargate UK plans – which also revolved around renting Nvidia GPUs from Nscale – due to the regulatory environment and energy prices in the country. OpenAI declined to comment on whether similar considerations are in play with Stargate Norway, although energy prices are notably lower in the nordics.

In the meantime, Nscale may have found another tenant for the London facility. According to Bloomberg, Google will rent the capacity left open by OpenAI’s decision to put the brakes on its investment.

Microsoft’s incessant efforts to push users onto Windows 11 may have taken a major blow thanks to Google’s new ChromeOS Flex launch – and it could be the perfect remedy for cost conscious businesses dithering over upgrading.

The tech giant has teamed up with electronics refurbishing company, Back Market, to let Mac and Windows users install the operating system on their ageing devices by simply inserting a USB and converting the device OS.

It’s something of a masterstroke by Google and comes at a critical time in the enterprise and consumer hardware markets. With Windows 10 support ending in October last year, millions of users worldwide are facing a ticking timebomb with upgrades.

Microsoft has been pestering users to make the shift for years at this point, even though upwards of 500 million device owners have a PC that’s too old to even upgrade.

Of course, you can continue using Windows 10 through the Extended Security Updates (ESU) program, but as ITPro previously reported, that could prove costly.

Research from Nexthink in September last year projected that costs for maintaining the ageing operating system would exceed $7.3 billion globally. Those staggering numbers paint a grim picture for those hoping to stick with the highly popular OS.

Combine the Windows 11 conundrum with the fact that hardware prices are expected to skyrocket this year, and the situation is even worse for enterprises and consumers alike.

Neither appear too keen on upgrading while RAM prices are astronomical, recent research shows. Given that confluence of issues, Google could be onto a winner with ChromeOS Flex.

What to expect with ChromeOS Flex

ChromeOS Flex is free to use for anyone, but it’s primarily aimed at use among academic institutions and businesses.

The tech giant has an enterprise-grade option, which will undoubtedly be a viable choice for smaller businesses looking to eke out device lifecycles – a tactic that will become more common amid price hikes.

The benefits here are tantalizing for IT admins, many of whom will be rubbing their hands at the prospect of managing an operating system, devices, and apps all within one ecosystem. If you’re a Google Workspace shop, it’s the perfect choice.

To get your hands on it, Back Market USB kits are available for as little as $3 (although they appear to be sold out as of 9 April). Alternatively, you can download ChromeOS Flex for free online.

If you’ve got an ageing device that’s struggling, the cloud-based operating system could be the one for you. A quick browse through the company’s sprawling list of certified models underlines the scope of the scheme. A total of 640 devices ranging back well over a decade are covered by the scheme.

Diving into the list, I found some of ITPro’s old archived reviews, including the Dell Latitude 3440 (2014) edition and Microsoft Surface Pro 3 devices are certified. If you’re somehow still clinging onto ancient tech, the benefits here are huge.

Sustainability gains

The Windows 11 push isn’t just a financial burden for enterprises and consumers, research shows it has a serious environmental impact. Research last year, for example, warned of a looming torrent of e-waste due to the changeover.

Many older devices which can’t run Windows 11 have ended up on the scrapheap. It’s a waste of perfectly good hardware and a slap in the face for consumers and businesses that are conscious of, or legally obligated to, improving sustainability.

Figures from businesswaste.co.uk found the Windows 10 end of life transition will generate around 12,805,100 kg of metals – equivalent to more than 8,000 cars.

Google’s own figures suggested that an estimated 61.3 metric tons of e-waste will be thrown away in 2026 alone. For perspective, that’s the equivalent of around 168 Empire State Buildings.

ChromeOS Flex could go a long way to help cutting down those eye-watering figures, providing there’s uptake, of course. Google claims converting around 10,000 devices to the operating system could prevent 50,000 pounds of e-waste from ending up on the scrapheap.

A tough time for Microsoft

This isn’t a direct dig at Microsoft. Indeed, Google is keen to emphasize that some older Apple devices can also be converted to ChromeOS Flex.

It undoubtedly comes at a difficult time for the tech giant though. Microsoft has bent over backwards to push users to its newer operating system, I begrudgingly upgraded my personal device some time ago.

All things considered, Microsoft hasn’t done a very good job of encouraging users to make the shift. Upwards of one billion devices worldwide were still running Windows 10 as of December 2025.

Admittedly, we can’t all cling onto the operating system we love. If that was the case then I’m sure a few of us would still be working on Windows XP. The back and forth between consumers and the tech giant has made it irritating, however.

Deadlines were set, then pushed back. Confusion over the prospect of continued support was rampant, and with the exorbitant costs associated with sticking to Windows 10, the company was clearly strong-arming businesses into making the shift.

Microsoft has also faced pushback from consumers over its incessant shoehorning of AI features within Windows – an issue it’s vowed to address in recent weeks.

The launch of the MacBook Neo adds insult to injury on this front as well. With the launch of the price-conscious laptop series, even the most ardent Windows users may be wooed, and I rank myself in this demographic.

Microsoft has tinkered with Windows 11 too much and forced us all to make hard choices with our beloved hardware and operating systems. Google is offering us a lifeline.

Anthropic has announced a major expansion of its partnership with Google and Broadcom, which will see it rely more heavily on Google’s tensor processing unit (TPU) chips to meet computational demand.

The frontier lab announced it has signed an agreement to use “multiple gigawatts of next-generation TPU capacity”, which would come online from 2027 onwards.

It’s unclear whether the specific TPU models deployed as part of the deal will be TPU v7 Ironwood, announced at Google Cloud Next 2025, or an unspecified future model. ITPro contacted Anthropic for clarification but did not receive a reply at time of publication.

The deal will see Anthropic make use of a further 3.5 GW of TPU compute capacity, per CNBC reporting. In its announcement, the firm explained the “vast majority” of this would be physically sited in the US.

Anthropic uses a mix of TPUs, Amazon Trainium chips, and Nvidia GPUs to fulfil its training and inference requirements. Amazon is still its primary training partner and cloud provider, with the firms having worked together on the million-chip Project Rainier cluster throughout 2025.

In October 2025, it had announced plans to expand its TPU footprint by up to one million chips throughout 2026, the equivalent of over a gigawatt of compute capacity.

Google says its TPUs, which Broadcom co-designs and Taiwan Semiconductor Manufacturing Company (TSMC) manufactures, offer highly competitive teraflops per watt compared to competing chips.

Unlike generalist GPUs, TPUs are application-specific integrated circuits (ASICs) that were designed specifically with machine learning and AI in mind.

“This groundbreaking partnership with Google and Broadcom is a continuation of our disciplined approach to scaling infrastructure: we are building the capacity necessary to serve the exponential growth we have seen in our customer base while also enabling Claude to define the frontier of AI development,” said Krishna Rao, CFO of Anthropic.

“We are making our most significant compute commitment to date to keep pace with our unprecedented growth.”

Anthropic's compute demand rising with revenue

Like its competitor OpenAI, Anthropic is looking to rapidly expand its compute capacity to meet rising customer inference demand driven by its popular tools such as Claude Code, Claude Cowork, and agentic tool use within cloud platforms like AWS Bedrock and Google Vertex AI.

In recent weeks, Anthropic announced new session limits for free, Pro, and Max subs for all weekday Claude use from 1pm to 7pm GMT. The move prompted criticism from users on platforms such as Reddit, with some reporting hitting rate limits within 30-35 minutes of vibe coding with Claude Code.

On 4 April, Anthropic removed allowances for third-party harnesses from the standard Claude subscription, effectively locking popular tools like OpenClaw behind a paywall.

In the same announcement as the new compute deal, Anthropic announced its run-rate revenue has surpassed $30 billion, up from $9 billion at the end of 2025. The firm added that over 1,000 business customers now spend more than $1 million on Anthropic annually.

Per reporting from Forbes, Anthropic reports its revenue as a gross figure, inclusive of revenue accrued from hyperscaler customers accessing its models.

This results in a higher top-line revenue figure, as it doesn’t include the cut Google Cloud and AWS take each time a user pays for Claude access via their platform.

ITPro asked for clarification on the precise way in which Anthropic calculates its revenue and if the $30 billion figure excludes hyperscaler revenue sharing, but did not receive a reply by the time of publication.

FOLLOW US ON SOCIAL MEDIA

Google has warned that “Q-Day”, the point where a quantum computer is powerful enough to crack current encryption techniques, could come as soon as 2029.

The new timeline comes in light of progress in quantum computing hardware development, quantum error correction, and quantum factoring resource estimates.

Quantum computers pose a particular threat to encryption and digital signatures, according to the company, and attackers are already stealing encrypted data they believe they'll be able to decrypt some time in the future.

Digital signatures, meanwhile, will also be under threat, requiring a transition to post-quantum cryptography (PQC) before the development of Cryptographically Relevant Quantum Computers (CRQCs).

"As a pioneer in both quantum and PQC, it’s our responsibility to lead by example and share an ambitious timeline," warned Heather Adkins, VP, security engineering, and Sophie Schmieg, senior staff cryptography engineer

"By doing this, we hope to provide the clarity and urgency needed to accelerate digital transitions not only for Google, but also across the industry."

Most firms aren't prepared for Q-Day

Estimates of when Q-Day is likely to arrive have varied widely, but most experts have suggested dates in the mid-2030s.

In 2025, the UK’s National Cyber Security Centre (NCSC) urged organizations to carry out PQC migration activities by 2031, and to have completed migration to PQC of all systems, services, and products by 2035.

NCSC CTO Ollie Whitehouse said at the time this transition will be a “complex change programme that makes fixing the Millennium Bug look easy”.

Research from Bain & Company in January this year found that 90% of organizations don’t yet have systems in place to defend against quantum security threats.

This was despite the fact that 71% expected quantum-enabled attacks within five years, with a third predicting them within three.

Notably, only one-in-ten said they had a roadmap in place to address the risks, with most waiting to see what happens and hoping a third party solves the problem first.

Google moving fast on PQC

Alongside its Q-Day warning, Google has announced that it's integrating PQC digital signature protection using ML-DSA into Android 17.

The tech giant said the proactive move aligns with recommendations outlined by the National Institute of Standards and Technology (NIST).

In a blog post confirming the move, Eric Lynch, Android product manager, and Dom Elliot, group product manager at Google Play, said the update aims to "establish a new, quantum-resistant chain of trust".

"This chain of trust secures the platform continuously—from the moment the OS powers on, to the execution of applications distributed globally. Android is swapping today’s digital locks for advanced encryption to help enhance the security of every app you download — no matter how powerful future supercomputers get.”

FOLLOW US ON SOCIAL MEDIA

Google has announced the launch of a new AI-powered dark web monitoring service, allowing enterprise security teams to keep tabs on emerging threats and exposed information.

The new dark web intelligence capabilities, offered through Google Threat Intelligence, use Gemini to analyze threat actors activities relevant to particular organizations.

The new AI capabilities aim to streamline threat intelligence operations, according to Google, removing the need for enterprises to manually input and update keywords on an as-they-go basis.

"Our new dark web intelligence capability uses Gemini to autonomously build an organizational profile that is specific to your business operations and mission, automatically adjusting as these are modified," the firm explained in a blog post penned by product manager Brandon Wood and product marketing manager TJ Alldridge.

The duo added that the new platform will "evolve" based on how teams react to intelligence flagged by Gemini, helping to "ensure the system's context is current".

Initial testing of the dark web intelligence tool found it's capable of analyzing "millions of daily external events" with 98% accuracy. The tech giant is keen to emphasize that this accuracy will be a key differentiator for enterprises, helping to remove false positives.

Similarly, Gemini will provide "reasoned answers" that provide a deeper level of context for security teams, explaining the "why" and "how" of potential threats.

"We are giving defenders their time back and ensuring they maintain the intelligence high ground in an increasingly automated threat landscape," the company said.

"Customers now have the ability to translate vast dark web data into precise, relevant insights delivered at the speed of AI with the goal of enabling your team to think and act faster than the agent-enabled adversary."

The launch of the service comes just weeks after support for Google's consumer-focused dark web report feature officially ended. The company confirmed plans to scrap the tool in December, claiming it was no longer effective.

At the time, Google said it planned to focus efforts on tools that give users "clearer, actionable steps" to protect information.

Google is all-in on agentic security

The launch of the dark web intelligence service marked one of several AI-related security announcements for Google at the 2026 RSAC Conference in San Francisco this week.

The tech giant announced plans to introduce new "agentic automation" features for Google Security Operations. This will enable enterprises to deploy agents aimed at autonomously identifying and responding to potential threats.

"Google Security Operations users can embed agents, including our Triage and Investigation agent, directly into workflows to accelerate mean time to respond," the company said in a blog post.

This particular agent is capable of autonomously investigating alerts to provide evidence for security operations teams, Google said, providing users with "verdicts and comprehensive explanations".

Elsewhere, Google Security Operations will be able to build and deploy their own security-focused agents from April through remote model context protocol (MCP) server support.

This will remove the need for users to host their own security operations MCP server client, according to Google. General availability is scheduled for early April.

A host of big tech firms have handed over $12.5 million in funding to advance open source security and try to eliminate "AI slop" bug reports.

Firms including OpenAI, Anthropic, AWS, Google, Microsoft, and GitHub have pledged funding for Alpha-Omega and the Open Source Security Foundation (OpenSSF), both security initiatives within the Linux Foundation.

The aim is to develop long-term, sustainable security solutions that support open source communities worldwide.

The move comes as open source maintainers contend with an unprecedented number of security reports, many of which are generated by automated systems.

Mark Ryland, director of the Office of the CISO for AWS, said these AI-generated reports are overwhelming their ability to review them.

"Many of the reports are of very low quality — a reality given rise to the new industry term 'AI slop'," he said. "Many projects have already elected to put guidelines in place for AI submissions, while others have shut down upstream contributions entirely to prevent a flood of AI-generated pull requests."

Closer ties with open source maintainers

The new investment will allow Alpha-Omega and OpenSSF to work directly with maintainers and their communities to make emerging security capabilities accessible, practical, and aligned with existing project workflows.

“Grant funding alone is not going to help solve the problem that AI tools are causing today on open source security teams,” said Greg Kroah-Hartman of the Linux kernel project.

“OpenSSF has the active resources needed to support numerous projects that will help these overworked maintainers with the triage and processing of the increased AI-generated security reports they are currently receiving.”

The GitHub Secure Open Source Fund is adding an additional $5.5 million in Azure credits and funding to provide training and expertise.

GitHub Security Lab, meanwhile, is improving the security advisory experience on GitHub and Private Vulnerability Reporting (PVR) features, with an eye on reducing the burden of low-quality reports and helping maintainers manage increased volume.

Google, meanwhile, will provide AI-powered tools like Big Sleep and CodeMender from Google DeepMind – already used to protect the company's own systems. It's also extending research initiatives like Sec-Gemini to open source projects.

“Our commitment remains focused: to sustainably secure the entire lifecycle of open source software,” said Steve Fernandez, general manager of OpenSSF.

