AMD and AI infrastructure company Cerebras have unveiled a joint system integrating AI infrastructure from both companies.

The joint solution, which was announced at AMD Advancing AI 2026, features AMD Helios – the star product of the show – and the Cerebras Wafer-Scale Engine.

According to Cerebras and AMD, it's intended to solve the issue of inference workloads having variable requirements when it comes to latency, throughput, token capacity, cost, and scale.

For example, agents and copilots need faster response times, whereas high-volume workloads prioritize maximum code generation, the organizations said.

By unifying Helios and Wafer-Scale Engine, the system optimizes the two stages of workflow independently: Helios is tuned to process prompts and large context windows, providing ultra-high throughput.

Wafer-Scale Engine, meanwhile, accelerates memory and bandwidth-intensive token generation with ultra-low latency.

“The demand for ultra-fast inference is growing at an unprecedented pace,” said Cerebras CEO and co-founder Andrew Feldman. “Partnering with AMD gives us an incredible opportunity to bring that performance to even more customers.”

AMD CEO Lisa Su added: “AI inference is becoming one of the largest infrastructure opportunities in AI and its growing diversity requires a more flexible approach.”

“Together with Cerebras, we are extending that leadership into the most latency-sensitive applications and creating a powerful new platform for real-time agentic AI.”

The joint solution will be available in the second half of 2026, initially through Cerebras Cloud.

The announcement comes alongside a number of other Helios partner announcements, including new commitments from Microsoft and Anthropic to roll out the system to their infrastructure.

FOLLOW US ON SOCIAL MEDIA

AMD has launched its latest Instinct GPUs, the MI455X, in tandem with bringing its Helios AI architecture into production at AMD Advancing AI 2026.

Helios, which was announced at last year’s annual conference, is the company’s flagship hardware product, integrating Epyc CPUs, Instinct GPUs, Pensando networking, and Rocm software stack.

“Helios is really the system that scales to be an AI factory,” said Andrew Dieckman, CVP and GM of data center and GPU at AMD. “We set out to build the world’s highest-performing rack and we believe we have delivered the world’s highest performing rack.”

According to AMD, Helios has 15% more AI compute, 50% more memory, and 50% more scale-out bandwidth than Nvidia’s competitor offering, Vera Rubin NVL72.

At the heart of Helios – the “engine” in Dieckman’s words – is the new AMD Instinct MI455X CPU. The MI400 series features increased HBM4 capacity and bandwidth, expanded AI formats with higher throughput, and standard-based rack-scale networking.

In Helios, the MI455X acts as the accelerator, delivering 40 petaflops of peak FB4 compute, with 423GB of HBM core and a peak memory bandwidth of 23.3 terabytes per second of memory per GPU.

In addition to the new MI455X GPU, Helios also features 6th Gen AMD Epyc server CPUs with 96 high-frequency cores and 1.6TB/s of memory bandwidth, as well as AMD Pensando and UALoE networking fabric.

This can scale up to 260TB/s bandwidth and host 72 GPUs on one domain.

AMD showcases Helios partners

With Helios already in production, AMD was keen to show off some of its partners that are deploying – or pledged to deploy – the system. These include OpenAI, Meta, Oracle, Microsoft, Dell Technologies, HPE, Lenovo, Bull, Cisco, Tensorwave, and more.

“Helios is a big step forward for AMD because we’re moving to a full rack-scale architecture where we provide all the components in a really co-designed engine and co-engineered rack, so that [customers] can get the full benefit out of the complete AMD end-to-end solution,” said Dieckman.

“We continue to be very focused on driving improved economics for our customers – making sure that we deliver the most tokens per dollar within the market.”

FOLLOW US ON SOCIAL MEDIA

Microsoft has announced it will deploy AMD’s Helios rack-scale solution as it targets data processing and AI inference gains for Azure customers.

The collaboration will see three new AMD EPYC CPU-powered virtual machine (VM) offerings made available through Azure. These include:

• HDv2 VMs
• HCx2 VMs
• ND MI455X v7 VMs

The first of these, Azure HDv2, is designed for large-scale data processing and to “empower massive agentic workload adoption”.

Featuring 500 sixth-generation EPYC CPU cores, HDv2 VMs boast 4TB of RAM, 32TB of local NVMe storage, and 500Gbps Azure Boost networking capabilities.

Azure HCx2 VMs, meanwhile, are designed primarily for silicon design and technical computing tasks, such as scientific simulations and engineering analysis.

This series builds on the design of Azure HX virtual machines, which the company launched in partnership with AMD in 2023.

HXv2 VMs will include 176 AMD sixth-gen EPYC CPU cores. These offer “significantly increased” per-VM and per-core performance and will also include 800GB InfinBand for “large-scale MPI-based simulations”.

Elsewhere, the ND MI455X v7 specifically targets “production-scale” AI inference improvements for Azure customers.

“ND MI455X v7 is designed for the reasoning, search and agentic workloads behind modern AI services,” Microsoft said in a blog post.

“Powered by the AMD Helios rackscale solution, it expands Azure’s infrastructure options for large-scale inference and is designed to deliver strong performance and efficiency for demanding AI workloads.”

The move makes Microsoft the latest in a string of big tech customers flocking to the Helios system. AMD chief executive Lisa Su said the deal builds on long-standing ties between the duo.

“AMD and Microsoft have spent years building high-performance infrastructure together, and today we're extending that partnership across the full stack of AMD AI solutions on Azure,” said AMD CEO Lisa Su.

“Microsoft's new AMD deployments mark an important milestone as we deliver leadership compute solutions to Azure customers and scale the next generation of AI infrastructure together.”

Microsoft targets “scale and choice” for customers

Helios is AMD’s first rack-scale system designed specifically for frontier model training and large-scale AI inference. The system combines AMD Instinct MI455X GPUs, AMD EPYC “Venice” CPUs, Pensando networking, and ROCm software..

Microsoft CEO Satya Nadella said the collaboration marks a significant expansion of Azure infrastructure capabilities.

“Customers are looking for AI infrastructure that is optimized for a wide range of workloads, from training and inference to data preparation, search, and reinforcement learning," he said.

"Through our collaboration with AMD, we are expanding the Azure infrastructure portfolio with AMD Helios to give customers the performance, scale and choice they need to build and run the next generation of AI applications.”

Helios roll-out pending

AMD unveiled the Helios rack-scale solution at its 2025 Advancing AI conference. The solution offers an alternative to – and competes with – Nvidia’s Vera Rubin and Grace Blackwell systems.

At the time, Su described the launch of Helios as a “game changer” that’s “purpose built for the most demanding AI workloads”.

A host of major tech providers, including Meta, OpenAI, HPE, and Oracle have since confirmed they’ll deploy the solution in data centers.

The chipmaker will begin shipping Helios to customers in the second half of 2026.

Although exact timelines haven’t been revealed, the collaboration with Microsoft comes just days ahead of AMD’s 2026 Advancing AI conference, in which customers can likely expect real-world examples of the system in action.

ITPro will be live on the ground in San Francisco for AMD’s annual conference this week.

Stay up-to-date with all the news, views, and announcements from the event by following our socials (below) and subscribing to our daily newsletter.

FOLLOW US ON SOCIAL MEDIA

AMD Advancing AI 2026 takes place this week at the Moscone Center in San Francisco, with the two-day event largely focusing on the second day, when we will hear from AMD’s CEO Lisa Su.

One notable thing, for me, is a sense that in many ways AMD and the wider chip market is in largely the same position as it was this time last year. Nvidia is still the talk of the town when it comes to AI infrastructure, with CEO Jensen Huang once again guesting at other hardware vendors’ flagship events.

AMD still has HPC pedigree (more on that below) but doesn’t have the cachet Nvidia does, while Intel trails further behind.

Despite this, there are still predictions to be made, so here are three things I expect to see at AMD Advancing AI 2026 and one I don’t.

Helios

AMD Helios, the company’s fully-integrated, rack-scale architecture solution, was unveiled at Advancing AI 2025 but we are yet to see any deployed (at least publicly).

This is despite HPE saying it would have a cluster available in its cloud in the second half of 2026 (for those counting, that’s now) and Oracle stating it will build its AI superclusters on Helios.

What better time to talk about Helios being out in the market than at Advancing AI?

My feeling is it will build on last year’s announcement by wheeling out some real-world examples of the hardware in use, whether as part of the HPE or Oracle partnerships or something else.

Altair MI450 chips

Annual company events in the tech industry are for product launches as much as anything else. Having already made several announcements about its gaming portfolio at Computex 2026 in May, we can probably discount major announcements around Ryzen or Radeon – although there may be a recap.

Instead, one of the flagship announcements could be the company’s highly anticipated next-gen GPU series the MI450 – code named Altair.

Oracle already announced in October 2025 that it would be putting MI450 into OCI once it’s available and rumour has it the company has already inked a deal to sell these chips to Anthropic. Speaking of which…

AI partnerships

AMD – perhaps more than other hardware companies – makes a lot of its relationship with AI businesses.

This includes the likes of Anthropic and OpenAI, with the latter’s CEO, Sam Altman, appearing on stage at Advancing AI last year, as well as Hugging Face, PyTorch, and Llama.

I also expect to hear more about what role AMD expects to play in the brave new world of agentic AI

Supercomputing and HPC

While we might hear about high performance computers (HPC) – for example Alice Recoque, which is supposed to start onboarding this year – there may be little conversation around supercomputing more generally.

This is because while an AMD machine has been ranked number one in the Top500 list of most powerful supercomputers consistently since 2022 and held the top two places since November 2024, it was overtaken by a Chinese supercomputer in the June 2026 rankings.

AMD Advancing AI 2026 takes place from 22-23 July, with CEO Lisa Su’s keynote happening on the morning of the second day. I will be on the ground reporting on all the news coming out of the event and providing the usual dose of analysis you have come to expect.

Follow here for all previous years’ coverage and everything from this year as it comes in.

Achieving true tech sovereignty could be a bridge too far for Europe, a senior Forrester analyst has told ITPro, as new research from the consultancy lays bare the scale of the challenge.

According to Forrester’s Global Sovereignty Forecast, the shift to bolster tech sovereignty will move “slowly over the next five years”, with many efforts to divest from US tech providers falling flat.

The report includes a ‘Tech Sovereignty Index’ used to measure a country’s ability to “develop, operate, and secure critical technologies independently” - and it paints a damning picture.

Between 2025 and 2030, European countries are only expected to make “modest” gains in building sovereignty. Germany’s score is expected to rise from 34% to just 36%, for example, while the UK from 30% to 32%.