“By directly empowering the maintainers, we have an extraordinary opportunity to ensure that those at the front lines of software security have the tools and standards to take preventative measures to stay ahead of issues and build a more resilient ecosystem for everyone.”

AI slop reports are skyrocketing

Concerns about AI slop bug reports have been voiced by a number of organizations, including the Python Software Foundation.

Developers behind cURL, an open source command line interface (CLI) tool which allows developers to transfer data, recently shut down its bug bounty scheme in response to a growing number of slop reports.

As ITPro reported at the time, lead maintainer Daniel Stenberg said the current volume of submissions is placing a “high load” on the security team.

FOLLOW US ON SOCIAL MEDIA

A high-severity vulnerability in Google’s Gemini Live in-browser AI assistant created significant privacy and security risks for Chrome users.

Gemini Live runs in a side panel, summarizing web content in real time and handling tasks like automated actions, and requires deep, privileged access to the browsing environment, including media devices and files.

The flaw, tracked as CVE-2026-0628, allows malicious extensions with basic permissions to 'hijack' the new feature, giving attackers access to webcams, microphones, and private files.

Palo Alto Unit 42 researchers said this vulnerability could have allowed malicious browser extensions with only basic permissions to escalate privileges and access the victim's camera and microphone without consent.

Thereafter, a threat actor could take screenshots of any website and access local files and directories.

“Today’s agentic browsers can act on your behalf — researching, reasoning and taking action without direct user input. While this can deliver meaningful productivity gains, in the absence of enterprise-grade controls these tools can take autonomous actions beyond IT oversight," warned Anupam Upadhyaya, SVP, product management, Prisma SASE, Palo Alto Networks.

"By inheriting a user’s browser session and accessing screens, files, cameras and microphones, agentic browsers can expand the attack surface through prompt manipulation and weakened web isolation, creating security and accountability gaps enterprises haven’t faced before."

How the Gemini Live flaw works

Researchers found that an extension with access to a basic permission set through the declarativeNetRequests API could have enabled an attacker to inject JavaScript code into the new Gemini panel.

The API allows extensions to intercept and change properties of HTTPS web requests and responses – needed for legitimate purposes, such as allowing AdBlock to stop requests that could lead to privacy-undermining ads.

However, when loaded within the new side panel, rather than a standard tab, a flaw emerges in the ability to intercept and change properties of hxxps[:]//gemini.google[.]com/app.

As a result, attackers could run arbitrary code at hxxps[:]//gemini.google[.]com/app under the new Gemini browser panel. As a privileged component of the browser itself, code running within the Gemini panel could access capabilities unavailable to the extension that injected the code initially.

Attackers could start the camera and microphone of the browser without asking for user consent, reach local files and directories of the underlying operating system, take screenshots of tabs showing any website that serves over HTTPS, and hijack the panel into carrying out a phishing attack.

"Innovation can’t come at the expense of security," said Upadhyaya. "If organizations choose to deploy agentic browsers, they must treat them as high-risk infrastructure, with runtime visibility, enforced policy controls and hardened guardrails built in from day one. Anything less invites compromise.”

Unit 42 shared the issue with Google in October via coordinated vulnerability disclosure, with Google releasing a fix in early January.

FOLLOW US ON SOCIAL MEDIA

Google is building a data center where the only water used will be in the kitchens, the company claims.

The tech giant announced a new data center under construction in Wilbarger County, Texas will use air cooling instead of water, with "clean" power added to the grid even before the data center comes online.

Google’s announcement comes amid a wider pushback against the impact of data centers – in particular in Texas – which are being built in a rush to meet growing demand sparked by AI.

Common concerns include increasing energy prices through high consumption and the fact these sites use significant volumes of water.

President Trump this week told data center building companies they need to start sorting out their own power supplies to avoid energy prices continuing to increase for people.

Google said that, once complete, the new data center will use "advanced air-cooling technology, limiting water consumption to only critical campus operations like kitchens."

Indeed, Google said the aim is to replenish more local water than it consumes and is working with local water charities. Google's head of data center public affairs for Texas, Andrew Hart, said the company was "proud" of its close collaboration with local leaders on this front.

"Texas is at the center of the world’s AI leadership, and our expansion into Wilbarger County is a testament to the state's incredible talent and infrastructure," he said in a statement.

The industry doesn't need to use water for cooling, but often does in a bid to cut power use. Evaporative cooling, for example, consumes vast amounts of water.

Alternatives include air cooling, which uses liquid carbon dioxide, while other operators use liquid immersion and closed loop practices. In short, there are options.

Google’s new data center power plans

When it comes to power, Google is working with local energy generation companies Wilbarger County through power purchase agreements (PPAs) to boost clean energy on the Texas grid.

The tech giant claimed it will add more than 7,800 megawatts of net-new energy generation and capacity.

Indeed, the new site will be co-located next to "clean" power plants built and operated by AES, which will also run retail and other services at the campus. Google said the co-location is a "key strategy" for easing demands on the grid while speeding along buildout of data centers.

The aim is for the project to be "power first" with the plants coming online ahead of the data center. That, along with $30 million announced in November to help local residents hit by rising energy costs, should help keep power affordable, the company said.

"We are proud of our long-standing partnership with Google and thrilled to bring them this advanced stage ‘power first’ clean co-located project,” said Kleber Costa, Chief Commercial Officer, AES US Renewables.

"By capitalizing on AES’s extensive development expertise and years of hard work in collaboration with the exceptional local leaders and communities, we are bringing to Google a site that is fully ready for construction, with land and interconnection agreements, as well as co-located generation under 20-year PPAs."

Texas pushback

Google's efforts to ensure its data center won't use much water and will generate clean energy come amid pushback against these facilities, largely focused on those two issues.

That's particularly true in Texas. In central Texas, the city of San Marcos recently rejected a data center project in the area after resident complaints — and a local judge — raised concerns about water use. As one activist said: "We are out of water."

Two other similar infrastructure projects remain in the works in the nearby area, with one in nearby Round Rock approved despite "hours of public opposition" at a hearing on the issue.

It's not just about water and energy, though. In north Texas, a "city sized" data center project faces criticism for the risk it causes to local Dinosaur Valley State Park, home to well-preserved dinosaur tracks, though water and power use are concerns too.

In Hood County, residents are also concerned about construction-related noise. As one resident noted to local reporters: "I get to pay property taxes to be audibly assaulted."

City and state authorities are quick to note that these projects not only bring in jobs but also tax revenues, important for areas facing restricted and deficit budgets.

FOLLOW US ON SOCIAL MEDIA

Google has added code validation capabilities to its Gemini CLI coding extension Conductor in a move aimed at tackling some of the challenges of using AI for software engineering.

The tech giant first unveiled the Conductor extension back in December, aiming to create context-driven development by shifting projects out of chat logs and into markdown files. Now, it's adding a new feature to help coders verify their work.

"Our new Automated Review feature allows Conductor to go beyond planning and execution into validation, generating post-implementation reports on code quality and compliance to the guidelines you’ve defined," the company said in a blog post.

Once the coding agent finishes its tasks, Conductor will generate a report where it reviews code, ensures everything meets user-set guidelines and compliance requirements, and runs a basic security review to look for critical vulnerabilities before code is merged.

This includes probing for hardcoded API keys or personal information that could leak, according to Google. Beyond that, Conductor includes test-suite validation.

"Instead of relying on manual execution, Conductor integrates your entire test suite directly into the review workflow," the post added.

"It runs all relevant unit and integration tests, then incorporates the results and coverage data into the final report to provide a unified view of whether the new code actually functions as intended within your existing ecosystem."

Google eyes security gains with Gemini CLI

The aim is for Automated Review to give developers detailed information on what needs improvement or addressing, offering a clear workflow that includes the exact file path to fix issues.

"This level of detail ensures that 'agentic' development doesn't mean 'unsupervised' development," the blog post noted.

"Instead, it creates a workflow where the AI provides the labor and the developer provides the high-level architectural oversight, backed by automated verification."

Google suggested more features were on the way, noting the latest updates are evidence of the company's aim to make "AI development safe, predictable and architecturally sound."

Trust but verify

The rise of AI coding tools has sparked concerns about errors introduced by agents and other automation tools – especially with code that looks correct and production ready, but contains security risks.

Indeed, AI-generated code is already the cause of one in five breaches, according to one survey.

Adding another layer of verification and supervision could be critical in stopping disastrous flaws before they cause havoc – especially given that developers are now falling foul of these on a frequent basis.

A recent survey found nearly half of software developers don't check AI-generated code, in part because it's harder to review code produced by AI than humans.

Nigel Douglas, head of developer relations at Cloudsmith, said while the feature could prove useful, it won’t address all the challenges presented by AI-generated code.

"An AI coding CLI without automated reviews is like a chainsaw without an ‘off’ button, but, unfortunately, these changes focus only on the code that’s been generated –completely skipping the upstream components it’s pulling in," he said.

"If an AI coding assistant suggests a package that doesn’t exist or has already been infected with malware, you’ll end up shipping vulnerabilities far faster than you can catch them.

“Peer reviews can’t work the way they’ve always worked when LLMs can generate thousands of lines of functional code in minutes. No human can – or should – read that fast.

FOLLOW US ON SOCIAL MEDIA

Unilever has signed a five-year deal with Google Cloud under which it will use the hyperscaler's AI, data, platform, and marketing capabilities to accelerate business transformation.

The aim is to build new capabilities in brand discovery, measurement, and AI-augmented marketing for Unilever’s global portfolio, which includes brands such as Dove, Vaseline and Hellmann's.

"Technology has moved to the core of value creation at Unilever,” said Willem Uijen, chief supply chain and operations officer at Unilever.

“As brands are increasingly discovered and chosen in environments shaped by AI, we must lead this shift," Uijen added.

"This collaboration with Google Cloud sets a new level in how technology can power commerce and growth in the fast-moving consumer goods industry, ensuring Unilever is agile, fit for the future, and equipped to unlock value at every level of the company.”

By migrating its integrated data and cloud platform to Google Cloud, Unilever plans to build an enterprise-wide, “AI-first digital backbone”.

The consumer goods giant said it hopes to generate demand faster, turn data into actionable insights, and respond to market shifts in a more agile way.

Unilever targets 'agentic commerce' gains

The deal will focus on three core pillars: agentic commerce and marketing intelligence, an integrated data and cloud foundation, and advanced AI.

Under the agreement, the two companies will work together to build next-generation marketing capabilities across brand discovery, conversion, and measurement to deal with shifts in technology and consumer habits.

Meanwhile, Unilever will transition key enterprise applications and data platforms to Google Cloud, creating a connected environment for scalable AI deployment across the value chain.

Notably, Unilever said the partnership will help fast-track its adoption of emerging technologies, combining Unilever's expertise with Google's AI capabilities.

“In partnering with Unilever as it boldly reimagines its business processes, we are not just modernizing legacy systems; we are deploying our advanced models, such as Gemini, to create a system of intelligence that reasons, learns, and acts, said Tara Brady, president, EMEA, at Google Cloud.

"This will set a new standard for agility and consumer engagement in the [consumer packaged goods] CPG sector.”

Unilever is one of the world’s largest suppliers of beauty and wellbeing, personal care, home care, and foods products, and has undergone a significant transformation over the last five years.

In 2023, for example, the company signed a deal with Accenture that enabled it to become a cloud-only operation. The move saw the company switch to running most of its cloud estate on Microsoft Azure.

Elsewhere, the consumer goods giant has focused heavily on a range of AI projects through its Horizon3 Labs, and already has 500 AI applications in place.

FOLLOW US ON SOCIAL MEDIA

State-backed threat actors from CRINK nations have come to rely on large language models (LLMs) as “essential tools” for researching and targeting victims, according to a new report.

The latest AI Threat Tracker report from Google Threat Intelligence Group (GTIG), produced in collaboration with Google DeepMind, details the numerous ways threat groups are already using AI to plan and carry out attacks.

Advanced persistent threat (APT) groups were tracked using Google’s own Gemini family of models to conduct targeted research on potential victims, probe vulnerabilities, and create tailored code and scripts.

For example, the China-based APT Temp.HEX was found using Gemini to file information on individual targets in Pakistan.

The as-yet-unattributed APT UNC6148 also used Gemini to seek out sensitive information tied to victims, such as email addresses and account details, as the first step in a targeted phishing campaign on Ukraine and the wider defense sector.

In response, Google disabled the assets associated with both groups. Other incidents saw attackers use public AI models to more directly fuel attack campaigns.

Iranian-backed groups such as APT42 were observed using Gemini and other AI models to research potential victims, then craft convincing phishing emails based on target biographies.

That same group was observed using Gemini to translate local languages as well as regional references and phrases.

North Korea-backed groups seized headlines throughout 2024 and 2025, as hackers infiltrated IT departments of major organizations including KnowBe4 with fake addresses and identities.

In the report, the North Korean-backed group UNC2970 was found using Gemini to plan attacks on cybersecurity defense companies and map job specifications.

AI-enhanced malware is gathering steam

The report also noted the growing risk presented by malware that uses AI to achieve novel capabilities such as preventing network detection.

HONESTCUE malware, for example, has been found to use API calls to Gemini to generate ‘stage two’ code. This is used to download and execute additional malware directly in the memory of target systems using CSharpCodeProvider, a legitimate .NET class for executing C# code.

Because the Gemini-produced code executes the secondary malware directly in memory, HONESTCUE infects target systems without leaving telltale artifacts on the victim’s disk.

Though the malware hasn’t been linked to specific attack campaigns to date, GTIG researchers said they believe its developer is a single threat actor or small group testing the waters for future attacks. This is backed up by evidence HONESTCUE has been tested on Discord.

Another example can be found in COINBAIT, a phishing kit created by the APT UNC5356 that shows signs of having been created using the vibe coding platform Lovable.

GTIG has previously warned that while AI malware is still nascent, it’s developing quickly. In the latest report, authors noted that while no “paradigm shift” has yet been unlocked by APTs, their exploration of malicious AI is ongoing and the technology will play a growing role in every stage of the attack lifecycle.

On the other hand, researchers discovered that threat actors are passing off jailbroken public AI models as handmade offensive tools.

For example ‘Xantharox’, a dark web toolkit advertised as tailor-made offensive AI toolset, is actually powered by open source AI tools such as Crush and Hexstrike AI via model context protocol (MCP), as well as public AI models like Gemini.

Threat actors are stealing API keys to enable this hidden activity, with GTIG warning organizations with cloud and AI resources are at risk. Users on platforms such as One API and New API, often those in countries with regional AI censorship, are also targeted for API key harvesting.

Model extraction puts AI developers at risk

Researchers also observed instances of APTs performing ‘model extraction’, in which attackers use legitimate access to frontier models such as Gemini to help train new AI and machine learning (ML) models.

Generally, attackers use an approach known as knowledge distillation (KD) in which a ‘student’ AI model is trained on the answers to specific questions based on the exemplar answers of the pre-existing AI model.

This can result in models with advanced capabilities such as frontier reasoning but none of the guardrails present in public AI models like Gemini. In the future, threat actors could then use

GTIG tracked over 100,000 prompts intended to expose and replicate Gemini’s reasoning capabilities in non-English languages, which were automatically counteracted by Google’s systems.

“Google’s latest AI Threat Tracker marks a specific turning point: we are no longer just worried about bad prompts, but the industrial-scale extraction of the models themselves,” wrote Jamie Collier, lead advisor in Europe at Google Threat Intelligence Group, in a LinkedIn post marking the launch of the report.

Google DeepMind and GTIG blocked attempts at model extraction throughout 2025, noting that the attacks were launched by private companies and researchers around the world rather than APTs.

Distilling secondary models from Gemini is a violation of Google’s terms of service and is considered theft of intellectual property (IP). The hyperscaler recommended organizations that provide AI models as a service should closely observe API access for signs of model extraction.

FOLLOW US ON SOCIAL MEDIA

The European Commission (EC) has officially approved Google’s proposed acquisition of cloud security firm Wiz.

In a statement confirming the move, the Commission said it concluded the deal would not raise competition concerns across the European Economic Area (EEA).

Teresa Ribera, European Commissioner for Competition, said the probe concluded that regardless of the acquisition, Google “stands behind Amazon and Microsoft” in terms of market share in areas such as cloud infrastructure.

“Based on its market investigation, the Commission found that there are several credible competitors that customers could switch to if Google were to bundle Wiz's multi-cloud security platform with its existing products, or in case Wiz's platform no longer worked with clouds other than Google's,” the EC said in a statement.

The EC also investigated whether Google would gain access to “commercially sensitive data” pertaining to cloud competitors that integrated with Wiz solutions, again confirming there was no risk.

“The market investigation confirmed that the data that Google would obtain as a result of the transaction is not commercially sensitive, and can generally be accessed by other security software companies,” the Commission said.

The confirmation marks the latest hurdle cleared by Google in its pursuit of the cloud security firm. In November last year, a US antitrust review approved the deal.

Google announced plans to acquire the cloud security company in March 2025 in a deal worth $32 billion.

The deal marked the largest acquisition in the company’s history, more than double the fee for its deal to acquire Motorola in 2012.

At the time, Google said a key factor behind the acquisition lay in addressing “two large and growing trends in the AI era” - cloud security and multi-cloud adoption.

Wiz was founded in 2020 in Israel before moving operations to the US. The startup reported rapid growth in the years preceding the deal. At the time Google announced the acquisition, headcount at the firm stood at 900 and the firm was valued at $12 billion.

FOLLOW US ON SOCIAL MEDIA

Google CEO Sundar Pichai has hailed the company’s growing strength in the AI race after its cloud unit recorded significant financial growth.

Alphabet’s 2025 financial year ended on a bang, with revenue surpassing $400 billion for the first time and Google Cloud revenue reaching $17.7 billion – a 48% increase compared to the same period last year.

Pichai claimed growth in this division is a clear signal of the company’s ongoing success in generative AI.

"Overall, we’re seeing our AI investments and infrastructure drive revenue and growth across the board," he told investors.

Google Cloud has been the tip of the spear for the tech giant in the AI race, leading on the roll-out of its flagship Gemini model range and development of in-house hardware for AI inference.

With Gemini, the growth signals are clear. Pichai revealed Gemini 3 Pro has recorded the “fastest adoption of any model in our history”.

Indeed, the AI app now boasts more than 750 million active monthly users, marking an increase from 650 million in the previous quarter. Pichai also told investors the company secured eight million paid seats for Gemini Enterprise packages in the space of four months.

Google Cloud is gaining ground, albeit slowly

Lee Sustar, principal analyst at Forrester, said Google Cloud’s revenue growth is a key indicator that Alphabet is gaining ground on cloud competitors such as Amazon Web Services (AWS) and Microsoft.

“Google Cloud's quarterly revenue jump of 48% over the same period a year earlier is decisive evidence that it is a full-blown enterprise challenger to AWS and Microsoft Azure,” he said.

The hyperscaler still lags behind AWS and Microsoft both in terms of market share and revenue, however.

Figures from Statista show AWS retains a dominant position in the cloud computing market with a 29% share, but both Microsoft and Google Cloud have been gaining ground in recent years.

Microsoft, Google Cloud’s closest competitor, still dwarfs it in terms of revenue. The tech giant’s recent earnings report shows its cloud unit recorded 26% growth across the quarter and surpassed $50 billion in revenue - nearly three-times that of Google Cloud.

Bullish spending plans

While Alphabet earnings highlight the continued success of its AI and infrastructure push, Sustar noted that growth “comes at a hefty price” as capital expenditure projections soared.

The company plans to spend between $175 billion and $185 billion across 2026, roughly twice what it spent last year and a significant increase on analyst expectations of $115 billion.

Pichai attributed this to supply constraints and future planning, with the company appearing keen to strike while the iron is hot with bullish infrastructure investment.

“We’ve been supply constrained even as we’ve been ramping up our capacity,” he told investors. “Obviously, our capex spend this year is an eye towards the future. We are constantly planning for the long-term.”

Much of this spending will go toward compute capacity to meet the aforementioned constraints, according to CFO Anat Ashkenazi, with around 60% of spending focused on upgrading existing hardware assets and construction of new data center infrastructure.

Google was famously caught off guard by the launch of ChatGPT in late 2022 and had a rocky start following a lackluster product launch with Bard in February 2023.

However, company appears to have cast off initial concerns about its ability to compete with Microsoft and OpenAI.

The launch of Gemini 3 Pro in November 2025 marked a significant turning point for the tech giant. The new flagship model vastly outperformed competitor options such as GPT-5.1 and Claude Sonnet 4.5 on multimodal reasoning benchmarks, for example.

MMMU-Pro testing showed Gemini 3 Pro scored 81% versus GPT-5.1’s 76% and Claude Sonnet 4.5’s 68%.

Hardware is another area in which Google has recorded positive signs. At Google Cloud Next 2025 the company unveiled ‘Ironwood’, its seventh generation TPU that, at the time, offered industry-leading capabilities for AI workloads.

While Google still uses Nvidia hardware across its infrastructure, this has become a notable USP for the company, giving it a degree of flexibility in meeting customer compute demands as well as a buffer against chip supply constraints.

As ITPro reported last year, the tech giant has invested heavily in its TPU range, and previous models including Gemini 2.5 Pro and Gemma 3 were trained using the hardware.

FOLLOW US ON SOCIAL MEDIA

A former Google software engineer has been convicted on seven counts of economic espionage and seven counts of theft of AI trade secrets after stealing sensitive company documents.

Linwei Ding, also known as Leon Ding, 38, stole thousands of pages of confidential information containing Google trade secrets related to AI technology. The jury’s verdict followed an 11-day trial before US district Judge Vince Chhabria.

"Silicon Valley is at the forefront of artificial intelligence innovation, pioneering transformative work that drives economic growth and strengthens our national security," said US attorney Craig Missakian.

"The jury delivered a clear message today that the theft of this valuable technology will not go unpunished. We will vigorously protect American intellectual capital from foreign interests that seek to gain an unfair competitive advantage while putting our national security at risk."

According to the evidence presented at the trial, while working at Google between May 2022 and April 2023, Ding stole more than two thousand pages of confidential information.

He then uploaded them to his personal Google Cloud account in December 2023, less than two weeks before he resigned from Google.

Working with Chinese companies

Prosecutors said Ding secretly hooked up with two China-based technology companies and began founding his own technology startup, applying for a Chinese government-sponsored “talent plan”.

This program was designed to encourage people working in research and development outside China to send their knowledge and research back to China in exchange for salaries, research funds, lab space, or other incentives.

In a PowerPoint presentation that formed part of his application, Ding claimed that his company’s product “will help China to have computing power infrastructure capabilities that are on par with the international level.”

The trade secrets he stole related to hardware infrastructure and software platforms that allow Google’s supercomputing data center to train and serve large AI models.

They included detailed information about the architecture and functionality of the company's custom Tensor Processing Unit (TPU) chips and GPU systems, for example, as well as its custom-designed SmartNIC AI supercomputer network interface card.

Ding will next appear at a status conference on 3 February and faces a maximum sentence of 10 years in prison for each count of theft of trade secrets.

“This conviction reinforces the FBI’s steadfast commitment to protecting American innovation and national security. The theft and misuse of advanced artificial intelligence technology for the benefit of the People’s Republic of China threatens our technological edge and economic competitiveness,” said FBI special agent in charge Sanjay Virmani.

“The FBI San Francisco division serves Silicon Valley and the companies who lead the world in innovation, and we are committed to safeguarding their work.”

FOLLOW US ON SOCIAL MEDIA

Meta plans to invest as much as $135 billion in AI infrastructure this year, nearly doubling its capital expenditure, as it seeks to catch up with key competitors in the generative AI race.

Meta posted results with $59.89bn in quarterly revenue, topping analyst expectations of $58.41bn. For the full year, Meta posted revenue just above $200bn, up 22% from the year before, with net income of $60bn versus $62bn a year before.

"We had strong business performance in 2025," said Mark Zuckerberg, Meta founder and CEO, in a statement. "I'm looking forward to advancing personal superintelligence for people around the world in 2026."

In a call with analysts after the results were announced, Zuckerberg predicted "a major AI acceleration" including the launch of agents that "really work", adding that Meta would release new models and products in the coming months.

"I expect our first models will be good, but more importantly, we’ll show the rapid trajectory that we’re on," he said. "Our vision is building personal super intelligence."

Zuckerberg suggested that might show up in the company's smart glasses or other wearables — an area OpenAI is hoping to enter this year too.

In that call, CFO Susan Li said Meta was putting up the cash to back that up with capital expenditures in the range of $115bn to $135bn.

Last year, Meta was reportedly set to spend around $60bn, but that was increased to up to $72bn later in the year. Microsoft and Alphabet spent between $80bn and $90bn, while Amazon's forecasted spend was expected to top $100bn.

Meta is scrambling to catch up

Meta's results came alongside quarterly reports from AI rivals including Microsoft, which saw shares slide amid concerns from investors that spending too much on AI and affiliated infrastructure was impacting growth.

The bigger concern for Meta may be that it's at risk of falling behind due to lack of investment, according to Mike Proulx, VP researcher director at Forrester.

"Meta, just like every other big tech company, is now first defined by AI. In the glasses race, it’s ahead. But in the model race, it still trails," Proulx said. "If Meta wants to deliver on its 'personal superintelligence' vision, it needs to catch up with Google and OpenAI.

“That means heavy investment and real focus."

Proulx added that Meta shifting investment out of the metaverse was a "positive move", but suggested that the messaging from Zuckerberg on upcoming model releases might not fill users with confidence.

"Meta also tried to temper expectations around its next set of models, calling them merely 'good' and hinting that bigger upgrades are coming later. But by then will it be too late?"

Meta's Llama model range proved highly popular with users when first released in 2023, and the company pivoted hard to focus on generative AI alongside industry rivals.

Subsequent releases have proved controversial, however. The launch of Llama 4, for example, saw company executives deny claims that model benchmarking results had been exaggerated.

The pace of change in the industry has also been rapid, with big tech providers such as Google, OpenAI, and Anthropic all engaged in a war of attrition to roll out more powerful and efficient models.

OpenAI's initial lead alongside Microsoft on this front has since been whittled down by competitors such as Anthropic and Google.

Over the last 18 months, the latter has made great strides in catching up with the duo, and some industry analysts suggested that the launch of its Gemini 3 model range in November placed it firmly in the lead.

FOLLOW US ON SOCIAL MEDIA

The environmental impact of AI is a growing area of study and one that businesses must begin to seriously consider.

When you power a data center with coal, it’s obvious that it’s having a detrimental impact on the environment. But is a ‘green’ data center really green? And to what extent might the benefits of AI outweigh potential environmental negatives?

In this episode, Jane and Rory discuss the shifting sustainability targets of the world’s public cloud giants, as caused by AI, and what they’re doing to get back on track.

Highlights

"Google had a 22% rise in scope three emissions in 2023 for example, and it pinned a lot of this on increases in data center capacity. Microsoft and AWS don't give exact figures, but they similarly attributed it to data center build out. AWS, in fact, said that its scope three rise was primarily from data center construction and fuel consumption by third party delivery service providers, so that all gets tallied up."

"One of the ways that they're beginning to do this is by using cleaner materials. So something that we covered is that Microsoft has actually begun using mass timber for its data center construction, which to some degrees, supplements steel and concrete. Google and AWS and Microsoft as well are also looking at methods to buy lower carbon concrete, which is just formulated in a different way, to reduce the upfront emissions associated with it."

"In all cases, this is not out of the hands of the company. The company has just made a decision that it would rather pursue aggressive data center expansion than meeting its 2024, 2025, sustainability targets."

"Ultimately, though, these sites can only be so efficient, and when you really, really drill down, one of the main things that they're going to have to do... is secure just clean power or data centers in order to prove credentials. The other thing that the hyperscalers will really have to do before, I think before 2030 is prove that the actual output of these data centers has some kind of net positive for the environment."

Footnotes

• Microsoft Environmental Sustainability Report 2025
• Amazon 2024 Sustainability Report
• Google 2025 Environmental Report
• Google emissions have surged 51% in five years – but it’s making solid progress in data center efficiency
• Data center carbon emissions are set to skyrocket by 2030, with hyperscalers producing 2.5 billion tons of carbon
• Small businesses are ‘flying blind’ on carbon emissions and struggling to meet sustainability goals – and the blame lies with big tech vendors
• Big tech’s solution for AI-related carbon emissions could be more AI
• Microsoft wants to drastically cut carbon emissions, so it’s building data centers with wood
• Can small modular reactors meet data center power demand?
• Google just confirmed the location of its first small modular reactor
• Hyperscalers go nuclear
• Gas-powered data centers: what's behind the boom?
• Could all data centers go solar?
• TPUs: Google's home advantage
• What is a tensor processing unit (TPU)?
• AI’s thirsty secret

Subscribe 

• Subscribe to The IT Pro Podcast on Apple Podcasts
• Subscribe to The IT Pro Podcast on Spotify
• Subscribe to the IT Pro newsletter
• Join us on LinkedIn

When DeepSeek hit the scene in January 2025, the Chinese AI company sent shockwaves through the industry.

The launch of the company’s R1 model, a fine-tuned version of its previous V3 range, prompted a market meltdown. Broadcom and Nvidia shares nosedived, wiping $593 billion from the latter’s market value alone.

DeepSeek was a big wake-up call for US counterparts for a number of reasons. Chief among them was the fact it proved Chinese AI developers weren’t as far behind as some industry stakeholders believed – and the fact its advances were open source.

Notably, the most jarring aspect of the DeepSeek R1 launch lay in the reputed cost of building it. Built on a shoestring budget compared to those offered by US counterparts, this raised serious questions about the huge financial outlay associated with AI training.

In the year prior, big tech talking heads repeatedly touted the prospect of unfathomably large model training costs. Speaking in July 2024, Anthropic CEO Dario Amodei predicted skyrocketing fees in this domain, suggesting that future models could cost upwards of $100 million to build.