This slow progress flies in the face of what has become a key recurring talking point in Europe over the last 18 months. Calls have been rising to cut reliance on US-based technology providers, largely due to a combination of rising geopolitical tensions and data privacy concerns.

Speaking to ITPro, Dario Maisto, principal analyst and lead author of the report, suggested the current focus on sovereignty is much like other tech industry trends: awash by hype.

“There is a lot of hype around sovereignty, because service providers see it as the money maker,” he told ITPro. “So why should an organization engage with me? Well, sovereignty.”

For many organizations, Maisto noted that the focus on sovereignty isn’t actually required from a compliance perspective, albeit in some limited capacities.

“I was watching a Red Hat presentation in the morning, and they said there are a lot of new regulations. No, there's not. There's none. There is no new regulation that mandates sovereignty in any way whatsoever,” he said.

“Nor is there any definition at any possible level, national, regional, international, for sovereignty. There are some requirements in terms of data residency, citizenship of people handling certain workloads, certain workloads being deployed on on-prem rather than public cloud … but there is no formal, overarching obligation towards sovereignty whatsoever.”

“Persistent dependencies” run deep in Europe

Maisto admitted that geopolitical tensions, particularly deteriorating transatlantic ties, have indeed prompted a rethink of how heavily European organizations rely on US providers.

The idea of a wholesale shift away from popular vendors would require a monumental effort, however. Combined, Amazon Web Services (AWS), Microsoft Azure, and Google Cloud control 65% of the European cloud market.

“There is not much that we can do about it. Certain SaaS applications, for example, do not even work outside of the hyperscalers clouds,” he said. “So you have two options as a client: either take the degraded, not complete, less serviced, less updated on-prem version, which you can, or you get the most updated, the fanciest, the shiniest cloud version.”

“But then you are forced to deploy on a hyperscaler because the APIs don’t work outside of the hyperscaler cloud. Moving these applications outside of the hyperscaler cloud, it's not a lift and shift.

“It's not something that the client can do.”

Elsewhere, Europe simply doesn’t have the capabilities when it comes to macroeconomic considerations. When gauging sovereignty credentials, Forrester’s index takes into account a range of factors, including:

• Government spending
• Tech workforce availability
• Data center capacity and autonomy
• Chip production
• Rare earth elements processing
• AI model creation

Forrester noted that “persistent dependences” in areas such as energy infrastructure, data center infrastructure, and supply chains could hamper the region’s ability to operate independently of both US and Chinese technology ecosystems.

Back down to earth

Maisto’s argument is that the reality of achieving true tech sovereignty often fails to acknowledge the significant investment required to reach this point – along with the potential disruption and legwork needed to move away from US-based providers.

“I think seven out of 10 international SaaS solutions are out of the US,” he said. “So how feasible is it that organizations replace 20 years of investment in Salesforce, in Workday, in ServiceNow overnight?”

Put simply, it’s an idealized vision borne out of the reactionary, tumultuous geopolitics at play in the world right now.

“So many people see sovereignty as a green field,” he said. “For organizations that have been investing millions, if not billions, in cloud migrations, task transformation, it's not about greenfield. It's a retrofit.”

These questions are what prompted the creation of what Maisto described as the “minimum viable sovereignty model” used in the Tech Sovereignty Index.

“How do I do the least possible investments now to get the sovereignty that I actually need, what is it I actually need to be compliant? Maybe to mitigate a risk, and what is it I can actually achieve?”

Lifting and shifting anyway

Despite the difficulties highlighted by Maisto, efforts to bolster sovereignty have been gaining traction across Europe over the last year. In April, authorities in France announced plans to move away from Windows in favor of Linux-based operating systems for some 2.5 million government devices.

Speaking at the time, David Amiel, minister of public action and accounts, said the country aims to become “less reliant on American tools and regain control of our digital destiny”.

At an EU-wide level, similar efforts are being made. As ITPro reported in June, the European Commission unveiled the technological sovereignty package.

This aims to reduce reliance on US tech and shake up the union’s capabilities across several areas, spanning AI, cloud computing, open source, and semiconductor manufacturing.

On the latter of these focus areas, Maisto noted that this is a “move in the right direction because we need more options” in terms of hardware manufacturing.

At present, Europe accounts for less than 10% of global chip manufacturing, according to Forrester. Crucially, it lacks the “large designers that can compete with the likes of Nvidia or Qualcomm”.

All told, European companies design only 1% of global chips.

Maisto suggested this once again shows the scale of the challenge faced both by individual European nations and the region as a whole under EU leadership.

Looking ahead, he said the focus for many will be to “manage the dependencies rather than avoiding them”.

FOLLOW US ON SOCIAL MEDIA

Channel partners are increasingly adapting infrastructure strategies to help customers navigate hardware pricing volatility and component shortages, according to new research from Westcon-Comstor.

The distributor found that 87% of channel partners have already taken action to reduce the impact of supply constraints on customer projects, with project delays identified as the single biggest risk created by ongoing pricing pressure and hardware availability issues.

The findings are based on a survey of more than 100 members of Westcon-Comstor’s global Tech Xpert community, comprising technical specialists and decision-makers from resellers, systems integrators, and managed service providers across EMEA and APAC.

According to the research, partners are responding to these challenges by improving infrastructure planning and adopting managed services, while some are even switching technology vendors.

“Hardware pricing volatility and supply constraints are reshaping customer buying decisions and having a profound impact on the IT channel,” commented Callum McGregor, chief operating officer and chief financial officer at Westcon-Comstor.

“In response, partners are helping customers plan earlier, adapt architectures and stay flexible so they can keep projects moving despite uncertainty.”

Adapting mitigation strategies

In terms of mitigation strategies, improved infrastructure capacity planning was found to be the most common, cited by 60% of respondents.

Meanwhile, almost a third (29%) said they had already switched vendors or platforms in response to hardware availability constraints.

Meanwhile, 27% of partners said they were using storage tiering or caching to reduce reliance on memory-intensive workloads, while a quarter (25%) reported moving workloads or infrastructure components to managed services.

Evolving buyer behavior

The research also highlighted a shift in buying behavior, with 75% of respondents reporting some form of change.

More than half (52%) said either they or their customers had accelerated hardware purchases to reduce the risk of future shortages or price increases – although almost a quarter (23%) reported delays or cancellations as organizations weighed up growing uncertainty.

The distributor revealed that hardware availability is also influencing infrastructure decision-making, with 45% of partners stating that supply chain constraints had accelerated cloud adoption among customers and 7% reporting a shift back to on-premises infrastructure.

A balancing act for partners

According to McGregor, the ability to balance immediate performance with long-term requirements is becoming “increasingly critical” in order for partners to maintain success in the current climate.

“It also demonstrates the strategic role that channel partners, supported by distribution, can play in building resilience and helping customers navigate the pressures that have quickly become a systemic feature of the market,” he added.

Elsewhere, Westcon-Comstor’s research found that 39% of partners identified technical challenges – including performance instability, operational complexity, and reduced resilience – as key risks linked to hardware shortages.

Around one-quarter (24%) also said supply constraints were limiting their ability to scale infrastructure capacity.

FOLLOW US ON SOCIAL MEDIA

Pure Data Centres Group has announced plans to invest more than €7.5 Billion to build a sprawling data center campus in Finland.

The deal represents one of the largest ever inward investment projects in Finnish history, and will see an initial pledge of €1.5 billion to develop a 110MW AI campus in Seinäjoki.

Long-term, Pure DC said the site will eventually scale to a 500MW campus capable of hosting high-performance computing (HPC) and AI workloads.

Phase one of the project is already fully leased, the company revealed, with the substation for the first data hall constructed and live.

Gary Wojtaszek, Executive Chairman & Interim CEO of Pure DC, said the project marks the beginning of one of Europe’s largest-ever AI infrastructure projects and will help put Finland at the forefront of the region’s AI transition.

“Countries building AI infrastructure over the next decade will shape the global economy for the next fifty years. Finland has all the ingredients required to lead this transformation - world class engineering talent, abundant renewable energy and a culture of innovation that gave rise to Nokia”, Wojtaszek said.

“We are not building data centers in Seinäjoki. Together with our partners, including the Seinäjoki Municipality, we are helping create one of Europe’s most important AI ecosystems capable of supporting global tech leaders as well as the next generation of Finnish entrepreneurs and innovators.”

Pure DC will work closely with local partner, SDC Ventures, as part of the development, which is expected to provide over 3,000 jobs during the decade-long construction projects.

Inside the Seinäjoki data center campus

Designated SJK01, the Seinäjoki campus is a sprawling 370 acre estate that, upon completion, will support 550MW of IT capacity. Pure DC said the data center site is designed specifically for “mass-scale” AI and machine learning workloads.

The site also has access to 700MVA of renewable power, meaning it will operate with “exceptionally low” energy costs and a stable grid connection.

Notably, the development will use repeatable 40MW ‘AI-ready’ modules, which come equipped with built-in Direct Liquid Cooling (DLC) capabilities to support high-density workloads.

The data halls located at the site are also based on a modular design aimed at streamlining efficiency and predictability. Jaakko Kiiskilä, Mayor of Seinäjoki Municipality, said waste heat from the data center will be repurposed to support the district’s heating network.

“It’s fantastic to see a British-headquartered firm thrive on the world stage by delivering the UK's largest inward investment into Finland, cementing our position as a world leader in the AI space,” said Peter Kyle, secretary of state for the UK’s Department of Business & Trade.

“Through our Modern Industrial Strategy, we're ensuring the UK remains a leader in AI, setting ourselves apart globally while making sure our trillion-dollar tech sector continues to thrive for decades to come.”

FOLLOW US ON SOCIAL MEDIA

Intel has announced a €5 billion investment to expand its Leixlip campus in Ireland, with plans to scale capacity to deliver Intel Xeon 6 and next-gen Intel Xeon chips built on its Intel 3 node.

Using existing capacity across cleanroom space, the company said the investment will see it expand current production output and advance research and development activities.

As part of the move, the chipmaker will upgrade existing fabrication facilities and install new manufacturing equipment.

Key infrastructure enhancements, including an expansion of the automated track system, aim to integrate disparate campus modules into a single, high-velocity production environment.

“This €5 billion investment represents a definitive commitment to maximize capacity at our Leixlip campus and increase what we can deliver to Intel Foundry customers," said Naga Chandrasekaran, executive vice president, chief technology and operations officer and general manager of Intel Foundry.

“By investing in our existing fabs with state-of-the-art technology and installing cutting-edge tools, we are not just increasing output of critical products like Xeon 6 and next gen Intel Xeon processors built on Intel 3, we are ensuring that Ireland remains at the forefront of the world’s most advanced manufacturing ecosystems, while strengthening the region’s role in the global technology landscape.”

Work has already started, according to Intel, and the company expects it to deliver jobs for specialized tradespeople, including construction workers and equipment installers.

Full-time tech-focused jobs at the chipmaker will also be delivered as part of the expansion scheme.

“Intel’s latest multi-billion-euro investment in Leixlip is a powerful vote of confidence in Ireland, our skills base and our position at the heart of Europe’s most advanced manufacturing ecosystem," said An Taoiseach Micheál Martin.

"At a time of rapid technological change and global competition, this expansion strengthens Ireland’s role in securing resilient semiconductor supply chains and reinforces our ambition to remain a global leader in innovation, productivity and sustainable economic growth.”

Intel builds on deep Ireland ties

Intel has been active in Ireland since 1989, and has invested more than €30 billion in the Leixlip campus, which opened in 2023, in that time.

In 2024, the company sold 49% of the plant to US private equity firm Apollo, but bought it back earlier this year at a higher price. Last August, the US government took a 10% share in Intel at a point when it was valued at less than a fifth of its current market capitalization.

With the EU’s Chips Act aiming to reduce reliance on foreign supplies of chips, Intel said the investment will strengthen Europe’s semiconductor supply chain and support the EU's tech sovereignty ambitions.

“Intel is one of Ireland’s longest standing and most strategically important investors," said Michael Lohan, CEO of Ireland's Industrial Development Authority (IDA).

"This project demonstrates the value of Ireland’s skilled workforce, innovation ecosystem and stable business environment, while reinforcing Ireland's leadership in advanced semiconductor manufacturing, supporting both European competitiveness and resilient global supply chains.”

FOLLOW US ON SOCIAL MEDIA

Irish data centers consumed a quarter of the country's electricity last year, almost as much as every home in the country combined.

New figures from the Central Statistics Office (CSO) show that data center electricity consumption rose to 7,663 GWh in 2025 from 6,973 GWh in 2024 – a 10% rise – to hit 23% of the national total.

For comparison, 2025 consumption stood at 18% for urban households and 9% for rural households. Consumption by all users other than data centers, including residential and other business customers, went up by just 2% over the same period.

The figures represent the latest phase in what's been a continuous rise. The CSO's quarterly tracking from 2015 to 2025 shows that data center electricity consumption grew from 291 GWh in the first quarter of 2015 to 1,991 GWh in the fourth quarter of last year.

Total metered electricity consumption rose 34% during the period, the officials noted.

"Newly compiled quarterly figures spanning 2015 to 2025 highlight a substantial increase in metered electricity consumption by data centers," said Dr Grzegorz Głaczyński, statistician in the Climate and Energy Division.

"Data center consumption has grown every single year without exception, more than doubling between 2015 and 2019 from 1,240 GWh to 2,490 GWh, and tripling again between 2019 and 2025, reaching 7,663 GWh."

Ireland is a data center haven

Despite its size, Ireland currently has more than 80 data centers, mostly in the Greater Dublin Area.

In 2021, the country introduced an effective moratorium on the building of new data center connections to the electricity grid, which remained in force until late last year.

Developers were required to either generate their own electricity on site, or relocate to regions outside the Greater Dublin Area.

Even during the moratorium, data center electricity consumption continued to rise steadily, with the International Energy Agency (IEA) predicting two years ago that it would represent a third of the country's electricity consumption by this year.

Now, under a new Large Energy Users (LEU) Connection Policy finalized late last year, the Commission for the Regulation of Utilities (CRU) requires new data centers to meet at least 80% of their energy demand from renewable sources.

This requires large data center developers to provide 100% on-site, flexible power generation, with at least 80% of annual electricity to be sourced from new, unsubsidized renewable projects within their first six years.

At the same time, a new Large Energy User Action Plan (LEAP) was introduced to encourage investment in energy intensive sectors such as semiconductors, pharmaceuticals, precision engineering, and hyperscale data centers.

The plan calls for data centers to be sited in green energy parks and co-located with renewable energy sources, including offshore wind.

It also allows developments outside these locations, however, in some cases until beyond 2030, in line with existing planning arrangements. One such area is Dublin, which has raised concerns among opposition parties.

These plans were subject to a fierce cross-party pushback last year, as ITPro reported at the time, largely due to concerns about the impact on local communities and the environment.

FOLLOW US ON SOCIAL MEDIA

IBM has cut the ribbon on a series of new mainframe offerings as the tech giant eyes data center efficiency gains.

The new IBM z17 and IBM LinuxONE 5 series mark the first time the company is offering rackmount options alongside single frame systems.

According to Tom McPherson, general manager for IBM Z and LinuxONE, the move here aims to help customers consolidate data center footprints without sacrificing performance.

“The number of mission-critical workloads is rising at an incredible pace, forcing organizations to make tough decisions about performance, AI integration, and infrastructure footprint,” McPherson commented.

“With these new IBM Z and IBM LinuxONE systems, we’re making it easier to run workloads where they make the most sense, while opening the door for a wider range of organizations to benefit from these technologies for the first time.”

IBM eyes data center flexibility

The new IBM z17 and LinuxONE 5 configurations support up to 82 cores and 18TB of memory across two processor drawers, according to IBM. The company said this represents a 20% increase in core count alongside a 12% increase in memory capacity.

The IBM z17 single frame is a fully packaged solution in an IBM rack designed to operate as a “complete enclosed unit ready to deploy now”.

With the IBM z17 rackmount, customers will have the option to install IBM Z components within their own industry-standard racks.

The company noted that customers can also co-locate IBM equipment with third-party components to “achieve the best fit-for-purpose installation” based on their needs.

“Each system is designed to help organisations reclaim space, improve energy efficiency, and integrate seamlessly into existing environments,” the company said in a blog post.

LinuxONE upgrades target AI performance

Elsewhere, the new IBM LinuxONE series comes in two dedicated configurations: the Rockhopper 5 and LinuxONE 5 Express options.

The Rockhopper 5 series is a “scalable, multi-drawer” configuration designed specifically for high-density workloads, according to IBM.

This comes equipped with on-chip AI acceleration features, as well as confidential computing and post-quantum cryptography capabilities in both single frame and rackmount configurations.

Meanwhile, the LinuxONE 5 Express option offers the same capabilities, albeit in a compact configuration. IBM said these are designed for smaller workloads, with a particular focus placed on cost-efficiency and scalability.

The z17 single frame and rackmount configurations, IBM LinuxONE Rockhopper 5, and IBM LinuxONE 5 Express series will be generally available for customers from 12 August.

FOLLOW US ON SOCIAL MEDIA

The US National Institute of Standards and Technology (NIST) has announced plans for a $20 million center to drive quantum research, development, and manufacturing.

The Quantum Manufacturing Engineering Center (QMEC) is being set up with the help of non-profit R&D organization SRI International, with the aim of strengthening the manufacturing of scalable, high-performance quantum components and systems.

“NIST is a world leader in quantum science and technology based on decades of fundamental research that helped launch the US quantum industry,” said under secretary of commerce for standards and technology and NIST director Arvind Raman.

“This public-private partnership with SRI International will accelerate the development of America’s quantum industrial base — the foundation upon which the quantum revolution is being built.”

The QMEC will recruit companies and quantum experts to conduct research and development aimed at boosting US manufacturing and strengthening supply chains to enable quantum technology production at scale.

Initially, the center will focus on areas such as cryostats and lasers, with the specific goal of improving the manufacturability of quantum-enabling components within three years.

This will include developing scalable processes for quantum chips and integrated photonic circuits and establishing standards and quality control methods for quantum technologies.

The project also aims to build a US-based supply chain for materials and components required for quantum systems, build expertise through programs focused on quantum, and accelerate commercialization pathways from laboratory to production.

“Our approach is results-driven and timeline-focused,” said Lawrence Lee, SRI’s QMEC program director. “Each project we undertake will have clear goals, defined milestones, and measurable outcomes that are informed by collaborative discussions with quantum technology companies, value chain partners, and end users."

US eyes competitive edge in quantum

NIST and SRI have been working together since 2019, establishing the Quantum Economic Development Consortium (QED-C) in response to the National Quantum Initiative Act – work that led NIST to identify quantum manufacturing as a critical gap in national efforts to develop a robust commercial industry.

Earlier this month, the white House issued an executive order, Ushering in the Next Frontier of Quantum Innovation, aimed at accelerating the deployment and commercialization of quantum computing, sensing, and networking.

Among other things, the EO directed the Secretary of Commerce to develop a plan for advancing the commercial readiness of quantum sensing, quantum-sensor manufacturing technology, and quantum-network-enhanced timing.

This will include prioritizing research, development, testing, and evaluation of applications and hardware for quantum sensing and quantum networking.

The new manufacturing engineering center will be a key component in the government’s push on this front, according to NIST.

"Quantum science promises to generate new knowledge and technologies that will supercharge scientific research and unlock enormous economic potential,” said deputy secretary of commerce Paul Dabbar.

“The new Quantum Manufacturing Engineering Center will bring together top experts to ensure both continued US leadership in quantum technologies and that we are the epicenter of manufacturing quantum systems at scale to drive advances in sensing, communications, encryption, computing, biomedicine and other critical areas.”

FOLLOW US ON SOCIAL MEDIA

Three-quarters of utility firms are struggling to forecast energy demands due to rapid data center infrastructure expansion.

A new study from Capgemini Research Institute suggests that eight-in-ten utilities predict more extreme and volatile demand patterns. A key factor here is that consumption patterns from AI are less stable and more difficult to model.

The biggest problem is uncertainty, the study noted, with utilities increasingly forced to plan for demand that may never materialize.

Two-thirds of electricity executives refer to “phantom” data center load requests - with 19% of these never materializing.

This forecasting uncertainty creates real problems with capital allocation, Capgemini warned, with utilities having to decide not only how much capacity to invest in, but where and when to prioritize grid modernization investments to support future demand.

Hyperscalers are feeling the pinch

Things are just as bad for hyperscalers, which need to make major infrastructure decisions against a backdrop of uncertain demand forecasts, grid availability, and connection timelines.

The problem is made all the harder by the geographic concentration of data centers, which places significant strain on local grids.

More than half of electricity executives are finding that load concentration is a major obstacle to reliable service, while large clusters of high-density facilities are creating localized bottlenecks that affect system stability and investment planning.

“AI is transforming electricity systems far beyond demand growth. It is exposing structural constraints in grid capacity, planning, and power availability, while making demand more dynamic and harder to predict,” said Claire Gauthier, global head of energy & utilities at Capgemini.

“The challenge is no longer only how much power is needed, but whether it can be delivered reliably, where and when it is required. Utilities have a defining role to play as system orchestrators, leveraging AI-enabled insights to balance grid and customer-owned resources, accelerate deliverable capacity, and enable the next phase of data-center growth.”