It’s safe to say that DeepSeek poured cold water on these bold claims and, in doing it with such bravado, created a sense of unease among US rivals. R1 went toe-to-toe with anything US providers could offer in terms of performance.

The model was trained on 37 billion active parameters and 671 billion total parameters, according to figures released at the time.

For context, benchmarks published by the Chinese firm showed it outpaced Anthropic’s Claude 3.5 and OpenAI’s GPT-4o in key areas, particularly in coding capabilities.

DeepSeek was an instant hit, with users globally flocking to the new application. Free to access, the DeepSeek app toppled ChatGPT as the top free application on the Apple App Store.

But one year on from the launch, where does DeepSeek stand? The initial furor and excitement has subsided to some extent, but the company is still plugging away – and so are users.

Figures from OpenRouter show token consumption for DeepSeek V3 0324 topped 7.27 trillion across 2025, ranking it fifth behind counterparts such as Claude Sonnet 4 and Gemini 2.0 Flash.

DeepSeek has been quietly innovating

DeepSeek’s V3 foundation model has had a number of updates across the last year, with V3.1 arriving in August. A hybrid open weight model, this marked a first major step toward agentic reasoning, combining thinking and non-thinking modes within a single system.

Once again, performance and cost were key talking points, performing well on coding benchmarks while offering more efficient compute costs compared to other models on the market.

Hot on the heels of V3.1, the Chinese firm launched V3.2 and V3.2-Speciale in December. The first of these, according to DeepSeek, places a strong focus on mimicking human reasoning while the Speciale range is designed primarily for mathematical contexts, boasting "maxed-out reasoning capabilities".

Both models offered equivalent performance to OpenAI’s GPT-5, with the team describing it as “your daily driver” in a clear nod to its use in a similar fashion to ChatGPT.

Let’s talk about cybersecurity

Wherever a conversation about DeepSeek arises, questions over security and data privacy haven’t been far away. Indeed, the R1 launch prompted an array of warnings from industry stakeholders about the risks associated with the model.

Indeed, February 2025 saw Australian lawmakers ban the application from government devices and systems, citing security concerns. Since then, lingering worries about the Chinese AI provider have persisted across both government and industry.

Speaking to ITPro in August last year, Andy Ward, SVP International at Absolute Security, said enterprises should exercise extreme caution when using the application, describing it as akin to “printing out and handing over confidential information”.

Research on the potential risks associated with DeepSeek have been highlighted in a slew of studies since its release. Less than a month after launch, researchers at Cisco warned the model contained “critical safety flaws” that left it open to an array of jailbreak techniques that could be pounced on by threat actors or leave enterprises vulnerable to data exposure.

Separate research from Harmonic Security found DeepSeek displays a disproportionately high risk of sensitive data exposure compared to other models out there on the market - even Chinese offerings.

Analysis by the company found the use of Chinese AI models is growing, with DeepSeek accounting for 25% of overall usage alongside offerings such as Alibaba’s Qwen range.

Crucially, however, DeepSeek accounts for 55% of sensitive exposure, which Harmonic researchers said was partly due to its popularity with coders.

DeepSeek V4 could shake things up again

The arrival of DeepSeek V4 appears to be imminent, according to recent speculation.

Reports from The Information, citing sources with intimate knowledge of the development process, said this latest iteration could arrive in early 2026.

This model will place a strong focus on code-generation features, a recurring theme in recent months amidst the arms race between Anthropic, OpenAI, and Google.

According to the publication, early benchmarks suggest the model will outperform the Claude and GPT ranges on this front.

As with the launch of DeepSeek R1, the Chinese firm is also expected to confirm another technical breakthrough on model development in terms of contending with lengthy code prompts.

If this speculation comes to fruition, we could see another scramble among US AI developers to one-up the company.

FOLLOW US ON SOCIAL MEDIA

Apple is turning to Google to help underpin its Siri service and “Apple Intelligence” capabilities, marking the latest sign the tech giant is gaining momentum in the AI race.

The duo have signed a multi-year partnership that will see Google support a major upgrade to Siri, which is expected to launch later this year.

Moving ahead, Apple’s foundation models will also be based on Google’s in-house Gemini AI model range, the company confirmed.

“After careful evaluation, Apple determined that Google's Al technology provides the most capable foundation for Apple Foundation Models and is excited about the innovative new experiences it will unlock for Apple users,” the duo said in a joint statement confirming the deal.

“Apple Intelligence will continue to run on Apple devices and Private Cloud Compute, while maintaining Apple's industry-leading privacy standards.”

The move comes in the wake of several months of speculation over closer ties between the two tech giants. Google already pays Apple billions of dollars to remain the default search engine on iPhones, for example.

According to reports from Bloomberg in August last year, talks between the two over a potential deal for Gemini to power Siri were in the works. Separate reports from the publication noted the deal could see Apple pay Google upwards of £1 billion a year to use the model range.

The deal also represents somewhat of a snub for OpenAI. Apple previously integrated ChatGPT into macOS and forged close ties with the company on Apple Intelligence, for example.

Opting for a deal with Google suggests the tech giant views the Gemini range as a more reliable option than OpenAI's flagship GPT series.

Google is pulling ahead in the AI race

From Google’s perspective, the deal with Apple marks a huge seal of approval for its Gemini model range. Wedbush analyst Dan Ives said the partnership is a “major validation moment for Google” in a Monday research note, per reports from Fortune.

Alphabet, Google’s parent company, hit a major milestone in the wake of the deal, surpassing a $4 trillion valuation for the first time. The surge means Alphabet is just the fourth company to hit such a milestone alongside Nvidia, Microsoft, and Apple.

A key factor behind the breakneck growth at Google has been its growth strength in the generative AI race. The company has recorded consistent growth and progress on this front over the last three years following a rocky start.

When ChatGPT launched in November 2022, Google was caught off guard by the release, prompting a rapid shift in focus led by CEO Sundar Pichai.

As ITPro reported at the time, CEO Sundar Pichai issued a ‘code red’ alert which saw the company reallocate personnel and resources in a bid to keep pace with OpenAI and Microsoft.

Recent events suggest that Google has firmly caught up with key competitors in the AI space, particularly with regard to model capabilities.

In November, Google announced the launch of Gemini 3. The new model range outperformed OpenAI’s GPT-5 model across a range of key benchmarks. That launch precipitated a similar ‘code red’ scramble at OpenAI, with CEO Sam Altman pushing for the company to release GPT-5.2 in quick fashion.

A huge victory for Google

I wouldn’t be surprised if somewhere in the Googleplex, Alphabet’s sprawling Mountain View HQ, there isn’t a proud executive today hanging up a plaque that reads “Google's Al technology provides the most capable foundation for Apple Foundation Models”.

This is a major victory for Alphabet, particularly Google DeepMind and Google Cloud, which have spent the past year turning Google’s cloud empire into the foundations for AI dominance.

At Google Cloud Next 2025, I wrote that the firm had demonstrated its clear strength in enterprise AI, backed by its colossal cloud infrastructure, home-grown TPUs for AI training and inference, and first-hand access to the world’s data.

No doubt Alphabet will shout about this deal from the rafters at Google Cloud Next 2026.

Beyond the short-term victory laps, there are also significant long-term reputational gains for both Google and Apple that can arise from this deal.

After first announcing Apple Intelligence at WWDC 2024, Apple has had a rocky 18 months of repeated delays. Users who bought the iPhone 16 Pro on the understanding that they’d soon have a revamped Siri are still waiting for the promised updates.

In the meantime, the 16 Pro has been eclipsed by an entire generation of devices.

Apple has long practiced a ‘slow but effective’ approach to both its hardware and software, in which it shows up years late to technological trends but executes them to such a high standard that its user base can’t complain. Apple’s approach to AI has been no different, as we’ve previously noted.

With the iPhone, we’ve seen it with ultrawide lenses, high refresh rate panels, and later this year could see Apple do it again with a foldable.

But Apple users have no qualms about adopting competing software in the meantime – in the form of Gemini, OpenAI’s ChatGPT, and Anthropic’s Claude. Clearly, Apple has decided that its customers have waited too long and Google is the most reliable partner for immediate results.

Logan Kilpatrick, product lead for Google AI Studio and Gemini API, took to X in July to announce that the firm had processed 980 trillion tokens in May, across its products and APIs. This number has since risen to 1.3 quadrillion monthly tokens, with no notable outages across Google’s AI portfolio nor significant limitations imposed on free tier users.

Note also that this is a “multi-year” deal, with Google involved in crafting a range of new foundation models for Apple. Here the company can leverage its unparalleled access to data across Search, YouTube, Google Photos, and more, at a scale Apple can’t match.

The exact interdependencies between the Gemini family of models and Apple Intelligence is still to be determined, with Apple insisting the service will continue to operate both on device and using its Private Cloud Compute infrastructure.

In all likelihood, however, technical information will be kept at a minimum as Apple seeks to uphold a seamless AI experience for its iOS and macOS users.

FOLLOW US ON SOCIAL MEDIA

Google has announced plans for deeper AI integration within Gmail to help users automate inboxes and tell you what to do next – though this iteration of AI Overviews is hopefully more successful at summarizing your emails than Google has been with search.

The tech giant revealed its Gemini AI will be set loose on Gmail inboxes to build an automated to-do list for users, as well as help dig out useful information from emails via search, and craft automated replies.

The AI-powered systems will first be available in the US in English, with wider availability across languages and regions in the coming months.

"Your inbox is full of important information, but accessing it has required you to become a power searcher," said Blake Barnes, VP of Product for Gmail, in a blog post.

"And even when you find the right emails, you are often left staring at a list of messages, forced to dig through the text to piece together the answer. That’s why we’re introducing AI Overviews."

AI Overviews is the same system used to generate summaries in response to Google search queries – which has been criticized since its launch for misleading information.

When it arrived in Search in 2024, AI Overviews was spotted telling people to glue cheese to pizza or eat rocks – and the challenges have continued, with recent reports that AI Overviews is offering misleading health advice.

How AI Overviews in Gmail will work

In Gmail, AI Overviews will have two primary functions: synthesizing an email conversation into a summary of key points and searching through an inbox to find information.

To use the latter, users simply have to ask a question in the Gmail inbox search bar, Barnes explained.

"Instead of hunting for keywords or digging through a year of emails, just use natural language, like 'Who was the plumber that gave me a quote for the bathroom renovation last year'?" he noted.

"Gemini’s advanced reasoning pulls the answer, instantly summarizing the exact details you need."

Both AI Overview functions are rolling out to Gmail immediately, though the inbox questions feature is only available to Google AI Pro and Ultra Subscribers, while the conversation summaries are available to all Gmail users at no cost.

Google also is testing a new AI Inbox, initially only for "trusted testers" before wider availability in the next few months. This will automatically summarize your messages and pull together a to-do list.

"AI Inbox is like having a personalized briefing, highlighting to-dos and catching you up on what matters," said Barnes. "This lets high-stakes items — like a bill due tomorrow or a dentist reminder — rise to the top."

AI-powered email writing tools are coming to Gmail

Google noted that AI Overviews isn't the first time that AI has been applied to Gmail, as it's been used to power spam blocking systems and to generate Smart Replies.

Indeed, alongside the addition of AI Overviews for searching messages, Google also introduced "Help Me Write", a tool to create or edit emails, as well as "Suggested Replies," which is an update to Smart Replies.

"Suggested Replies can instantly draft an initial response in your tone and style, leaving you free to refine it before giving it your approval," said Barnes. "You can also use the new Proofread feature for advanced grammar, tone and style checks so everything is polished before you send."

The combination of AI Overview summaries of emails paired with Suggested Replies and Help Me Write raises the specter of messages being read by AI and responded to by AI, with only a quick glance and tap of a button by humans.

Help Me Write and Suggested Replies are available to everyone at no cost, while the Proofread tool is being held back for Google AI Pro and Ultra subscribers.

Can you turn off Gmail’s AI features?

Beyond AI Overviews' advice to eat rocks, it's worth noting that generative AI systems continue to make mistakes, so business users – and anyone else – should be aware that accuracy is not guaranteed.

And that's a problem with AI being added to tools, Google or otherwise, without users requesting or desiring them.

This new tranche of AI tools for Gmail are initially only available to American users, but once they start rolling out more widely, it will be possible to turn them off – though that will also mean users lose key tools like spell checking, according to reports from The Verge.

Existing AI features can be managed in Settings in Google Workspace.

A spokesperson for Google told ITPro users will be able to switch off the new features, should they choose.

Users are able to toggle 'smart features and personalization' off in their settings "at any time".

"If you do, you return to the classic Gmail experience," the spokesperson explained. "We believe AI should be an opt-in partner, not a forced requirement."

FOLLOW US ON SOCIAL MEDIA

Google-owner Alphabet is buying data center and energy firm Intersect in a deal valued at $4.75 billion.

The move comes amid a race to rollout data centers – and the energy systems required to power them – to fuel AI.

In the last year alone, data center investment topped $61bn, though concerns continue over whether energy systems will be able to keep up with demand.

As part of the deal, Alphabet will acquire the $15bn in data center and energy assets Intersect already has in operation or is working to build, Reuters noted, adding that by 2028 Intersect had planned to have 10.8 gigawatts of power either up and running or in development.

"Intersect will help us expand capacity, operate more nimbly in building new power generation in lockstep with new data center load, and reimagine energy solutions to drive US innovation and leadership," said Sundar Pichai, CEO of Google and Alphabet.

Google targets energy innovation

Google already owns a minority stake in Intersect. However, moving forward Alphabet said Intersect would operate separately from itself and Google, keep the company brand, and continue to be led by CEO Sheldon Kimber.

The aim is for Intersect's team to work closely with Google's technical infrastructure division on joint projects, including a data center and power site already under construction in Haskell County, Texas, the duo said.

Intersect noted it will work to develop emerging technologies to diversify energy supply while also boosting Google's data centre projects.

"Intersect has always been focused on bringing innovation to the industry and we look forward to accelerating at scale as part of Google," said Sheldon Kimber, founder and CEO, Intersect.

"Modern infrastructure is the linchpin of American competitiveness in AI. We share Google’s conviction that energy innovation and community investment are the pillars of what must come next."

The deal doesn't include currently operating assets in Texas or California, which run as separate companies and will be bought out by existing investors, Kimber said in a blog post.

Powering AI

With infrastructure expansion gathering pace worldwide, experts have questioned whether supply will be able to keep up with demands placed on data centers as a result of AI.

Moreover, similar questions have been raised over whether current energy systems are up to the task of meeting future power demands. Data center power demand is predicted to top 106GW in the next ten years, according to a BloombergNEF report.

Google has already said it will pause AI data centers to help balance overloaded grids, and signed deals for renewable energy sources – including with modular nuclear reactor developers.

Kimber said deals like Alphabet's acquisition of Intersect are "pillars of what must come next".

"AI today is stuck behind one of the slowest, oldest industries in the country: electric power," he wrote.

"The country has racks full of GPUs that can’t be energized because there isn’t enough electricity for them. The grid is a patchwork operating system that’s been running for a century. An engineering miracle in its time, but not built for the AI era."

Bring your own power

Kimber said the US model is increasingly "bring your own generation".

"Companies like Google are thoughtfully building AI the responsible way," he added. "In the case of Intersect, that has meant renewables paired with flexible backup sources and energy storage, often on the same site as the data centers themselves. This approach can significantly accelerate speed to power and unlock scarce transmission capacity."

While Google has invested in renewables — including a 99% carbon-free data center in Norway — its greenhouse gas emissions have leapt 48% since 2019, driven by electricity use for data centers.

FOLLOW US ON SOCIAL MEDIA

Google Cloud has announced the first details of its new Partner Network, which is set to formally roll out in the first quarter of 2026.

The cloud giant is positioning the revamped channel initiative as a shift away from tracking program work to recognizing partner contributions across the full customer lifecycle.

Available to all partner types and sizes, the framework introduces a Diamond tier for the highest-performing partners, new competencies, as well as improved automation across the program.

The program will roll out in Q1 2026 and will include a six-month transition window to help partners adjust to the new structure.

In a blog post, Google Cloud said the initiative’s three core pillars – simplicity, outcomes, and automation – will better reward successful co-sale efforts, high-quality service delivery, as well as shared innovation with ISVs.

“Google Cloud Partner Network is being completely streamlined to focus on real-world results,” commented Colleen Kapase, Google Cloud’s VP of channels and partner programs.

“This marks a strategic shift from measuring program work to valuing genuine customer outcomes”

Google Cloud's streamlined framework

According to the tech giant, the new Partner Network has been designed to recognize partners for contributions such as pre-sales influence, co-innovation, and pre-sales support.

There’s also greater recognition for investments in skills, real-world experience, and successful customer outcomes.

On the structure front, the initiative is moving from a two-tier to a three-tier model with the introduction of a new Diamond category to the existing levels of Select and Premier.

This will be reserved for the highest-performing partners that consistently deliver "exceptional customer outcomes” across Google Cloud and Google Workspace, the hyperscaler said.

New competencies on the way

Partners will also see current specializations replaced with a new competency framework as Google Cloud reimagines how it rewards technical and sales capabilities.

New competencies will assess partners on two factors: capacity – measured through certifications and sales credentials – and capability – which will be tracked through validated pre-sales and post-sales contributions.

These competencies will be earned independently of tier status, with an additional Advanced Competency level signaling higher designation.

Increased automation

Across the entire program, Google Cloud is also doubling down on its automation focus.

Building on the introduction of its Earnings Hub and Statement of Work Analyzer tools, the Partner Network Hub will now use AI-driven automation to track partner progress automatically.

Successful customer engagements will be applied across all eligible tiers and competencies to help drive down time-consuming administrative and reporting tasks.

“This radical simplification eliminates redundant reporting and ensures seamless, comprehensive recognition for the outcomes delivered,” Kapase said.

FOLLOW US ON SOCIAL MEDIA

Cohesity has announced an expansion of its strategic partnership with Google Cloud in a bid to strengthen cyber resilience and drive enterprise AI adoption through new product integrations.

Building on the pair’s existing long-term collaboration, the agreement includes new solutions geared towards helping organizations better protect their data, meet regulatory requirements, and unlock greater value through enterprise AI deployments.

Additions include integrations between Google Cloud’s AI and security offerings and Cohesity’s enterprise AI assistant, Gaia, its Data Cloud platform, as well as Cohesity DataProtect.

“This collaboration with Google Cloud represents a bold step forward in redefining how enterprises protect, secure, and gain insight from their data, while reframing sovereignty from a regulatory burden into a source of trust and advantage,” said Cohesity’s CEO and president, Sanjay Poonen, in an announcement.

AI-driven insights

At the heart of the agreement lies Cohesity’s enterprise AI assistant, Gaia. Building on existing integrations with Google Gemini models, the vendor plans to embed Google Cloud’s Vertex AI Search capabilities into Gaia to enable instant, grounded answers to structured user queries.

The tech giant said this will offer citations pulled from source files stored within Cohesity’s immutable data lake.

There’s also a new integration with Gemini Enterprise in the pipeline, which will allow AI agents built on Google’s platform to securely access historical enterprise data stored within Cohesity’s Data Cloud.

The firm said the integration will enable “seamless connection” with existing workflows.

Additionally, customers can now leverage a Google Cloud-hosted version of the Gaia platform, designed to offer improved performance and response through Google’s global infrastructure.

Security enhancements

On the security front, the Cohesity Data Cloud already incorporates Google Threat Intelligence to help customers detect and respond to threats to their data stores, with Google’s indicators of compromise (IOCs) allowing for both proactive hunting and reactive threat scanning within the Cohesity platform.

The collaboration also pairs Cohesity’s Cyber Event Response Team (CERT) with Google Cloud’s Mandiant Incident Response team to support joint customers in their recovery from potentially destructive cyber attacks.

The pair said they will offer deeper integration between the Cohesity Data Cloud and Google Security Operations to automate security operations workflows and detect threats hidden within backup environments.

A planned cyber resilience SaaS solution will offer cloud-isolated recovery to help reduce recovery times.

Digital sovereignty

The expanded partnership also focuses on data sovereignty as compliance requirements continue to increase in complexity.

Cohesity is certified as a Google Cloud Ready – Regulated & Sovereignty Solutions partner, meaning customers can deploy the firm’s Data Cloud offering within the Google Cloud Data Boundary to comply with data residency and sovereign control requirements.

Cohesity FortKnox on Google Cloud also promises immutable, geographically-restricted cyber vaulting with policy-based controls geared towards meeting sovereignty requirements.

Additional highlights

Elsewhere, the partnership expands data protection capabilities for Google Cloud services such as Compute Engine and Cloud Storage, with additional support planned for GKE, BigQuery, and other data-centric services.

Cohesity and Google Cloud will also collaborate on joint go-to-market programs, co-selling initiatives, as well as integrated marketing campaigns. All components of the Cohesity Data Cloud are available via the Google Cloud Marketplace.

Google Cloud CEO, Thomas Kurian, said the agreement underlines the companies’ shared commitment to helping organizations “unlock the full potential of their data.”

“By combining Google Cloud’s AI and security innovations with Cohesity’s cyber resilience foundations, we’re enabling enterprises to accelerate AI transformation while building a more resilient, compliant future for customers everywhere in the world,” he commented.

FOLLOW US ON SOCIAL MEDIA

Google DeepMind CEO Demis Hassabis thinks there’s one particular area of the technology industry that should have investors worried about an AI bubble: startups.

When asked about the possibility of a looming bubble during a recent podcast appearance, Hassabis told Professor Hannah Fry that early-stage startups raising huge sums are a serious cause for concern.

“There is a lot of talk right now about AI bubbles,” he said. “I think there are parts of the AI ecosystem that are probably in bubbles.”

“One example would be, you know, just seed rounds for startups that basically haven’t even got going yet, and they’re raising at tens of billions of dollars valuations just out of the gate,” Hassabis added.

“It's sort of interesting to see. How can that be sustainable? My guess is probably not, at least not in general.”

Hassabis isn’t the first industry stakeholder to raise the alarm over huge startup valuations in the AI space. Earlier this year, a report from Silicon Valley Bank (SVB) warned about the rise of ‘zombiecorn’ startups – those that raise vast sums and will ultimately fail as they have no clear route to market.

SVB noted that the long-term viability of many high-flying startups in the AI space is questionable, with many recording poor revenue growth and collapsing quickly post-funding.

Speaking to ITPro at the time, Tom Glason, CEO and co-founder at Scalewise, said the report showed rampant generative AI hype among venture capital (VC) firms was creating a new wave of doomed startups.

“The AI boom has fueled a wave of overfunded startups that look healthy on the surface, but are commercial hollow underneath,” he told ITPro. “These so-called ‘zombiecorns’ raise huge rounds but fail to build sustainable revenue or viable unit economics.”

Big tech might swerve an AI bubble

Elsewhere, Hassabis suggested he’s less concerned about an AI bubble impacting big tech providers, mainly because there’s “a lot of real business underlying” growth in this domain.

However, he did point toward a potential “overcorrection” as a result of the huge hype and subsequent investment in the space.

“I think for any new, unbelievably transformative and profound technology, of which, of course, AI is probably the most profound, you're going to get this over correction in a way,” he said.

“When we started DeepMind, no one believed in it. No one thought it was possible. People were wondering what’s AI for anyway? And then now, fast forward 10-15 years, and obviously it seems to be the only thing people talk about in business,” Hassabis added.

“It’s almost an overreaction to the underreaction. So I think it’s natural. I think we saw that with the internet, with mobile, and I think we’re seeing, or going to see it, again with AI.”

Google is insulated from a potential bubble

Notably, Hassabis suggested that Google will be insulated from an AI bubble if that situation arises.

“I don't worry too much about, are we in a bubble or not, because from my perspective, as you know, leading Google DeepMind and with Alphabet as a whole, our job, and my job, is to make sure we either way, we come out of it very strong, and I think we're very well positioned,” he told Fry.

A key factor here, he noted, is the diversity across the tech giant’s products. While Google has rolled out AI features across a range of domains, including Workspace, Chrome, and Search, AI isn’t the sole focus or money-spinner for the firm.

“I think we are tremendously positioned either way. So if it continues going like it is now, fantastic, we’ll carry on,” Hassabis added.

“If there’s a retrenchment, fine, I think we’re in a great position because we have our own stack with TPUs. We also have all these incredible Google products and the profit that all makes.”

Hassabis’ bullish position on Google’s outlook in the event of an AI bubble runs counter to recent comments from CEO Sundar Pichai.

During a recent interview with the BBC, Google chief executive Sundar Pichai told the broadcaster “no company is going to be immune, including us”.

Pichai noted that current hype and surging investment in the technology bears some similarities to the early days of the web, with rampant “excess investment” that could come back to haunt some big tech providers.

Again, though, the party line at Google appears to be that the tech giant is well positioned to withstand any potential market chaos in the event of a bubble bursting.

FOLLOW US ON SOCIAL MEDIA

Apple, Google, and other browser makers have rolled out patches for zero-day bugs that are already being used by threat actors in "sophisticated" attacks.

Google noted that an exploit for one of the bugs exists in the wild already and was spotted by its Threat Analysis Group – that largely works on serious attacks led by state actors or similar – and was sorted via coordination with Apple engineers.

"For these Apple- and ANGLE-related issues, the quiet, coordinated disclosure strongly suggests the vendors viewed the bugs as high-risk and potentially already known to capable adversaries," Douglas McKee, director of vulnerability intelligence at Rapid7, told Dark Reading.

One of the bugs impacts other browser makers using Chromium, including Microsoft Edge and Vivaldi, which have also rolled out patches this week.

Patches for Apple

Alongside a set of other updates, Apple issued emergency patches for two issues in WebKit for devices running versions of its software before iOS 26. The flaws, CVE-2025-14174 and CVE-2025-43529, were credited, in full or in part, to Google Threat Analysis Group.

The first meant that accessing a webpage with "maliciously crafted" content could lead to arbitrary code execution. "A use-after-free issue was addressed with improved memory management," Apple said in a support document.

For the second, Apple said that malicious websites could lead to memory corruption, saying the issue was "addressed with improved validation."

"Apple is aware of a report that this issue may have been exploited in an extremely sophisticated attack against specific targeted individuals on versions of iOS before iOS 26," the company said in a support document detailing both issues.

The patches for the zero-day flaws are available for devices going back to iPhone 11, iPad Pro 12.9-inch 3rd generation, iPad Pro 11-inch 1st generation, iPad Air 3rd generation, iPad 8th generation, and iPad mini 5th generation.

Apple gave little extra detail about the zero-day flaws, saying it "doesn't disclose, discuss or confirm security issues". However, the tech giant issued a patch for the bugs alongside a set of other security issues that included further fixes for WebKit, the ScreenTime tool and more.

Google patches flaws

Google revealed its flaw via an update to the Stable Channel for the desktop version of its Chrome browser, crediting the discovery of CVE-2025-14174 to the Google Threat Analysis Group as well as Apple Security Engineering and Architecture.

Google had initially patched the flaw last week without any details beyond a "high" rating, but has now filled in a few details.

That flaw, labelled as a high-risk vulnerability, was first spotted December 5 and is caused by out-of-bounds memory access in Chrome's Almost Native Graphics Layer Engine (ANGLE), which is also used by WebKit, hence the impact on Apple.

"Out of bounds memory access in ANGLE in Google Chrome on Mac prior to 143.0.7499.110 allowed a remote attacker to perform out of bounds memory access via a crafted HTML page," said the CVE.org support page for the flaw.

"Google is aware that an exploit for CVE-2025-14174 exists in the wild," Google added in a blog post.

The company also patched two other medium-level flaws in the stable channel at the same time.

FOLLOW US ON SOCIAL MEDIA

Google DeepMind has announced plans to build a new automated research lab following a deal with the UK government.

Built from the ground up to be fully integrated with Gemini, the lab will synthesize and characterize hundreds of materials per day, cutting the time needed to identify new materials.

Superconductors that operate at ambient temperature and pressure could allow for low-cost medical imaging and reduce power loss in electrical grids, the firm said, while other novel materials could lead to advanced batteries, next-generation solar cells, and more efficient computer chips.

The agreement also includes plans to work on AI-enabled research in areas like fusion energy.

"DeepMind serves as the perfect example of what UK-US tech collaboration can deliver - a firm with roots on both sides of the Atlantic backing British innovators to shape the curve of technological progress," said technology secretary Liz Kendall.

"Science and technology are at the heart of our mission to drive a new era of national renewal – and partnerships like this will help us go further, faster."

As part of the deal, scientists will be given priority access to Google DeepMind’s existing AI models, including tools like AlphaGenome for DNA sequencing, AlphaEvolve for designing algorithms, weather forecasting models, and a new AI 'co-scientist': a multi-agent system that acts as a virtual research collaborator.

Meanwhile, a new AI system, Gemini for Government, aims to cut bureaucracy and automating routine tasks. Another version of Gemini, grounded in the national curriculum, will be available for teachers following a pilot in Northern Ireland that helped save teachers an average of 10 hours per week.

DeepMind targets AI security boosts

DeepMind also announced plans to expand its work with the UK’s AI Security Institute. This will involve giving access to its proprietary models, data and ideas, and producing joint reports and publications sharing findings with the research community.

Similarly, the collaboration with the institute will bolster existing security and safety research, with DeepMind taking part in technical discussions to tackle complex safety challenges.

The deal expands the existing relationship between the government and Google, which this summer announced plans to help modernize outdated government IT by migrating services to the cloud. More than a quarter of public sector systems were reliant on outdated tech, the government said.

In September, Google unveiled plans to invest £5 billion in UK AI infrastructure and research, including the development of a new data center in Waltham Cross.

"This partnership will make sure we harness developments in AI for public good so that everyone feels the benefits," said prime minister Keir Starmer.

"That means using AI to tackle everyday challenges like cutting energy bills thanks to cheaper, greener energy and making our public services more efficient so that taxpayers’ money is spent on what matters most to people."