What’s driving energy consumption?

The projected increase in electricity consumption is largely down to AI training and inferencing, according to Capgemini.

Training and inference processes currently account for around 25% of data center energy consumption, yet this is projected to reach 60% within the next three to five years.

The consultancy noted that this not only places significant strain on data center infrastructure, but also leaves other IT workloads on the sidelines as enterprises prioritize AI-related activities.

Data centers are increasingly shifting from backup-only approaches toward primary behind-the-meter (BTM) and near-site solutions. Nearly three-in-ten say they already deploy on-site power solutions and 39% plan to add on-site/BTM within the next couple of years.

Operating independently

Looking ahead, the majority (86%) of respondents see the ability to operate independently from the grid as a competitive advantage.

Around three-quarters of utilities and data center executives told researchers they were looking to try and establish a diversified energy mix to ensure reliability and long-term resilience.

Part of this lies in the fact that renewable energy alone cannot provide enough continuous power at scale for large data centers and AI workloads.

Battery energy storage systems (BESS) are emerging as a possible solution, with nuclear Small Modular Reactors (SMRs) expected to take time to deploy.

More than two-thirds of electricity and data center executives globally are looking to natural gas as a near‑term, transitional solution until renewable energy and storage technologies can scale.

"For both energy providers and data-center operators, the key challenge is no longer only scaling capacity, but doing so under uncertainty, speed constraints, and rising system complexity,” said Gauthier.

“Success will depend on the ability to align infrastructure investment, energy sourcing, and AI-enabled operations to manage both the scale and volatility of demand, while balancing reliability, cost, and sustainability.”

FOLLOW US ON SOCIAL MEDIA

While every CTO’s job varies, some stand out more than others, and being CTO of one of the USA’s national laboratories is one such case.

In eastern Tennessee, south of the city of Oak Ridge and close to the Clinch River, sits the aptly named Oak Ridge National Laboratory – or ORNL for brevity – and its CTO, Scott Atchley.

ORNL is one of two facilities – the other being Argonne National Laboratory in Illinois – that make up the Leadership Computing Facilities. These two locations, both of which are funded by the US Department of Energy (DoE), are home to some of the fastest and most powerful supercomputers not just in the US but in the world: Aurora at Argonne and, at ORNL, Frontier.

Atchley, a 16-year veteran of ORNL, tells ITPro that there have only been two weeks in his tenure as CTO that he’s regretted taking the job.

“That’s when we had deployed Frontier,” he says.

Full power failure?

Yes, even in one of the most advanced computing facilities in the world, it’s still entirely possible to have a relatively small fault cause chaos.

“[Frontier is] our current supercomputer, it's an HPE Cray EX system and one of the benchmarks – not the goal, but one of the benchmarks – for the system is that it would reach exascale, or 10 to the 18 flops,” says Atchley.

Supercomputers’ power is only announced twice a year – at ISC in Germany, which takes place in June, and SC Conference in the USA, which takes place in November. It’s from the announcements at these conferences that the Top500 list of most powerful supercomputers is drawn.

With the deadline for ISC 2022 fast approaching, Frontier was failing to perform.

“We were only about halfway [to 10¹⁸ flows], and we could not figure out why,” says Atchley. “We were having daily calls with HPE and AMD, and the deadline was coming.

“We were in a full panic, and this was the first time I thought, ‘my gosh, we might have bought the wrong machine’.”

Fortunately for Atchley and everyone else involved, he was wrong, and once the problem was identified, it was a “simple fix”.

“It was a one-line software change, and all of a sudden the performance shot up to right at just under an exaflop,” he says. “When they're new, [supercomputers] are very fragile. So, we rebuilt the machine during the day, and the next night we did the runs, and we finished it at about six in the morning.

“There were probably 50 people in line watching … and when the job looked like it was going to finish, all the chatter went quiet, nobody wanted to jinx it – we had been trying all night long, and it was really exciting when we saw the result that we actually exceeded an exaflop.”

Putting power to use

Top500 does have its limitations – only companies that run the LINPACK benchmark and submit the outcomes are measured, for example. There are also claims that it’s not particularly scientific, and Jack Dongarra, one of the creators of LINPACK, has himself admitted that its usefulness is limited.

“This benchmark reports the performance of this one application with its floating point operations and message passing. So there is one application and one number that represents the performance for the computer system. The weakness is that this is just one application and one number,” he told HPCWire in 2002.

Limitations and weaknesses don’t mean useless, however, and Atchley pushes back at the idea that Top500 is entirely meaningless.

“A lot of people will knock the Top500 … but we've actually had jobs on the system doing real science exceed an exaflop in performance, which is really cool, and so we're glad,” Atchley tells ITPro. “Our users love the machine, it's doing some incredible science.”

From aerospace to outer space – and back

Speaking of users, who is making use of Frontier? One case study Atchley gives is from the engineering firm GE. In partnership with French firm Safran, they have been developing a new type of jet engine.

“If you look at engines when commercial airliners came out in the 60s and early 70s, the engines were really narrow, really long,” says Atchley. “If you flew[today] and looked out the window, the engines now are really short, but really wide.

“Making them wider makes them more energy fuel efficient, but they're to the point now that if they make them any bigger, the gains they get in the efficiency they lose because of the drag on the nacelle.”

One option is to get rid of the nacelle – the housing around the engine – altogether. This isn’t a new idea, Atchley explains: “NASA has known about this for 50 years.”

The only issue is that using them on a commercial aircraft would deafen the passengers and crew, because they are – or would be – so loud.

It was only once Frontier came online with its exascale capabilities that GE and Safran were able to measure in high resolution the swirls of turbulence coming from the engine.

“Now they're tweaking the design to minimize those, and so that is the huge breakthrough, is understanding turbulence at a really fine scale. They can't do that in a wind tunnel. This engine is so big, there's no wind tunnel on earth that would fit it, so the only way they can do it is in a computer,” says Atchley.

As powerful as Frontier is, however, it can only model a 14-blade engine. In 2028, though, Frontier is expected to be superseded by the Discovery supercomputer, which may open up new opportunities for the likes of GE.

Once it’s online, ORNL will be home to three supercomputers: Frontier, Lux (which is expected to come online in 2026), and Discovery. At least for a time. Frontier will, ultimately, be decommissioned just as its predecessor, Summit, was.

“The challenge you get on these machines is even if the hardware doesn't run out or doesn't fail, the software, the vendors don't want to support something that's now 6,7,8 years old,” says Atchley.

“You think you want to have the latest software, you want to have the latest security patches, and so eventually you have to turn the machine off, but we had universities [and] industrial partners asking us, ‘can you run Summit longer?’ – we'd love to, but we can't.”

Sometimes there is a second life for these apparently beloved behemoths that isn’t just the recycling route.

“IBM was able to sell a portion of it to an oil and gas company, and so it is still doing work for them somewhere else.”

When the CrowdStrike update and associated outage adversely affected Windows devices in July 2024, many organizations were impacted. Just outside Houston, IT operations at the Spring Bank Independent School District (ISD) were among those affected.

Spring Branch ISD has over 31,000 children under its care across 46 schools, spanning 57 buildings and 6,000 staff members.

The organization faced significant challenges when the CrowdStrike outage hit, and it came at the worst possible time, given that summer school classes don’t run on Fridays.

“During the summer, school districts are primarily off on Friday, we work 4-10, so we’re closed on Fridays," Troy Neal, Spring Branch ISD’s executive director of cybersecurity and technology, told ITPro at Pure Accelerate 2026.

“[We] woke up Friday morning, turned on the news, and there’s this banner running across that says ‘CrowdStrike outage’ and I’m like ‘we’re a CrowdStrike customer, is this really affecting us?’

“Outages can be lots of things, so I looked at my phone, had some messages, and [I said] let’s start checking the systems, and [I] notice[d] the phone system was down.”

This posed a problem for the organization, given that these systems are hosted on-prem, Neal explains. Further investigation showed staff members couldn’t access some critical services.

At this point, the scramble to recover, or at the very least establish what was happening, was on.

Disaster recovery with Everpure

Spring Branch ISD works with a range of hardware vendors for storage, recovery, and backup purposes, Neal explains. Among these are Hitachi, Dell, Veeam, and Everpure, formerly Pure Storage.

High-capacity, high-performance storage is critical for the organization, which manages more than 60,000 endpoints spanning Chromebooks, iPads, and various Windows devices at schools across the district.

When Neal started at Spring Branch ISD some seven years ago, the organization was in the process of refreshing its storage capabilities due to the sheer volume of CCTV footage they stored.

“We have 4,400 cameras in our school district,” he explains. “We store about four to six petabytes of data on just CCTV, so we started doing our due diligence.”

This exploratory process saw Neal and colleagues scour the market for options, eventually narrowing the field down to Dell, Everpure, and Hitachi. Budgetary constraints were a concern, he notes, particularly given that Spring Branch ISD is publicly funded.

An initial offer saw Neal faced with a $7.1 million price tag for six petabytes across two locations over 10 years. He felt this was actually good value, but it was outside of his budget, which was just over half that amount.

Hitachi ultimately met Spring Branch’s requests, and the organization secured vital storage infrastructure to meet its increasingly demanding data needs. But it was here that the relationship with Everpure blossomed.

“We were really impressed with Pure at the time,” he says. “In the cyber space, 3-2-1 is the standard for backup strategies. I like 4-2-1, [as] you can never have enough backups.”

“I’m a layered defense kind of guy for everything we do, and so we had an opportunity with some extra bond funds to bring in another solution to be our secondary backup.”

That deal saw Spring Branch ISD adopt FlashBlade//S and FlashBlade//E solutions, which provide users with high-speed, scale-out backup storage. Notably, Neal says that when the CrowdStrike outage hit, these proved vital.

The road to recovery

After the initial furore of the CrowdStrike outage and the impact on Spring Branch ISD dissipated, Neal says he was in some way glad that it was a “CrowdStrike outage, not a Spring Branch ISD outage, and caused by an adversary.”

Focusing on the task at hand, Spring Branch found itself in a similar situation to many organizations elsewhere in the world: dealing with a botched patch.

“We’re in the office by 6.30 am that morning,” he says. “We pulled out our incident response plan, and [said] this is a live exercise.”

“Our first thing was their patch,” Neal adds. “That didn’t work too well.”

The initial workaround saw Neal’s team revert to a Veeam backup, which he says was “just a regular-sized virtual machine (VM) that we had”. This was expected to take around 45 minutes for a single VM.

Spring Branch ISD has over 200 VMs across its IT estate, meaning the recovery time could take days. With summer school students and staff expected back in on Monday morning, this could prove disastrous.

“We [said] this isn't going to work, let’s try Pure too.”

The resulting backup “took minutes,” Neal notes, adding “this is the performance we paid for and why we chose Pure.”

Leaning on Everpure backups, Neal says critical services were back up and running within around four hours. Thereafter, ancillary services were up by around noon.

“About six hours later, we were up and running, and then we did some cleanup work over the weekend. We also had to wait on some of our owners of some of these systems to test them, validate them, and then by the end of the weekend, we rank a whole other snapshot of every single VM,” he says.

“We had some good clean states going forward, and Monday came around, and summer school went on. Nobody knew the difference.”

Looking ahead, Neal says Spring Branch ISD plans to continue - and expand - on its relationship with Everpure. The organization is going back to a 3-2-1-style cluster, and recently brought in another FlashArray//C solution for additional backup.

Notably, the school district is also moving to EverGreen//One, the company’s storage as a service (STaaS) platform. This aims to unify on-prem and cloud storage resources, providing users with a single data storage platform based on a subscription. Neal notes that the flexibility and ease of use associated with this setup were particularly appealing.

“I’m going with a subscription-based model. So I don’t care what the hardware is, because that’s Pure’s problem to figure out and manage,” he jests.

Colt DCS has cut its greenhouse gas emissions across Scope 1, 2, and 3 categories by 27% since 2019 as part of its efforts to decarbonize the design, development, and operation of its data centers.

The firm’s 2025 Sustainability Report reveals it achieved 100% renewable electricity for Scope 2 emissions and a 90% renewable share of its total energy consumption across Scopes 1, 2 and 3.

"We are proud of the progress made in reducing absolute emissions in 2025 as we continue advancing towards our 2045 net zero target. Whole life carbon assessments are helping customers and partners identify key sources of value-chain emissions and understand the impact of design and procurement decisions," said Anthea van Scherpenzeel, the company's head of environmental sustainability.

"Through close collaboration with suppliers and industry partners, we are confident we can continue delivering projects at speed and scale while making steady progress towards our SBTi targets.”

Last year, the company expanded its Global Reference Design (GRD), developing a standardized blueprint it said is flexible enough to adapt to local requirements while keeping consistent core principles.

The GRD sees energy-efficient systems, lower embodied carbon materials and near-zero wastewater cooling technologies incorporated into all new data center designs.

According to Colt, the aim here is to maintain annualized Power Usage Effectiveness (PUE) below 1.2, along with waste diversion rates of 75% in construction and demolition and 90% in operations.

So far, the company has reached TRUE Zero Waste certification across three sites - London North, Tokyo Inzai 1-3, and Osaka Keihanna - and hit a global PUE of 1.41.

In terms of water use, the GRD incorporates chilled water systems using low-Global Warming Potential (GWP) refrigerants with near-zero water consumption, alongside rainwater harvesting for indoor use.

Meanwhile, closed-loop liquid cooling systems recirculate coolant, transferring heat without relying on evaporation.

Dina Nassar, sustainability analyst at Colt DCS, said efforts to improve sustainability capabilities come in direct response to AI-related pressure, which is creating “new efficiency challenges” for data center operators globally.

"Our revised GRD underpins the efficiency gains made in 2025, by ensuring lower embodied carbon materials and advanced cooling technologies are baked in at the design stage.”

Data center sustainability in the spotlight

The figures published by Colt come amidst rising concerns over data center efficiency and the environmental impact of infrastructure worldwide.

A recent UN report concluded that AI will be using the same amount of water as 1.3 billion people by 2030, along with 945 terawatt-hours of electricity, while CO2 emissions will rise to 399 million ​tons.

Similar research from PwC warned the data center boom is increasing electricity consumption faster than the supply of clean energy in some parts of the world, leading to higher power prices and increased reliance on fossil fuel generation.

In the US, the International Energy Agency (IEA) predicts that most of the increase in data center electricity demand through 2030 will be met by natural gas, holding back decarbonization efforts.

Earlier this month, the EU announced plans to develop minimum energy-efficiency standards for data centers, saying that there's an urgent need for them to become more energy efficient, reuse waste energy such as heat, and use more renewable energy sources, with a view to becoming carbon-neutral by 2030.

FOLLOW US ON SOCIAL MEDIA

Construction has started on the UK’s next national supercomputer, claimed to be 50 times more powerful than the country’s current model, ARCHER2.

Owned by UK Research and Innovation (UKRI) and hosted at the University of Edinburgh, the system is being built with the help of a £750 million investment from the UK government.

According to the university, the supercomputer will empower "world-changing" discoveries in fields such as aircraft engineering, extreme weather events, and drug discovery for cancer.

"This supercomputer has the potential to help UK researchers accelerate frontiers, unlock new paradigms, and develop solutions to challenges that were previously impossible," said Professor Liz Baggs, vice-principal for research and innovation and chair of food and environmental security at the University of Edinburgh.

"This includes creating the next generation of medicines, revealing untold stories about our planet, and, most excitingly, leading to discoveries that we can’t even imagine yet.”

The new machine is expected to be able to deliver at least a billion billion calculations per second, a marked increase compared to the 20 million billion delivered by ARCHER2.

This means the system will be able to carry out incredibly complex calculations in hours rather than days, and solve much larger problems than ever before.

Notably, the supercomputer will be more efficient than existing models, with surplus heat generated being used to warm university buildings. Developers also plan to assess whether it could also be used to warm local homes by warming water in disused mines.

To reduce the environmental impact, site demolition has been kept to a minimum, and there's a range of sustainability projects planned, including tree planting, protecting ancient trees, and conservation projects for local wildlife.

Construction work on the building, which is roughly the size of a supermarket, is being carried out by Robertson Construction Central East.

“You would never guess from this ordinary-looking building site just how vitally important it will be for the UK and how its contents could impact on all of our lives positively," said Professor Mark Parsons, director of the Edinburgh Parallel Computing Centre (EPCC) at the University of Edinburgh.

Building on ARCHER2’s success

ARCHER2 – also housed at the University of Edinburgh – played a big part in Covid-19 drug discovery, helped Rolls Royce to improve aircraft engine efficiency and sustainability, and enabled engineers to make wind farms more efficient.

The ageing supercomputer has generated £8 per £1 invested, enabling more than £4.2 billion in benefits for the UK economy, according to an independent report by London Economics.

The new supercomputer nearly didn’t happen, having been shelved by Labour when it came into power, and then reinstated a year later.

“The commencement of construction in Edinburgh marks a pivotal moment for the UK’s scientific infrastructure. We are providing the UK’s brightest minds with the 'industrial-scale' tools required to solve society’s most complex challenges, from decoding the next generation of life-saving medicines to engineering a net-zero future," said AI minister Kanishka Narayan.

"This £750 million investment isn't just about speed; it’s about economic growth, and ensuring the UK remains the premier destination for global innovation."

FOLLOW US ON SOCIAL MEDIA

IT, in its current form, relies entirely on networks. From office LANs to transatlantic fibre-optic cables, to WiFi and cellular connections, no business can really carry out its day-to-day tasks without using at least one network.

Yet, with all the attention focused on AI and its vociferous appetite for GPUs, CPUs, and even storage, the network risks being overlooked.

Without both investment and new ways of engineering the network, connectivity could derail progress in digital transformation, the user and customer experience, data science, and, of course, AI.

Recent research from vendor Cisco found that three in four enterprises expect to hit the limits of their campus and branch networking capacity within two years, and they are seeing network traffic double year on year. Already, some enterprises say they are holding back on scaling up AI investments because of network constraints.

Unlocking the problem

“You have to look at where bottlenecks are,” Paul Stringfellow, CTO at IT consultants Gardner Systems, told ITPro.

“This is less an enterprise network issue, but more a backend issue. So, looking at how you ensure high capacity [connections] between CPUs, GPUs, and back-end storage.”

As Stringfellow points out, there have been advances in high-speed connectivity, including faster Ethernet speeds and RoCE (RDMA over Converged Ethernet), which allows remote direct memory access over Ethernet technology.

These, though, along with Infiniband and 400Gbps and even 800Gbps Ethernet, are mostly targeting data centres and cloud providers.

For client devices, developments in WiFi should improve both connectivity for users and for IoT devices. Technologies such as WiFi 6 and WiFi 7 – and in the future, WiFi 8 - look to increase bandwidth and provide reliable connections, even as more users or devices connect to the WLAN.

“WiFi also needs to be included in this conversation, so many users now use WiFi as the primary office connection. The idea of cabled connections is foreign in a lot of places,” says Stringfellow.

The move to cloud and SaaS-based applications, as well as widespread home working, means that much of an enterprise’s data traffic has shifted to WiFi, with users never reaching for an Ethernet cable. Effective WiFi has become essential, with poor connections directly affecting productivity.

“Enterprise Wi-Fi has evolved significantly and now carries far more than simple user connectivity,” Phil Huang, business development and field application manager for the UK and Ireland at vendor D-Link, told ITPro.

“Modern Wi-Fi networks routinely support collaboration platforms, voice and video applications, IoT devices, and a wide range of cloud-based services,” he says. “In that sense, Wi-Fi has become a strategic access layer rather than just a convenience technology.”

But, he says, there are still significant differences between wired and WiFi networks. And improvements in WiFi connection speeds are pushing bottlenecks further into enterprises’ core networks. This, in turn, is driving investments in the wired LAN too.

“Newer Wi-Fi 6E and Wi-Fi 7 access points are capable of exceeding 1Gbps throughput, meaning a standard Gigabit uplink can quickly become the bottleneck,” says Huang.

As a result, businesses are moving to both 10Gbps connections and 2.5GbE. This is moving closer to “standard” 1Gbps connections in cost. “The premium over traditional Gigabit Ethernet has reduced dramatically, making 2.5GbE an easy investment for organizations refreshing their infrastructure,” he says.

Installing 2.5GbE, and even 10GbE, is simpler than fibre-optic or even higher-speed copper LANs. But it still works well for most office deployments, at distances under 100 metres.

Campus to cloud

The bigger challenges lie both in connectivity from businesses to data centers and the cloud, and within the data center itself.

Here, growing data volumes and the need to reduce latency are prompting CIOs to look at how their networks are designed, as well as the hardware they run on.

“As SaaS and AI adoption grows, hybrid work reshapes access patterns and digital security follows users beyond traditional perimeters,” Nauman Raja, a director analyst at Gartner, told ITPro.

“Enterprise networks are no longer built around the data centre. Instead, they now gravitate around the internet.”

AI too is playing its part: expensive GPUs are a waste of money if the network cannot keep up. Network architects need to meet the high-bandwidth demands of uploading data for model training, usually to the cloud, and the low-latency needs of inference.