MORE FROM ITPRO

• Google jumps on the agentic AI bandwagon
• Demis Hassabis thinks Meta's multi-billion dollar hiring spree shows it's scrambling to catch up in the AI race
• Google says you shouldn't worry about AI malware

Google CEO Sundar Pichai claims software development has become “exciting again” since the rise of vibe coding, but some developers might disagree while bogged down in flawed code.

Speaking during a recent podcast interview, Pichai suggested the trend that’s taken the industry by storm will have positive long-term implications for both devs and non-technical individuals alike.

“It’s making coding so much more enjoyable,” he told host Logan Kilpatrick. “I feel things are getting more approachable, it’s getting exciting again, and the amazing thing is it’s only going to get better.”

Pichai noted that non-technical staff internally at the company are using AI to support workflows and finding them to be a valuable addition to the toolkit.

“There’s a sharp increase in the set of people who have submitted their first CL’s (changelists) and it’s because these tools are making it more accessible,” he said.

“Maybe you’re a product marketing person, you have an idea… in the past you would have described it. Now maybe you’re kind of vibe coding it a little bit and showing it to people, so you can tangible see that come to work.”

Pichai is the latest in a string of big tech execs to hail the potential of vibe coding over the last year. Indeed, Google Brain founder Andrew Ng specifically highlighted the democratizing effect of AI in coding earlier this month, with workers outside the software profession flocking to the tools.

“The bar to coding is now lower than it ever has been,” Ng said. “People that code, be it CEOs and marketers, recruiters, not just software engineers, will really get more done than ones that don’t.”

Developers aren’t quite sold on vibe coding

Software developers don’t appear fully sold on the use of AI for coding, however, with research showing that although there’s a growing appetite for these tools, they often create more problems than they solve.

Findings from Google’s own State of DevOps report last year, for example, revealed more than one-third (39%) of developers had little trust in AI-generated code even though they use them on a daily basis.

Meanwhile, Stack Overflow’s 2025 Developer Survey found nearly half (46%) of developers “don’t trust the accuracy” of AI outputs when used for coding tasks. This marked an increase compared to 31% of devs who aired concerns on this front in 2024.

A key factor behind this skepticism is that the “productivity gains” delivered by the technology are negated by developers having to debug faulty code.

More than two-thirds (67%) of respondents to a survey from Harness said they now spend more time debugging flawed AI-generated code compared to when they were operating in a manual capacity.

Keep security teams in the loop

Pichai did acknowledge that there are risks associated with vibe coding, particularly in terms of security – and he’s not wrong.

A study from Veracode in July this year found nearly half of code generated with AI is fully secure, while a separate survey from Cloudsmith pointed to a growing trend of complacency among developers using these tools.

The survey found only around one-third of devs review AI-generated code before deployment, leaving enterprises open to an array of potential vulnerabilities.

In the meantime, Pichai said vibe coding should typically be used in an experimental capacity, and security teams should always be involved in the process.

“I’m not working on large codebases where you need to really get it right,” he said. “The security has to be there, so those people should weigh in.”

AI tools are the worst they'll ever be

Notably, the Google chief believes this is the worst vibe coding and AI tools will ever be, pointing to examples like autonomous taxi company Waymo. Initial skepticism and concerns about safety have given way to acceptance among commuters.

With AI coding tools, the same rules apply. Continuous improvements mean the technology will become more reliable moving forward, helping cement its place in the mainstream not only for non-technical individuals but for devs working on the frontline.

“There's something I always used to tell about Waymo, and whenever people talked to me about it is, ‘remember, this is the worst Waymo will ever drive’,” he said.

“This version of all these tools…vibe coding with Gemini 3 in AI Studio, it’s both amazing to see and it’s the worst it’ll ever be. Both are simultaneously true. You’re going to see a lot of progress ahead.”

MORE FROM ITPRO

• Want to supercharge your vibe coding skills? Here are the best AI models developers can use to generate secure code
• Vibe coding security risks and how to mitigate them
• ‘Slopsquatting’ is a new risk for vibe coding developers – but it can be solved by focusing on the fundamentals

Google has officially unveiled Gemini 3 Pro, its new state of the art LLM with record-breaking scores across almost every AI benchmark.

The new model is intended to improve every Google service that uses Gemini, including its dedicated app, coding tools, and AI in search.

Google stated that Gemini 3 Pro is much better at handling requests in their intended context and providing useful answers that don’t resort to flattery.

The model debuted in the number one spot across text, WebDev, and vision in LMArena, with Google having claimed Gemini 3 Pro is the “best model in the world for complex multimodal understanding”.

Across multimodal reasoning benchmarks, Gemini 3 Pro was found to consistently outperform competition such as GPT-5.1 and Claude Sonnet 4.5.

In MMMU-Pro, for example, the model scored 81% versus GPT-5.1’s 76% and Claude Sonnet 4.5’s 68%.

ARC-AGI-2 is a rigorous benchmark for testing the capability of AI model reasoning across a series of abstract visual puzzles. Easy for humans to complete but difficult for today’s LLMs, it’s considered a true challenge for frontier models.

Gemini 3 Pro scored 31.1% in tests, far in excess of GPT 5.1’s 17.6% and Claude Sonnet 4.5’s 13.6%.

Gemini 3 is a game changer for Google shops

In other areas, the performance gap between Google’s model and the competition is even more stark.

Gemini 3 Pro scored a new record score of 23.4% at MathArena Apex, a benchmark that tests LLMs on their ability to solve mathematical problems upon which they weren’t trained, compared to just 1% by GPT-5.1 and 1.6% by Claude Sonnet 4.5.

Although the model doesn’t beat the latest version of Claude at the most common coding benchmarks, it greatly improves upon Gemini 2.5 Pro.

Google has also emphasized how its reasoning and visual understanding helps Gemini 3 Pro to do more with the coding capabilities it has, for quicker overall resolution of common developer tasks.

In a demo, Google showed how Gemini 3 Pro could turn an image of a chessboard into an interactive game, as well as a back-of-the-napkin sketch of a website into a functioning page.

Gemini 3 Deep Think, the multi-step reasoning variant of the new model, improves upon its benchmarks at the cost of far slower responses.

In Humanity’s Last Exam, an intense benchmark designed to press LLMs with 2,500 difficult questions across a wide range of subject areas, Gemini 3 Deep Think scores 41%, compared to Gemini 3 Pro’s 37.5% and GPT-5 Pro’s 30.7%.

Gemini 3 Deep Think is still undergoing safety tests but will become available to Google AI Ultra subscribers within the next few weeks.

Gemini 3 Pro was trained using Google’s tensor processing units (TPUs), dedicated hardware for AI training and inference that Google cites as key to its efforts to develop AI sustainably.

Google Antigravity

Alongside the much-anticipated launch of Gemini 3 today, Google also revealed Google Antigravity.

The product is described as a new ‘agent-first’ IDE, with a focus on elevating developers into a manager of AI agents.

Using Gemini 3’s agentic coding features, as well as its reasoning and tool use, Antigravity is intended to allow developers to set agents multi-step, complex coding tasks and receive evidence proving the job has been completed to a high standard.

In a demo, Google showed how a developer could use the tool to build a flight lookup web app, returning flight data based on a flight number provided by the user.

The tool is then capable of creating an implementation plan, writing the code, and then testing the app by opening the Chrome browser. Finally, it provides the user with screenshots of its tests, which the user can critique in order to immediately update the UI of the finalized app.

These features are powered by a combination of Gemini 3, Gemini 2.5 Image (better known as Nano Banana), and Google 2.5 Computer Use.

In another demo, Google showed how enterprise developers can delegate to multiple ‘background agents’ within Antigravity, without needing to stay in the same chat window while tasks are being completed.

Google added that Antigravity is capable of supporting Claude 4.5 Sonnet in addition to its own models.

Antigravity is now in public preview for free, with support for other models including Claude Sonnet 4.5 and OpenAI’s GPT-OSS.

Enterprises with the Vertex AI and Gemini Enterprise subscriptions have access to Gemini 3 from today and the model is also available in the Gemini app, the Gemini API within AI Studio, Antigravity, and Gemini CLI.

The model will be priced at $2 per million input tokens and $12 per million output tokens, for prompts under 200,000 tokens.

Prompts over this limit will cost $4 per million input tokens and $18 per million output tokens.

MORE FROM ITPRO

• Google Brain founder Andrew Ng thinks everyone should learn programming with ‘vibe coding’ tools
• Google wants to take hackers to court
• Google CEO Sundar Pichai sounds worried about a looming AI bubble

We better all hope claims of an AI “bubble” don’t come to fruition, because we’re all going to suffer if it eventually bursts.

That's according to Google CEO Sundar Pichai, who told the BBC in an interview that there was "irrationality" in the AI hype that could impact every company.

"I think no company is going to be immune, including us," he told the BBC, though he stressed Alphabet and Google were in a good position to withstand any potential market shocks.

Pichai compared the hype and investment around AI to the early days of the web, admitting there was "excess investment", but adding that the impact of the internet was worth the costs.

"I expect AI to be the same. So I think it's both rational and there are elements of irrationality through a moment like this."

While concerns still linger over the long-term return on investment from AI adoption, this hasn't stopped big tech companies committing to bold spending plans over the next decade.

Google-rival OpenAI has committed $1.4 trillion in spending over the next several years for cloud and data center infrastructure, despite predictions it will post just $20 billion in revenue itself this year.

At the same time that such wild spending is happening, Pichai told the BBC that anyone using AI should "not blindly trust everything they say."

"We take pride in the amount of work we put in to give us as accurate information as possible, but the current state-of-the-art AI technology is prone to some errors,” he said.

Google's growing energy concerns

Alongside potential market disruptions, Pichai also warned about the long-term impact on the energy sector.

Energy used by data centers in the US is expected to triple by 2028, making up 12% of total demand, and globally AI is predicted to use as much energy as Japan by 2030.

Pichai noted that meeting those supply-related challenges were necessary to keep the economy moving.

"You don't want to constrain an economy based on energy, and I think that will have consequences," he said.

Of course, there are consequences beyond financial ones. When asked about AI's energy requirements, Pichai admitted surging power demands had indeed caused Alphabet's climate targets to slip, though he stressed its 2030 net zero goal was still possible.

Work disruption

In the sweeping interview, Pichai also pointed to the long-term impact of AI on individual workers. The Google chief said that people who adapt to AI will do better in their work than those who don't – despite his own admission of the flaws in the technology that meant it remained prone to errors in accuracy.

Pichai’s comments come in the wake of repeated warnings about how the technology will shape future labor markets.

Entry-level workers, for example, are expected to bear the brunt of the impact from AI, while earlier this month analysts warned that AI job cuts could even prompt a revival of offshoring, meaning any rehiring after AI benefits fail to materialize won't help those laid off.

These concerns have been tempered by some industry stakeholders, however, with OpenAI co-founder Andrej Karpathy recently insisting agents were nowhere near ready to begin stealing jobs.

"It doesn't matter whether you want to be a teacher [or] a doctor," he said. "All those professions will be around, but the people who will do well in each of those professions are people who learn how to use these tools."

"We will have to work through societal disruptions," he said.

MORE FROM ITPRO

• Jensen Huang says 'the AI race is on' as Nvidia shrugs off market bubble concerns
• Is an 'AI' bubble about to pop?
• The tech industry is becoming swamped with agentic AI solutions – analysts say that's a serious cause for concern

Tired of scam messages purporting to have a package for you? So is Google – and it's lawyering up to fight back.

Google said it is adopting a multifaceted approach to takedown a phishing as a service (PhaaS) operation known as 'Lighthouse', not only suing those responsible, but backing bipartisan US legislation to take on such scams and rolling out new AI-based tech to protect users.

"That text message you got about a 'stuck package' from USPS or an 'unpaid road toll'? It’s not just spam. It’s the calling card of a sophisticated, global scam that has swindled victims out of millions of dollars," said Google's general counsel Halimah DeLaine Prado in a blog post.

"Bad actors built 'Lighthouse' as a phishing as a service kit to generate and deploy massive 'smishing' (SMS phishing) attacks."

Those attacks arrive via a text message claiming to have a delivery or warning of an unpaid road toll, with a malicious link where victims are urged to enter their email, banking data, and more.

According to Google, the Lighthouse operation has impacted over one million victims spanning 120 countries, stealing information on anywhere between 12.7 million and 115 million credit cards in the US alone

“This represents a five-fold increase in these types of attacks since 2020,” DeLaine Prado noted.

Google said attacks often make use of legitimate brands and their trademarks on malicious websites, with the tech giant spotting at least 107 website templates using its own branding on fake sign-in screens.

Google getting tough on scams

Google said it is taking legal action in the hopes of dismantling the "core infrastructure" of the Lighthouse operation.

"We are bringing claims under the Racketeer Influenced and Corrupt Organizations Act, the Lanham Act, and the Computer Fraud and Abuse Act to shut it down, protecting users and other brands," DeLaine Prado noted.

The lawsuit is being brought against 25 unnamed people believed to live in China, seeking a restraining order and damages. Of course, given the individuals accused of running Lighthouse are not known, the intent isn't to necessarily target them.

Instead, Google is also asking web hosting providers to block Lighthouse associated IP addresses and domains.

Alongside the lawsuit, Google has thrown its weight behind a trio of bills currently working their way through US Congress: Guarding Unprotected Aging Retirees from Deception (GUARD) Act, Foreign Robocall Elimination Act and Scam Compound Accountability and Mobilization (SCAM) Act.

Those bills would see the establishment of taskforces to target such scams — and funding to investigate them.

Legal actions aside, Google said it is also developing tools using AI to better spot and flag such scams in a bid to better protect users.

Tough fight ahead

While the actions by Google have been welcomed, one industry expert said such efforts may be like playing whack-a-mole. They might knock one down, but another will just pop up again.

"Groups like Lighthouse appear regularly, and while legal action can disrupt them, these operations often re-emerge using alternative infrastructures," said Carl Wearn, head of threat intelligence and analysis & future ops at Mimecast.

"Copycat phishing as a service models will continue to grow, exploiting people’s instinctive trust in familiar digital channels like email and SMS."

While the increase of these scams – which not only now impersonate delivery firms and toll threats but governments and banks to trick victims – may spark more lawsuits from brands following Google's lead, Wearn said that "lasting impact will depend on public awareness, taking a moment to pause, verify and think before clicking."

MORE FROM ITPRO

• Cheap cyber crime kits can be bought on the dark web for less than $25
• Phishing tactics: The top attack trends
• Employee phishing training is working – but don’t get complacent

A tensor processing unit (TPU) is an application-specific integrated circuit (ASIC) that is specifically designed for the purpose of accelerating high-volume mathematical and logical processing tasks that are typically involved with machine learning (ML) workloads. These were designed back in 2015 by Google and released for third-party use in 2018.

The first TPUs were designed back in 2015 by Google and released for third-party use in 2018.

Over successive generations, TPUs have evolved from modest accelerators to massive supercomputer-scale pods. Early TPU generations focused on inference (and later training) with limited precision and high throughput, and were used to power early DeepMind breakthroughs such as the Go-playing program AlphaGo.

Google progressively improved memory bandwidth, chip count, interconnects, and energy efficiency, with recent models focused largely on meeting the massive demand for generative AI hardware. The latest generation, Ironwood, takes a major leap forward particularly when it comes to running “thinking” AI workloads (i.e., inference and reasoning at huge scale).

The Ironwood TPU is a seventh-generation chip designed for AI inference workloads, featuring 4,614 teraflops of peak compute, 192 GB of HBM memory, and 7.2 TB/s of memory bandwidth per chip. Scaled up to a pod of 9,216 chips, it delivers 42.5 exaflops of compute power and features a 1.2 TB/s bidirectional interconnect network. It also includes a specialized "SparseCore" to accelerate large-scale data tasks like those in recommendation systems.

Ironwood will become generally available towards the end of 2025 – customers will soon be able to test Google Cloud’s claims of its competitive performance for training AI and efficiently running inference on AI models.

The pros and cons of TPUs

Google has staked a lot of its AI dominance on TPUs, on which it heavily relies for training and inference its own flagship Gemini models. It argues that TPUs offer far greater cost per watt than graphics processing units (GPUs) such as those offered by Nvidia, with Ironwood pitched as a peer to the likes of Nvidia’s Blackwell chips.

“TPUs offer several compelling pros for modern AI development, particularly in deep learning. They are highly efficient for training and running large AI models like Large Language Models (LLMs),” Petr Baudis, CTO & chief AI architect at Rossum, tells ITPro.

One of their most significant advantages is the ability to scale thousands of chips together into ‘TPU pods,’ making them an ideal architecture for massive model training. When fully utilized, TPUs often provide a lower cost per unit of work (better $/training hour). Lastly, since Google controls the supply, users are less dependent on GPU shortages, which can be a critical advantage.

But TPUs aren’t better than GPUs across the board, as Baudis also explained.

“The primary limitation is that they are only available on Google Cloud, which results in significant vendor lock-in. They also have a smaller developer ecosystem compared to Nvidia's CUDA model, which is everywhere. Finally, TPUs work best on large, well-structured ML workloads and are not always ideal for smaller or irregular tasks."

TPUs vs GPUs designed by Nvidia, AMD, or Intel

In the past few years GPUs have become known as the go-to chip for AI training and inference, which makes Google’s reliance on TPUs all the more notable.

"TPUs are designed for AI tensor operations versus a GPU by Nvidia/AMD/Intel which is a more flexible architecture targeted at a broader set of applications (goes beyond AI) and will have general-purpose processing cores that are both smaller and in greater numbers than equivalent TPU cores," explains Alvin Nguyen, senior analyst at Forrester Research.

Using in-house chips also saves Google from the kind of supply chain limitations other hyperscalers face when it comes to acquiring in-demand GPUs.

“From a business standpoint, Google's TPU technology provides significant advantages, primarily centered on strategic independence,” Baudis tells ITPro. “It isn't a hostage to Nvidia's pricing or supply bottlenecks, and it can reliably offer AI capacity to customers even when GPUs are sold out elsewhere." But he cautions that in his opinion, while TPUs may offer better efficiency for LLM-specific workloads compared to GPUs, this is a smaller, variable factor that will vary quarter by quarter based on current hardware generations.

Nguyen adds that TPUs come with trade-offs but undeniably help Google to consolidate the control it has over its own AI ecosystem.

"Using TPUs allows Google to target the specific workloads they want to accelerate with an ASIC that they design and manufacture, providing them with more control over their infrastructure and tech stack (minimizes their dependence on NVIDIA and others for their technology, especially important with how in-demand GPUs are for AI)," he tells ITPro.

Google isn’t the only firm relying on its own silicon, however, with Baudis pointing to AWS’ has Trainium and Inferentia as well as Microsoft’s Azure Maia. These aren’t theoretical competitors either: Anthropic already use a mix of TPUs, Trainium, and Inferentia chips for their AI workloads.

Crucially, Nvidia still dominates the software ecosystem, CUDA remains the 'default language of AI compute.' Therefore, TPUs likely give Google an advantage mainly in cost, scale, and internal efficiencies, not total market control."

The future of TPU

Beyond the hardware, the broader strategic context around Google’s TPUs matters greatly. In October 2025, Anthropic announced it would access up to one million TPUs from Google Cloud, representing “well over a gigawatt” of compute capacity coming online in 2026. The deal is reportedly worth tens of billions of dollars.

“Google’s TPU strategy will age well if the world keeps moving towards giant AI models," says Baudis. "It supports huge, dense AI training at an extreme scale very well. The risks all seem relatively unlikely – either a shift towards very small models (while increasing intelligence demands) or a paradigm shift to a different technology than transformer-based neural networks. Smaller shifts in neural architectures, etc., that cannot be accommodated by existing TPUs can likely be absorbed by new incremental generations of the chip design. "

With regards to the Anthropic deal, Baudis adds that Anthropic will serve as a “demand anchor tenant” for Google, bringing several notable benefits.

“Firstly, it amortizes the costs associated with TPU datacenter build-outs,” he says. “Secondly, it validates the performance and efficiency of TPUs at the frontier scale required by a leading AI research company. Crucially, this partnership is expected to pull independent software vendors (ISVs) and enterprises toward the TPU software path, influencing their choice of models and tooling.”

Nguyen adds that the deal may help popularize Google’s TPUs, just as the hardware becomes more widely available. This could drive partners to optimize their AI services for Google Cloud more broadly. As enterprises look to avoid AI vendor lock-in, the TPU could become an ever more popular option.

A host of leading open weight AI models contain serious security vulnerabilities, according to researchers at Cisco.

In a new study, researchers found these models, which are publicly available and can be downloaded and modified by users based on individual needs, displayed “profound susceptibility to adversarial manipulation” techniques.

Cisco evaluated models by a range of firms including:

• Alibaba (Qwen3-32B)
• DeepSeek (v3.1)
• Google (Gemma 3-1B-IT)
• Meta (Llama 3.3-70B-Instruct)
• Microsoft (Phi-4)
• OpenAI (GPT-OSS-20b)
• Mistral (Large-2)

All of the aforementioned models were put through their paces with Cisco’s AI Validation tool, which is used to assess model safety and probe for potential security vulnerabilities.

Researchers found that, for all models, susceptibility to “multi-turn jailbreak attacks” was a key recurring issue. This is a method whereby an individual can essentially force a model to produce prohibited content.

This is achieved by using specifically-crafted instructions from the user that, over time, can be used to manipulate the model’s behavior. This is a more laborious process than “single-turn” techniques, which involve manipulating a model with a single effective malicious prompt.

Multi-turn jailbreak techniques have been observed in the wild before, particularly with the use of the Skeleton Key method, which allowed hackers to convince an AI model to produce instructions for making a Molotov cocktail.

Success rates with individual models varied wildly, the study noted. Researchers recorded a 25.86% success rate with Google’s Gemma-3-1B-IT model, for example, while also recording a 92.78% success rate with Mistral Large-2.

Researchers also recorded the highest success rate for single-turn attack methods with both these models.

Different strokes for different folks

The varied success rates recorded by Cisco lie in how these models are typically used, researchers noted. This rests on two key factors: alignment and capability.

In the case of 'alignment', this refers to how an AI model acts in the context of human intentions and values. 'Capability', meanwhile, refers to the model’s ability to perform a specific task.

For example, models such as Meta’s Llama range, which place a lower focus on alignment, showed the highest susceptibility to multi-turn attack methods.

Researchers noted that this is because Meta made a conscious decision to place developers “in the driver seat” in terms of tailoring the model’s safety mechanisms based on individual use-cases.

“Models that focused heavily on alignment (e.g., Google Gemma-3-1B-IT) did demonstrate a more balanced profile between single- and multi-turn strategies deployed against it, indicating a focus on “rigorous safety protocols” and “low risk level” for misuse,” the study said.

AI model flaws have real-world implications

Researchers warned that flaws contained in these models could have real-world ramifications, particularly with regard to data protection and privacy.

“This could translate into real-world threats, including risks of sensitive data exfiltration, content manipulation leading to compromise of integrity of data and information, ethical breaches through biased outputs, and even operational disruptions in integrated systems like chatbots or decision-support tools,” the study noted.

Notably, in enterprise settings, they warned these vulnerabilities could “enable unauthorized access to proprietary information”.

Concerns over AI model manipulation have become a common recurring theme since the advent of generative AI in late 2022, with a steady flow of new jailbreak techniques emerging on a regular basis.

MORE FROM ITPRO

• Researchers swerved GPT-4's safety guardrails and made the chatbot detail how to make explosives in Scots Gaelic
• What good AI cyber security looks like today
• AI security blunders have cyber professionals scrambling

Google has hit back at claims of a massive Gmail breach, suggesting recent reports stem from a “misunderstanding of infostealer databases”.

In a post on X, the tech giant insisted Gmail's defenses are “strong, and users remain protected” in a bid to calm customer concerns.

“The inaccurate reports are stemming from a misunderstanding of infostealer databases, which routinely compile various credential theft activity occurring across the web,” Google said.

“It’s not reflective of a new attack aimed at any one person, tool, or platform.”

The company’s sharp response comes after a host of publications reported that 183 million Gmail accounts had been compromised in what one outlet described as a "huge data breach".

Speculation over a breach mounted after a sizable dataset was added to the Have I Been Pwned database, which allows web users to check if their credentials have been compromised.

In a blog post dissecting the dataset, which was collated by researchers at Synthient, creator Troy Hunt noted these records were the culmination of several years’ worth of infostealer activity, instead of a newly discovered breach.

Regardless, several outlets pounced on the situation. In response to a comment under his original blog post, Hunt echoed Google’s response to the coverage.

“I think they're deliberately misleading and designed to drive eyeballs on ads whilst the truth gets buried somewhere further down in the story,” he wrote.

Gmail breach claims are a dime a dozen

This isn’t the first time Google has been forced to push back against reports of a Gmail breach. Indeed, it’s the second in the space of three months.

In late August, reports of a major incident that impacted some 2.5 billion Gmail accounts began circulating online, once again prompting a statement from the tech giant in which it refuted the claims.

In a blog post, Google described reports as “entirely false” and said there was no risk to users.

"We want to reassure our users that Gmail’s protections are strong and effective,” the company said.

“Several inaccurate claims surfaced recently that incorrectly stated that we issued a broad warning to all Gmail users about a major Gmail security issue. This is entirely false.”

MORE FROM ITPRO

• Warning issued over critical flaws spotted in TP-Link routers
• CISA issues alert after botched Windows Server patch exposes critical flaw
• Enterprises can’t keep a lid on surging cyber incident costs

Commercially-viable, scalable quantum computing could be just a few years away according to Sundar Pichai, CEO at Alphabet,

Speaking with Salesforce CEO Marc Benioff during a fireside chat at Dreamforce 2025, Pichai stated he was “bullish” about the potential of quantum computing in the next ten years, suggesting it could be realized even sooner.

In a long conversation on the future of both quantum and AI, Pichai highlighted the achievements of Google DeepMind as well as the Nobel Prize awarded to its CEO Demis Hassabis and director John M. Jumper.

“We believe in that kind of deep, fundamental R&D with a view of translating real problems,” Pichai said.

“That same quantum computing team, no different from our TPU team, plans to have commercially-available, real, at scale quantum computers in a few years. And I think they will get there.”

In 2024, Google announced its proprietary quantum chip Willow had demonstrated usable error reduction for quantum computing, a major step for achieving scalable quantum computing.

Just this week, Google announced that Willow had run a test algorithm 13,000 times faster than the best corresponding classical algorithm, run on a top supercomputer.

Benioff pointed out that quantum computing could undermine all cryptography, hinting at a need for post-quantum cryptography in the wake of any significant quantum advancements.

“I think in three five years’ time, we will have that moment where from a cryptographic standpoint we have to adapt to quantum, for sure,” Pichai stated, adding that people will need to work together on establishing safeguards.

Google isn’t the only company chasing a quantum leap with its own chip design. In February, Microsoft hailed its Majorana 1 quantum processing unit (QPU) as the first such chip of its kind to support a million qubits at once.

But it’s not all consensus. At the start of the year, Jensen Huang, CEO at Nvidia, poured cold water on quantum optimism with a prediction that “truly useful” quantum computing could be 20 years away, prompting stock plunges at companies such as D-Wave and Rigetti Computing.

Huang later walked back his comments and Nvidia has since committed to further quantum research – though continues with its no-chip quantum chip strategy.

ChatGPT reflections

Elsewhere in the conversation, Pichai reflected on the days immediately after the launch of ChatGPT and how Google responded to claims that it had a major competitor in AI and search for the first time in years.

He pointed out that Google began using transformers in 2017, following the publication of its paper Attention Is All You Need, and had made strides with internal chatbots such as BERT then Language Model for Dialogue Applications (LaMDA) – the latter of which drew headlines after an engineer went public with claims that it was sentient.

When ChatGPT launched, Pichai said, Google had been sitting on its chatbot model over quality concerns but could easily have beaten OpenAI to it.

“I would argue pretty much all of us working on it, we knew in a different world we would have launched our chatbot maybe a few months down the line,” he said.

“We hadn’t quite gotten it to a level where you could put it out and people would be ok with Google putting out that product.”

He also dismissed claims that he had been alarmed or panicked by the launch, with contemporary reports suggesting that Google initiated a state of ‘code red’ and scrambled developers internally to compete with OpenAI.

“For me when ChatGPT launched, contrary to what people outside felt, I was excited because I knew the window had shifted,” he said.

“We had been building this technology for so long, we were so AI-native, everything in the company, every decision we had made. I had decided to take a full-stack approach in AI, we were investing all the way from infrastructure, we built our own chips, to world-class research teams in Google Research, Google Brain, Google DeepMind.”

MORE FROM ITPRO

• Google CEO says more than 25% of the company's code is now AI-generated
• Nearly half of enterprises aren't prepared for quantum cybersecurity threats
• Sundar Pichai says DeepSeek has done ‘good work’ showcasing AI model efficiency

A raft of AI models from OpenAI, Anthropic, and Google have been officially cut from the GitHub Copilot service.

First announced in late September, users have had ample time to prepare for the deprecation cut-off date, which was confirmed in a blog post on 23 October.

The move comes as the company looks to usher users away from older models toward up-to-date options from the aforementioned providers.

“As part of our commitment to a fast, high‑quality Copilot experience, we regularly evaluate and retire older models in favor of newer, more capable ones,” the company said in September.

Anthropic models including Claude Sonnet 3.7, Claude Sonnet 3.7 Thinking, and Claude Opus 4 will be deprecated as part of the move while Gemini 2.0 Flash will be scrapped.

OpenAI models set for the chopping block include:

• GPT o3
• GPT o1 mini
• GPT 03 mini
• GPT 04 mini

What does this mean for GitHub users?