Organizations also need to consider how they connect their existing enterprise applications to AI models. Conventional networks are rarely built with these demands in mind.

“I don't think we're having a bandwidth problem. We're mostly having a design or infrastructure problem,” Wim Remes, a technology consultant who has also run operations for a managed security service provider (MSSP), told ITPro.

“Compute and data are not always in the same place. Either we bring data to the compute or we bring the compute to the data. The paths we create on the network need to be the shortest.”

“The top-level issue, to me, is that we’ve ignored ‘the network’ for about a decade and a half,” he adds.

“Does a full-stack developer know the basics of networking? The clear answer is no. The stack in ‘full stack’ stops at the web server at best. At the same time, many a CIO also doesn't really care about the network anymore.”

The challenge is to bring together new and emerging hardware technologies, such as 400 or 800Gb Ethernet and RoCE, with a network architecture giving the performance data-heavy, latency-intolerant applications need.

“Bad network design, such as routing internet-bound traffic through centralised data centres for software as a service (SaaS) application, or for end-user collaboration, such as video conferencing, will result in poor user experience and inconsistent access for both remote and on-site users,” Gartner’s Raja warns.

“As demand fluctuates with remote working patterns, network architectures must allow dynamic bandwidth allocations,” Raja adds.

“In many cases, smart traffic management and automated policies can reserve sufficient capacity for business-critical applications even during peak usage.”

That, though, means thinking less about the network as hardware, but as an operational issue across the business. After all, no one wants colleagues, or worse still, customers, locked out of applications because someone “tokenmaxxing” is blocking the LAN.

Chinese supercomputer LineShine has secured the number one spot on the 2026 Top500 list, marking the second from the country to take first place since Sunway TaihuLight in 2017.

The rankings, which were announced today at ISC26 in Hamburg, mean LineShine ends El Capitan’s run as the number one supercomputer.

Housed at the National Supercomputing Centre in Shenzhen (NSCS), LineShine was built by the Shenzhen Cloud Computing Centre.

The cluster is composed of around 14 million Arm-based cores, and recorded a debut measurement of 2.198 Exaflop/s on the HPL benchmark.

That marks a 20% performance boost compared to El Capitan, the organization noted, which comes in at 1.809 exaflops.

El Capitan, Frontier, Aurora, and JUPITER Booster all remain high on the list, occupying the second through to fifth spots on the Top500. This means three of the top five are still housed in the United States.

Under the hood of LineShine

LineShine is based on the ‘LingKun’ platform, which boasts 304-core LX2 processors running at 1.55 GHz. The system also used the proprietary LingQi interconnect along with the Kylin operating system.

The developers behind the LX2 processor have not been disclosed by the Shenzhen Cloud Computing Centre, however, analysis from Jon Peddie Research earlier this year described it as ‘Huawei's LX2’ - suggesting involvement by the tech giant.

While LineShine secured first place on the HPL benchmark, it also bagged the number one spot based on the HPCG ranking, recording 22 Petaflop/s.

The HPCG (high-performance conjugate gradient) ranking system is an alternative supercomputer benchmark which measures real-world performance, including internal data transfer speeds and memory subsystems.

HPL, by contrast, measures peak computational speeds.

Elsewhere, on the HPL-MxP benchmark, LineShine debuted in fourth place at 7.92 Exaflop/s. Top500 attributed its “modest” performance on this front to its CPU-only design.

A CPU-based supercomputer?

As Tom’s Hardware reported earlier this year, LineShine’s CPU-only design is rare, with the system among only a handful of supercomputers that don’t rely on accelerator chips such as GPUs or APUs.

Fugaku, developed by Riken and Fujitsu, for example, features a CPU-only architecture.

A key factor here could lie in US-imposed chip bans on Chinese manufacturers, the publication noted.

“The list demonstrates that there is no single dominant technology path to leadership-class computing,” Top500 said in a statement. “Instead, vendors are pursuing a variety of CPU, GPU, APU, and custom-accelerator approaches coupled with different interconnect and system designs.”

FOLLOW US ON SOCIAL MEDIA

What is HPE? For just over a decade, since its split with what's now HP Inc, the answer has been clear: HPE is an enterprise hardware company specializing in servers and storage.

The additions of Cray and SGI to its portfolio bolstered this position: While they primarily opened the door to HPC and supercomputing for HPE, fundamentally there's a lot of crossover between typical data centers and these behemoths.

Even the name – Hewlett Packard Enterprise – invokes a sense of a company with a long history of making specialist kit for organizations' data centers or server rooms.

At HPE Discover 2026, though, you would be forgiven for thinking HPE's primary competitor wasn't Dell or IBM but Cisco.

Both of the first day keynotes were chock full of networking. I'm not just talking about announcements, of which there were some, I'm talking about the overall vibe as well. Both CEO Antonio Neri and head of networks, Rami Rahim, spoke about the importance of networking as a foundation for modern enterprise architecture, particularly for AI workloads.

Rahim offered a cautionary tale about the Millennium Tower in San Francisco, a residential high-rise that began sinking into the ground and listing to one side shortly after its completion due to problems with its foundations. His argument: if businesses don't lay solid foundations based on networking technology, then they're setting themselves up for failure. Their agentic AI ambitions, in particular, will become weak and start to wobble.

Such was the emphasis on networking all week, you would be forgiven for thinking that the company had moved away from its traditional business areas. In fact my first question to Antonio Neri when I spoke to him was: "Is HPE a networking company now, then?"

"Yes, it is," came the answer. He expounded on how this plays into HPE's expectation that it will grow into an $11 billion company, adding that "in my mind, together with our competitor, Cisco, [HPE has] the most complete networking portfolio".

"Our competitor Cisco" – what a phrase, and one that would have seemed slightly crazy just a few years ago.

It's not that HPE has had no networking business before this – Aruba Networks has been a core part of what HPE has to offer for years – but I don't think anyone would have openly declared Cisco a competitor in 2015.

Undeniably the Juniper acquisition does strengthen HPE's position in the data center – it can now offer the whole package, from scale up to scale out. Neri was keen to point out the AMD Helios, which should be coming this autumn, uses HPE Juniper Networks for scale up. Possibly more to come on that in July when we get to hear from AMD's CEO Lisa Su at AMD Advancing AI summit, though we shall see.

The question that I've been asking myself throughout the conference and I've voiced already on the 19 June edition of the ITPro Podcast and reaction video, is this: is the focus on networking a perpetual state of affairs, or is HPE just burnishing its latest acquisition?

My expectation is that for the next 6 months minimum Juniper's star will very much be in the ascendence at HPE. After that? Well, maybe the foundations will have settled and networking will have equal importance to the rest of what the company does. HPE Discover Barcelona, the company's European conference that takes place 2-3 December this year, may offer some more insight into the direction of travel.

Until then, you can catch up on all my coverage from Las Vegas here and if you want to know more about the tower and its ongoing issues – including attempts to remediate that made it worse – I recommend this blog, and also the Wikipedia page. It makes for quite a read.

While data center developers are looking forward to a prolonged period of expansion, they're still being held back by a shortage of available power.

In a mid-year update to its annual Data Center Power Report, Bloom Energy has found that developers are citing higher electricity prices, increased water consumption, and strain on grid reliability as the concerns most likely to influence projects.

Community scrutiny has only ramped up over the past six months, and, as of May, in the US, at least 18 state bills and 86 local moratoriums have been proposed. More than six-in-ten developers say they are planning to source their own power if the grid is unavailable, and nearly a third of these onsite-powered sites are expected to incorporate carbon capture by 2030, reflecting growing pressure to address emissions concerns while expanding power capacity.

"Access to power remains the biggest constraint to data center growth, but it is not the only issue. Community concerns are increasingly shaping which projects move forward," said Natalie Sunderland, chief marketing officer at Bloom Energy.

Developers are looking forward to adding significant capacity between now and the end of the decade, as AI adoption accelerates. Inference is now driving sustained demand for new data center capacity, and now accounts for more than 50% of AI compute, reflecting the transition from model building to real-world applications.

However, there's a gap in readiness that risks slowing AI innovation. Chip developers expect high-density architectures and rack-level DC designs to be adopted in 2028, a full year ahead of data center developers' plans.

"Our findings suggest that solutions that reduce strain on local infrastructure while helping developers bring new capacity online faster – such as clean onsite power – will play an important role," said Sunderland.

"We believe the winners in the next phase of AI buildout will be those that can grow in a way that works for both operators and the communities that host them."

The research follows repeated warnings that the power requirements of AI data centers are growing too fast for national grids to keep up. In a recent report, Wood Mackenzie concluded that with grid transmission build-outs five to ten years away, projects, markets, and consumers are all at risk. Colocated generation and flexible interconnection models aren't actually fixing the problem, said the firm, as these projects face huge technical, regulatory, and economic hurdles.

The problem's only set to get worse, according to real-estate services firm Savills, with only 850MW of power capacity delivered across EMEA since the start of this year, down by 11% on the same period last year.

Gartner has also warned that 40% of data centers could face constraints in power availability by 2027: consumption is expected to grow 160%, with the power required to run AI-optimized servers reaching 500TWh, more than two-and-a-half times as much as in 2023.

It's HPE Discover 2026 this week, but before the main conference kicks off the company welcomes its channel partners to the Partner Growth Summit.

As you can see from the main image, the theme for this year's Partner Growth Summit is "The Power of One". Follow along with this live blog for all the announcements from Simon Ewington's keynote, channel and partner ecosystem leader at HPE and keep an eye on our main HPE Discover page for all the announcements from the conference throughout the week.

Hello, and welcome to our live blog from HPE Partner Growth Summit 2026, I'm managing editor Jane McCallion.

This is the first Partner Growth Summit since HPE acquired Juniper Networks in July 2026. In December, at HPE Discover Barcelona 2025, the company talked about "cross-pollination" between Juniper and its existing networking business, Aruba, in terms of technology. Based on this week's strapline "the power of one", we may have more to hear about how that's working from a channel point of view.

Perhaps unsurprisingly, Antonio Neri will also be on stage at some point during this keynote, but we're kicking off with Simon Ewington who is talking up "the power of One". He starts with a thank you to the HPE channel partners, saying 64% of all business is done through the channel. This rises to 90% in networking.

Ewington is talking about the changes that have happened over the course of 2026. The "Power of One portfolio" is a three part thing, he says.

"First, we want to help others unlock greater relevance by delivering one portfolio built on innovation that keeps you ahead in the customer market. Second, to help partners [build] expertise, so that you can position that portfolio by providing the right tools, right training, and the right knowledge, and last to help partners as well higher knowledge when you sell and you win."

The announcements are coming thick and fast! First is some new channel-only offerings – we will have a write up of this available shortly – which include HPE Morpheus VM Essentials and a VM Essentials offering for partners themselves, which will be available 'free' for three years (there is a "nominal" support fee).

Ewington also talks up HPE's networking prowess and associated security.

"We're enabling you to meet strictest data residency and regulated requirements with air gaps and solvent ready private cloud solutions, which then can be combined with partner delivery, and then most obvious opportunity we create in one portfolio is through our worldwide HPE networking business," he says

"To take advantage of our merging networking portfolio, refine existing networking competencies into one, and adding a new rounding competency, which we've adopted to capture that opportunity," he adds. This means integrating the Juniper and HPE channel programs into a single one – which means moving everyone to Platinum, Gold, and Silver levels.

Antonio Neri, CEO of HPE, is now on stage with Ewington.

"First and foremost, thank you for all the support you provide to us as enjoying customers, and for the spectacular tools we deliver," says Neri.

He also talks about what has happened in the five months between the acquisition of Juniper in July 2025 and January 2026.

"Consider the large scale companies coming together, and we continue to deliver profitable growth for our shareholders ... so I'm very proud, because, as you said, we are building investment in business in the industry," he adds

There's a lot of hinting, nods and winks about what's going to be announced tomorrow. Big netowrking announcements (no surprise there) and the importance of AI. More on that tomorrow, where I'll be running a live blog on ITPro.

Back to Ewington, he's asking what HPE needs from partners in FY 2027. Neri starts off with a thank you to the partner community, adding that the role of HPE is to make partners "relevant", which grows business for everyone and increases innovation.

"Our role is to grow with you," he says, but it's not just abotu 2027. He's summing up by saying that the technological challenges and disruption that's happening now "we're all experiencing" but there's business opportunity there. And with a big round of applause he has left the stage

Now Phil Mottram is on stage for a "tight session" as it's sort of a "three in one".

"We're going to hear about the big trends we're seeing in the market, and where our customers are coming to us for answers"

Joining Mottram for the first segment are Maeve Culloty and Neil MacDonald.

MacDonald is addressing one of the biggest issues for partners (I've heard this not at HPE but elsewhere too) is supply challenges.

"We know that the supply environment is difficult and that it's added a lot of work, too much work for you and your teams, you're dealing with supply availability challenges, you're dealing with higher prices, pricing changes, and you're dealing with customers who all have serious budget challenges because of the dynamics in the industry, and what we don't expect this place situation to resolve any time soon," he says.

MacDonald reflects on changes the company has made recently to help partners, including faster delivery starts and smart choice SKUs.

BIG ANNOUNCEMENT: The company is moving away from a 14 day period to finalise a sale/quote validity. That's now 30 days.

"Starting tomorrow we are moving our standard full mobility out to 30 days for new servers, storage, and relay pricing," says MacDonald.

"Thank you, everyone, for working with us through a pretty rocky time over the last few months," he adds. "During the first half of the year, commodity costs were changing in an extremely dynamic way, which is why we have to shorten our core validities, and our goal right now is to be able to stand behind and extend that ... validity."

Another announcement, this time from Culloty, is a change to how credit works. She says that partners are asking for more credit capacity.

"We're working in a very dynamic environment right now, and ... I'm proud to announce ... an increase to approve to credit items 150% making it easier for you to sell and upsell to your customers," she says.

Fidelma Russo, CTO at HPE, joins Mottram on stage now along with the company's CTO of worldwide technology.

Russo talks about how AI is changing things across the board and – another familiar refrain from infrastructure vendors in general – that businesses are discovering that private cloud can be overly expensive for some workloads. This presents an opportunity in private cloud for everyone.

We'll be hearing more about this in Russo's keynote on Wednesday it seems.

"We are seeing a lot of demand for on-prem infrastructure, especially around AI, and so that's where we see a lot of adoption around private cloud AI for customers who are seeing that from a cost perspective, from a compliance perspective, from a sovereignty perspective, and from control perspective, that this is the best way to deploy their AI workloads. So I think that's a huge opportunity for partners, especially with services," Russo says.

Mottram wasn't joking when he said this would be speedy, we're now onto Rami Rahim, former CTO of Juniper and now in charge of networking at HPE. This is obviously one of the most interesting things for partners this week, as it's been a significant transition especially for Juniper partners.

Asked how it's going, Rahim says: "It's a lot of work, I'm not going to lie, but I would say that we are ahead of schedule at this point. I think any successful integration is going to start with culture and cultural alignment. Now, [I've been] really very positively surprised to see just how aligned from a cultural standpoint Juniper and HPE were right from the get-go, which allowed us to focus on what's important right from the very beginning."

"We want to get the integration right. We could not let the integration de-focus us for what is really critical, and that is our customers, our partners, ensuring that there's no disruption to any of our customer deployments, and very importantly, innovation," he adds.

Rahim is also talking about AI as an opportunity for partners, that embracing AI and getting it right it's a new route to increase profitability for partners.

A spokesperson from Myriad360 is syaing that an opportunity for partners is that "network is becoming strategic again".

"Network... is foundational in any AI competition," he says, "but the interesting thing is it's not alone: so is infrastructure, so is data security ... and many, many more."

That's it from Mottram, Ewington is back on stage to wrap-up (I think)

Ok, so an announcement: for the rest of this FY the Juniper and HPE networking partner programs will continue to operate as two separate programs, but from 1 November 2026 they will be unified into a single stream under HPE Partner Ready Vantage.

Juniper partners will be onboarded at the level they're currently on, which gets a round of applause from the audience – sure there are probably some relieved delegates to hear that officially confirmed.

That's it for the Partner Growth Summit keynote, but stay with our sister site ITPro for all the announcements from this week. Judging by all the hints dropped to channel partners there are going to be a lot!

Cisco aims to consolidate and simplify infrastructure management to compensate for rapid AI advances and surging customer demands, according to the company’s UK&I CTO.

Speaking to ITPro following the company’s annual Cisco Live event, Rob Lay said a slew of announcements made at the conference last week intend to shore up infrastructure optimization capabilities for enterprises.

Chief among these announcements was the launch of the Cisco Cloud Control platform, which aims to consolidate disparate infrastructure management platforms through a single interface.

The move here, according to Lay, comes in direct response to the growing sprawl of disparate, overlapping solutions enterprises contend with, be that from Cisco itself, or third parties.

“Cloud Control is all about bringing together all of those different management platforms that we've got, because we've got a really broad portfolio, bringing together the likes of Catalyst Center and Meraki Dashboard and Nexus One, and the Collab Command Center, and Cisco Security Cloud Control,” he told ITPro.

“All of those pieces, you’re bringing all that together so you’ve got that single point of management.”

Cisco describes the platform as a “new way” to run critical infrastructure that ties together five key elements, including:

• Cross-domain telemetry
• Purpose-built models
• Trusted agents
• Cloud Control Studio
• Cisco AI Canvas

The launch of Cisco Cloud Control is the culmination of several years’ worth of evolution for the networking giant, according to Lay. Indeed, Cisco has been moving consistently toward a “platform approach” when it comes to network and infrastructure management.

“We've talked for a couple of years now about this idea of platform, and I think that it becomes very, very difficult to really have a true platform until you've got that consolidated management capability,” he added.

While in previous years this focus was spurred on by evolving cloud demands, particularly the shift toward hybrid and multi-cloud setups, the advent of generative, and more recently agentic, AI required somewhat of a rethink.

The performance demands of agentic AI are huge, Lay told ITPro. This means performance, stability, and consistency of cloud and networking capabilities are now paramount.

Running parallel to the performance requirements, enterprises are also dealing with an increasingly sprawling array of agents working behind the scenes, which again requires an overhaul of visibility and the ability to track, monitor, and orchestrate these autonomous bots.

“I think it's also driven by the landscape that we're seeing from an AI perspective as well. We can't get through a conversation these days without mentioning AI, but when you look at the challenges that our customers are facing – the need to be able to adapt, manage, respond to stuff at much, much faster speeds is driving a necessity,” he explained.

“We can't have customers bouncing from console to console to console anymore. So it's about bringing together all those different capabilities [and] unifying that experience for our customers.”

Cisco’s ‘AgenticOps’ push

Agentic AI certainly is a key factor in why Cisco customers are demanding more intuitive, unified management capabilities in 2026, Lay told ITPro.

Indeed, the Cloud Control Panel represents the underlying foundation for the firm’s ‘AgenticOps’ operating model. The company envisages a future where humans and agents are working together “side by side”, and the features rolled out by the company aim to simplify adoption.

The firm’s Agent Builder tool, for example, enables users to build custom agents for Cloud Control “tailored to their own policies and workflows”. These can be connected to more than 50 third-party platforms, either through native connectors, or the Model Context Protocol (MCP).

Elsewhere, the App Builder also allows enterprises to build and publish applications and workflows for Cloud Control using natural language prompts. This, the company noted, is based on OpenAI's Codex agent.

Tackling AI’s “network problem”

While agentic AI features and customization capabilities are a key focus for Cisco, Lay noted that Cisco’s bread and butter, networking, is the area in which it’s still emphasizing the importance of relying on high-performance capabilities.

Customers have worked their way through the early days of the generative AI boom, which was dominated by chatbots, and with agents, they’re now focusing heavily on delivering clear-cut use cases that “add value to their organizations”.

Throughout this process, he noted, it’s become clear that “AI rapidly becomes a network problem”.

“It's really, really difficult to enable an AI platform to work properly unless you've got that very highly capable network that underpins it all,” he explained.

“It is kind of true to say that Cisco now is delivering that critical infrastructure for the AI era, because regardless of whether we’re talking about the data center network and the need to connect to GPUs, whether we’re talking about the fact that we’ve got customers, service providers, and hyperscalers now building data centers further and further apart - how do we work to that?”

With the previous generation of generative AI focused largely around chatbots, network traffic typically “spiked” in short bursts. With agentic AI always on and running away in the background, that “profile just goes up and stays high in terms of network demand”.

“That puts a very different challenge and requirement on the network to deliver that real consistent high efficiency, high quality network connectivity,” he said.

Lay said these fundamental “network challenge” issues have prompted Cisco to drive networking innovation in recent years to enable it to meet surging customer demand.

Part of the challenge here, he noted, has been emphasizing that customers simply cannot continue “sweating assets” if they’re to have success with agentic AI (and remain secure).

In a recent discussion with a customer, Lay revealed an IT leader boasted about having a Catalyst Switch that had been “up for 12 years”.

“That’s great, right? That shows our kit is reliable and stable, but in today’s world that’s quite a risk for our customers, not just from a security perspective, but fundamentally, it’s not going to support any of the sort of network traffic types that you need for today,” he told ITPro.