So what does this mean for users? Essentially, their use across “all GitHub Copilot experiences” will be removed, the company noted. This spans Copilot Chat, ask and agent modes, inline edits, and code competition tasks.

As a result of the move, GitHub has advised users to update workflows and integrations to use supported models.

“Copilot Enterprise administrators may need to enable access to alternative models through their model policies in Copilot settings,” GitHub said.

“As an administrator, you can verify availability by checking your individual Copilot settings and confirming that the policy is enabled for the specific model.”

GitHub added that no action is required on the part of admins to remove deprecated models.

If you’re wondering which alternative models are best following the deprecation cut-off date, GitHub has you covered.

Here’s a handy list of options:

MORE FROM ITPRO

• Everything you need to know about GitHub Models
• What Thomas Dohmke’s departure means for GitHub
• GitHub just unveiled a new AI coding agent for Copilot

Researchers have uncovered a new class of Android attack based on a years-old data-stealing browser attack method.

The technique, dubbed Pixnapping by US-based researchers, allows attackers to stealthily steal sensitive information displayed by other apps or even websites without users ever realizing their data has been compromised.

This data can include two-factor authentication (2FA) codes, private messages, and even financial information.

In proof-of-concept tests, the team was able to recover sensitive information from widely used apps and websites such as Signal, Venmo, Google Authenticator, Gmail, Google Maps, and Google Accounts.

Notably, the malicious app was able to steal 2FA codes from Google Authenticator in less than 30 seconds, all without requiring any Android permissions or displaying suspicious activity to the user.

Pixnapping allows a malicious Android app to 'snap' pixels from other apps or websites by exploiting Android APIs and a GPU hardware side channel known as 'GPU.zip', which leaks information about how the graphics hardware processes visual data.

How Pixnapping attacks work

Pixnapping works in a three-stage process, with the first being invoking a target app, such as Google Authenticator, to cause sensitive information to be submitted for rendering.

The next step involves inducing graphical operations on individual sensitive pixels rendered by the target app - for example, the pixels that are part of the screen region where a 2FA character is known to be rendered by Google Authenticator.

A side channel such as GPU.zip is then used to steal the pixels operated on during step two, one pixel at a time. Steps two and three are repeated for as many pixels as needed to run optical character recognition over the recovered pixels and recover the original content.

The researchers demonstrated successful attacks on modern Google and Samsung phones, including the Pixel 6 through Pixel 9 ranges, and the Galaxy S25, running Android versions 13 through 16.

However, because the core mechanisms used by Pixnapping are typically available in all Android devices, the vulnerability is likely to affect a wide range of smartphones across manufacturers.

“Conceptually, it is as if any app could take a screenshot of other apps or websites without permission, which is a fundamental violation of Android’s security model,” said researcher Riccardo Paccagnella, assistant professor in Carnegie Mellon University’s Software and Societal Systems Department.

Initial patches failed to solve the issue

In February, the research team disclosed its findings to Google, which rated Pixnapping High Severity and began tracking it as CVE-2025-48561.

While Google attempted to mitigate the issue by restricting access to certain APIs, the research team said it later discovered a workaround that restored the attack’s effectiveness.

There's no evidence that the vulnerability has been exploited in the wild. However, as of October 13, researchers said Android remained vulnerable – although Google plans to issue an additional patch in its December Android security bulletin.

"Fixing Pixnapping will likely require changes to core Android mechanisms, for example, by allowing apps to prevent other apps from drawing over their sensitive content," said Paccagnella.

To prevent Pixnapping, the researchers recommend that users keep their Android devices updated with the latest patches as soon as they are released.

They plan to release Pixnapping’s source code once effective patches are available, to support further academic research and industry defenses.

MORE FROM ITPRO

• DNS Security 101: How to safeguard your business from cyber threats
• Why this Cybersecurity Awareness Month is particularly urgent
• How the channel can help enterprises short up supply chain security

Microsoft has announced plans to invest £22 billion ($30bn) across the UK in a bid to drive development of the country’s AI infrastructure.

The investment, which represents the largest financial commitment the tech giant has ever made in the UK, includes $15bn in funding to expand cloud and AI infrastructure.

In a blog post outlining the pledge, Microsoft president Brad Smith revealed this will also include investment to build the country’s largest supercomputer, boasting more than 23,000 Nvidia GPUs.

Development of the supercomputer will be led in collaboration with Nscale.

In a separate statement, Microsoft CEO Satya Nadella said the move builds on long-standing ties between the tech giant and the UK.

Nadella said the company has around 6,000 employees in the UK alongside multiple data center regions and "some our our most important AI and research labs".

"Microsoft has proudly invested and operated in the United Kingdom for more than four decades, opening our first office outside the United States here in 1982," he said.

"Our commitment to that partnership is stronger than ever. Just two years ago we announced a $3.2 billion investment in the United Kingdom’s AI infrastructure," Nadella added. "Today we’re doubling down on our investment."

Prime minister Keir Starmer welcomed the announcement, adding that the funding pledge is a “powerful vote of confidence” in the UK’s burgeoning technology sector.

“This commitment will not only strengthen our digital infrastructure and support thousands of highly skilled jobs, but also ensure Britain remains at the forefront of global innovation as we deliver on our Plan for Change,” he said.

A big week for British tech

The announcement from Microsoft comes amidst a big week for the British technology sector and coincides with US president Donald Trump’s second state visit to the UK.

A host of major technology companies have all made significant funding commitments in recent days.

Google, for example, unveiled the opening of a new $1 billion data center at Waltham Cross on the outskirts of London as part of a sweeping £5 billion investment scheme across the country.

Funding from the tech giant aims to build out cloud and AI infrastructure alongside investment in research and development (R&D).

Meanwhile, Nvidia announced plans to invest upwards of £11 billion on 16 September. Working alongside partners like Nscale and CoreWeave, the chipmaker said it plans to create Europe’s largest GPU cluster in the UK, underpinned by 120,000 Blackwell GPUs.

MORE FROM ITPRO

• How Satya Nadella's first decade at Microsoft took it from tech giant to undisputed industry titan
• Microsoft saved $500 million by using AI in its call centers last year
• Microsoft says AI is finally having a 'meaningful impact' on developer productivity

Google has officially opened the doors of its $1 billion data center in the UK, part of a £5 billion investment into local infrastructure and the tech giant's first data center in the country.

The opening of the data center in Waltham Cross, in Broxbourne on the edge of London, comes as US President Donald Trump is set to arrive in the UK for a state visit that is expected to include a series of deals between the US and UK on technology and nuclear power.

Google said in a statement that the "state of the art" data center would be used to help meet demand for AI-powered services including Google Cloud, Workspace, Search, and Maps.

The £5 billion investment includes data centers like the one at Waltham Cross, as well as Google's capital expenditure, R&D, and AI-related engineering over the next two years in the UK, including DeepMind, the British-founded AI science company bought by Google in 2014.

"Google's investment in technical infrastructure, expanded energy capacity and job-ready AI skills will help ensure everyone in Broxbourne and across the whole of the UK stays at the cutting-edge of global tech opportunities," said Ruth Porat, President and Chief Investment Officer at Alphabet and Google, in a statement.

Inside Waltham Cross

Google bought the 33-acre site in Waltham Cross in 2020, saying last year that the project "represents an investment of $1 billion dollars in the country".

Technical details about the data center have been sparse, but Google has been keen to share information about energy plans – no wonder given the sensitivity around energy and water use by data centers.

Earlier this week, The Guardianreported that planning application documents for Google's next UK data center, in Thurrock in Essex, shows it will take up 128 acres and is expected to emit more than half a million tonnes of carbon dioxide a year – equivalent to 500 short-haul flights a week.

For the newly-opened Waltham Cross site, Google said it was designed to "minimize" environmental impact, including by using more advanced air cooling technology rather than traditional water-based cooling systems.

"The facility uses advanced air-cooling technology to limit water usage to domestic use and is also equipped to support off-site heat recovery, meaning heat from the data center can be re-routed and provided free of charge to help warm local homes, schools or businesses," the company said in a statement.

Google has also signed deals with local energy providers to encourage renewables, including a previously announced partnership with Scotland's ENGIE for wind generated energy as well as teaming up with Shell to use battery storage systems to help address intermittency issues with renewable sources.

Last year, Google said UK operations were "on track to operate at or near 90% carbon-free energy in 2025." This year, Google has adjusted its goal.

"Between the Shell alliance and Google's other clean energy initiatives, Google's UK operations are projected to run at or near 95% carbon-free-energy in 2026."

Keeping locals on board

The rise of AI has sparked a wave of data center construction, which isn't always welcome in locations where construction takes place.

In Waltham Cross, locals last year complained that noise from the construction site was keeping them awake at night, ruining local views and lowering house prices. The Daily Mail quoted one resident as saying: "I moved here for peace and quiet… It won't benefit me. It won't benefit locals."

Google said it had set up a community fund to be managed by Broxbourne Council to help support local economic development, and was "providing direct support" for local charities and social enterprises.

Councilor Corina Gander, Leader of Broxbourne Council, said in a statement provided by Google: "The Council has worked closely with Google to maximise the positive impact of the development. Google is heavily investing in community-based projects and is making an important contribution to the local economy."

MORE FROM ITPRO

• Majority of English data centers use less water than a 'typical leisure center' as operators embrace new cooling methods
• $200 billion to build a data center? Soaring AI demands mean costs could skyrocket within a few years
• Plans announced to resurrect former steelworks as a ‘green’ data center

The UK government has revealed its push to encourage AI use is delivering marked benefits for developers.

As part of an AI trial across government, more than 1,000 tech workers across 50 different departments trialed AI coding assistants from Microsoft, GitHub Copilot, and Google Gemini Code Assist between November 2024 and February this year.

Figures published following a review of the scheme shows developers are saving around one hour each day, equivalent to around 28 working days a year.

Technology minister Kanishka Narayan said the trial scheme highlights the benefits of rolling the technology out across government.

"For too long, essential public services have been slow to use new technology – we have a lot of catching up to do," said Narayan.

"These results show that our engineers are hungry to use AI to get that work done more quickly, and know how to use it safely."

Most of the time savings from the AI assistants came from using them to write first drafts of code that experts then edit, or using them to review existing code.

With only 15% of code generated by AI assistants being used without any edits, engineers have been taking care to check and correct outputs where needed, the government said.

The trial appears to have been popular with government coders, with 72% of users agreeing they offered good value for their organization. Nearly two-thirds (65%) reported that they were completing tasks more quickly and 56% said they could solve problems more efficiently.

Notably, more than half (58%) of participants said they would prefer not to return to working without these solutions.

Tara Brady, resident of Google Cloud EMEA, said the tech giant is “thrilled to see the positive impact” its AI coding tool delivered for government workers.

"This landmark trial, the largest of its kind for Gemini Code Assist in the UK public sector, underscores the transformative potential of AI in enhancing productivity and problem-solving for coding professionals, and highlights the successful collaboration stemming from Google Cloud’s Strategic Partnership Agreement with the UK government."

Prime minister Keir Starmer has been highly vocal about the government’s plans to roll out AI across government and public services since taking office. Downing Street hopes to save taxpayers more than £45 billion through the use of AI.

Questions remain over government AI coding gains

Martin Reynolds, field CTO at software delivery platform Harness, welcomed the move but questioned whether the plans go far enough. A key factor here, he noted, lies in the volume of manual remediation required by developers using AI-generated code.

"While AI is creating an initial velocity boost, 85% of government AI-generated code still needs to be manually edited by engineers,” he said.

“That's before it enters the more manual downstream stages of delivery, such as testing, security scanning, deployment, and continuous verification, which are essential to getting code into production safely and reliably," Reynolds added.

The quality of AI-generated code has become a recurring talking point in recent months. A recent study from Fastly, for example, found developers often find themselves manually remediating faulty code, which ultimately negates the time savings delivered by the technology.

Nigel Douglas, head of developer relations at software supply chain security firm Cloudsmith, also voiced concerns about potential security issues, saying there's not much evidence of secure-by-design thinking.

Given the critical nature of the work conducted by developers in key government departments, this should be a key focus moving forward.

“Without security-aware tooling or policy enforcement, you can easily see over-enthusiastic use of AI coding assistants unknowingly introducing vulnerabilities into one of this country’s most critical software ecosystems," he said.

"We’re getting past the point where it’s acceptable for software development teams to ‘hope for the best’ - you’ve got to be able to verify the provenance of the ingredients flowing through your software supply chain and into production systems, and you need tools to help respond to newly emerging threats that may impact what you’ve already deployed.”

MORE FROM ITPRO

• Not all software developers are sold on AI coding tools
• AI coding tools are booming – and developers in this one country are by far the most frequent users
• AI coding tools are finally delivering results for enterprises

Google Cloud has announced the launch of a new multi-cloud data transfer service for enterprises based in the UK and European Union (EU).

In a blog post Jeanette Manfra, senior director for global risk and compliance at Google Cloud, said the new Data Transfer Essentials scheme comes in direct response to cloud interoperability principles outlined in the EU Data Act.

“Data Transfer Essentials is a new, simple solution for data transfers between Google Cloud and other cloud service providers,” Manfra said.

Notably, under the terms of the EU Data Act, cloud providers are allowed to “pass through costs to customers” for data transfers. Manfra noted that the Google service will come “at no cost to customers”.

Key competitors including Microsoft and Amazon Web Services (AWS) also provide data transfer options for customers. In Microsoft's case these come at a cost to the enterprise, while AWS allows EU customers to transfer data at a reduced rate.

“The original promise of the cloud is one that is open, elastic, and free from artificial lock-ins,” Manfra wrote.

“Google Cloud continues to embrace this openness and the ability for customers to choose the cloud service provider that works best for their workload needs.”

What to expect with Google Cloud's data transfer scheme

According to Google, the new service is specifically designed for “in-parallel” processing of workloads distributed across multiple cloud providers, giving enterprises greater flexibility.

“Data Transfer Essentials enables you to build flexible, multi-cloud strategies and use the best-of-breed solutions across different cloud providers,” Manfra said.

“This can foster greater digital operational resilience – without incurring outbound data transfer costs from Google Cloud.”

Data transfer costs, also known as “egress fees”, have become a point of contention in the cloud computing industry in recent years.

With enterprises flocking to hybrid and multi-cloud strategies, the costs associated with transferring workloads to alternative providers have been a major hurdle.

Indeed, egress fees were a key talking point in a recent probe into the state of the UK’s cloud computing industry. The investigation, conducted by the Competition and Markets Authority (CMA) following an initial probe by Ofcom, found egress fees created significant barriers to multi-cloud adoption.

“Customers face both commercial and technical barriers when seeking to multi-cloud or switch their cloud provider and many currently think that the costs outweigh the benefits,” the CMA reported in early August.

“Barriers to multi-cloud negatively affect many customers’ ability to use and integrate multiple public clouds, and this limits customers’ ability and incentive to exercise choice over their cloud provider."

MORE FROM ITPRO

• Say goodbye to walled gardens, Oracle is doubling down on multi-cloud
• What is 'multi-cloud by design'?
• How is hybrid cloud security different from multi-cloud or single cloud security?