Customer mindsets are shifting on this front, Lay added, and this is an area Cisco is keen to capitalize on and keep pace with.

“When they’re looking at refreshing their technology, they’re not just looking for the current in-support version of what they had previously, they’re looking at what do we need our network to do in two years’ time, in three years’ time,” he commented.

“That’s what we need to be planning for and investing for today.”

FOLLOW US ON SOCIAL MEDIA

The UK is running on fumes in terms of data center capacity, Savills has warned, with surging demand pushing infrastructure to its limit.

Analysis from the real estate services firm found the national vacancy rate has fallen from 27% in 2016 to just 8% in the first quarter of this year.

Things are even tighter in London, where the figure stands at just 7%, thanks to existing cloud ecosystems, network density, and established availability zones. The city now accounts for 1,637 MW of the UK’s total live capacity of 1,803 MW.

“While activity remains anchored to tier-one locations around London, with Manchester cited secondary, beyond these two cities activity is limited reflecting how difficult it remains to deliver network density, occupier ecosystems and timely access to power," said Rupert Duckworth, associate director, Savills Data Centre Advisory.

"Having said that, we are forecasting growth in the North East given its renewable infrastructure and grid connectivity, and by a planning context that is more receptive to large-scale investment. Even so, this is best understood as a medium-term diversification story rather than an immediate release valve for the broader market.”

The firm said that national capacity is likely to increase in the long-term. However, outside the London area, the cloud market is still fairly shallow, as developers favor building on existing infrastructure networks where connectivity is already established and development risks are lower.

In 2024, 191 MW of IT capacity was delivered in the UK as a whole - a record high - followed by an even greater 231 MW last year.

According to Savills, this increase doesn't appear to be continuing, with just 48 MW delivered in the first quarter of this year. Of the 242 MW currently under construction across the UK, around 66 MW is scheduled to complete between Q2 and Q4.

This, the firm noted, shows just how limited the immediate delivery window is, relative to the headline pipeline.

“National supply is expanding on paper, but much of the pipeline remains early-stage and uncertain," said Duckworth. "Pre-letting has become a core feature of the market, with future capacity increasingly committed well before completion."

Capacity constraints could rise

Looking ahead, Savills said the UK is unlikely to see any fall in data center demand, capital, or strategic relevance; the big question is which projects can deliver capacity in a more constrained, regulated, and execution-sensitive environment.

To a large degree, delivery is now shaped by energy strategy as much as by real estate fundamentals.

"Market conversations suggest that, in some of the most constrained West London locations, developers are now planning for multi-year waits for firm capacity, long enough for operators to start to consider alternative power solutions less as a contingency plan than a strategy; using them to bridge the gap to grid capacity and secure position in the right locations,” said Lydia Brissy, Savills' director of European research.

FOLLOW US ON SOCIAL MEDIA

Meta is putting $115 million behind a new program designed to train workers to help build data centres and other AI infrastructure.

America’s Workforce Academy (AWA) will cost nothing to those attending, who will be guaranteed a job upon completion, and initially launch this year as a pilot in Louisiana, Ohio, Indiana, and Texas.

Meta didn't reveal how many positions would be available, which companies would employ these workers, whether they would be short-term construction roles that are usually required for building data centres, and – as Reuters noted – if the jobs would be unionized.

The AWA will help ensure there's enough trained workers to complete the skilled tasks needed to build AI infrastructure, but also acknowledges looming changes to the workplace, with big tech providers claiming the technology will have a serious impact on white-collar jobs.

Meta recently slashed 10% of its workforce, about 8,000 jobs as part of a pivot towards AI. The AWA investment is a small slice of its overall AI investment, with the company expected to spend between $125 billion and $145 billion on AI infrastructure and data centers this year alone.

The $115m AWA investment is an initial pledge to cover the first year pilot in those four states, Meta noted.

"The AI revolution is bringing change but also historic opportunities," said Dina Powell McCormick, Meta president and vice chairman.

“Skilled workers electrified rural America one pole at a time. They manned the factories that built the arsenal that won World War II. Now a new generation will pour the foundations and lay the fiber that secures American strength in this new age."

Will tradespeople be the real winners with AI?

The move by Meta comes just weeks after claims that tradespeople could rank among the real ‘winners’ of the generative AI race.

Nvidia CEO Jensen Huang predicted that the economic benefits of AI won’t be limited to the tech sector, but will also extend to trade workers,

"Electricians, plumbers, iron workers, technicians, builders, this is your time," he told students during a commencement speech at Carnegie. "AI is not just creating a new computing industry, it is creating a new industrial era."

However, many of these jobs are short term. Reuters noted one Meta project would require 1,800 workers to build, but only 100 to operate once the data center was complete.

Workers needed – for now

Meta said in a statement that the US labor market is short of hundreds of thousands of skilled workers and that this is "an incredible opportunity for these American heroes to power America's future."

Rachel Peterson, Meta’s VP for data centers, said filling construction and trades skills gaps requires the same "long-term thinking" that Meta is bringing to AI itself.

"The AI infrastructure we’re building today requires an incredible workforce to make it a reality," Peterson said. "America needs hundreds of thousands of skilled tradespeople — electricians, mechanics, fiber technicians and more — and this program creates clear, accessible pathways into those careers."

Meta will partner with key industry players as part of the AWA, including National Urban League, the Associated Builders and Contractors (ABC), and CBRE.

The AWA will make use of the ABC's existing educational ecosystem. A similar previous effort, fiber installation training program Level Up, saw 35,000 applications in its first week.

"The sustained demand for data center construction technicians means the industry needs an all-of-the-above approach to grow the construction talent pool," said Michael Bellaman, ABC president and Chief Executive Officer.

FOLLOW US ON SOCIAL MEDIA

AMD has announced plans to invest £2 billion in the UK as the chip maker aims to accelerate AI innovation and research, and expand access to compute resources.

The five-year plan, announced at London Tech Week, will see the firm form new strategic partnerships with Imperial College London and Oriole Networks, along with support for the University of Cambridge's Zenith AI supercomputer and Sunrise fusion AI system.

“This investment is a major vote of confidence in Britain’s place as a global AI superpower. We’ve got the talent, the world-class universities and the ambition to lead, and partnerships like this help turn that potential into real progress,” said chancellor Rachel Reeves.

"It will drive more cutting-edge research here in the UK, open up opportunities for people to build the skills they need for the jobs of the future, and speed up breakthroughs that can improve people’s lives and grow our economy.”

The collaboration with Imperial College London is aimed at advancing computational science and supporting research that relies on large-scale computing resources, including healthcare innovation and climate modeling.

The two also plan to explore opportunities to optimize AI models, scientific workflows and data-intensive applications on AMD compute platforms and AMD ROCm open software.

“The United Kingdom has the talent, research excellence and ambition to help lead the next era of AI,” said AMD chief executive Lisa Su.

“AMD is proud to deepen our commitment to the UK and work with partners across government, academia and industry to expand access to the compute infrastructure needed to advance sovereign AI, accelerate discovery and drive long-term economic growth.”

AMD pens Oriole Networks deal

AMD has also announced a collaboration with Oriole Networks in support of the UK's Advanced Research and Invention Agency (ARIA) Scaling Inference Lab.

This is a national initiative designed to address critical AI infrastructure bottlenecks and support what's expected to be the world's first large-scale AI system powered by a pure photonic network.

AMD and Dell Technologies, meanwhile, are working with the University of Cambridge on expanding national AI infrastructure. This includes plans to develop the new Zenith AI supercomputer and the Sunrise fusion AI system developed in collaboration with the United Kingdom Atomic Energy Authority (UKAEA).

Zenith is a new UK AI-for-science platform, funded by the Department for Science, Innovation and Technology (DSIT) and UK Research and Innovation (UKRI), while Sunrise is a second AI supercomputer being built now, funded by the Department for Energy Security and Net Zero (DESNZ), owned by UKAEA, and operated by the University of Cambridge.

Together, they will support a broad range of AI-for-science applications including healthcare research, climate modelling, materials science, engineering simulation, fusion research, and scientific AI model development.

The AMD investment comes alongside the launch of a new £1.1 billion government plan to boost the UK's chip industry and attract more AI funding.

“This investment reflects the strength of Britain’s talent, research and ambition in AI - but also the infrastructure we are putting in place to match it,” said technology secretary Liz Kendall.

"With world-class chip designers, leading universities, and partners such as AMD choosing to invest here, we are building the compute capability needed to power innovation, drive growth, create jobs, and ensure the most advanced AI technologies are developed in the UK."

FOLLOW US ON SOCIAL MEDIA

The UK government is continuing its push to boost the country's AI infrastructure capabilities with a £1.1 billion plan to back home-grown chip firms.

The new AI Hardware Plan, announced by technology secretary Liz Kendall at London Tech Week, will see £750 million allocated for the development of a new national AI supercomputer.

This includes £400 million for next-generation AI chips, of which £150 million is an advanced commitment to buy novel chips from both existing British firms and innovative startups.

"AI is the defining currency of economic and hard power in today’s world and the countries that control the hardware behind it will hold the keys to the future," said Kendall.

"The UK is already a global leader in chip design, and I believe this is a race Britain can win. To do that, we must back more British AI – and that means investing in the chips, computing power and skilled people behind it."

The new national AI supercomputer will be one of the most advanced in the world when deployed in 2030, based on a heterogeneous mixed chip system using both proven and next-generation processors.

The government hopes to see British-designed chips used in the system, which will form part of the UK’s AI Research Resource for researchers, startups, and public services.

£150 million will go to buy next-generation inference chips, with a further £250 million for more specialized chips, Downing Street confirmed. An additional £120 million will fund a new AI Hardware Innovation Programme to help British companies design, develop, and test innovative novel chips.

At least £20 million will be used to expand the Scaling Inference Lab, delivered by ARIA and CommonAI, to help companies prove their technology, attract investment, and secure partnerships with global tech firms.

Elsewhere, £45 million has been pledged to build skills capabilities, including doctoral training and undergraduate bursaries to train more engineers, chip designers, and technicians.

Finally, a new fund led by Silicon Valley investors Playground Global and backed by up to £150 million from the British Business Bank - its biggest investment ever - will invest in UK-based AI hardware companies.

AI Hardware Plan welcomed by industry

The announcement by Kendall came after prime minister Keir Starmer announced a £400 million investment for specialist AI chip purchases.

The two announcements have been welcomed by industry stakeholders as a positive step toward bolstering the country’s infrastructure capabilities – a focus that will be vital if the country is to position itself as a leading AI economy in the future.

Andy McLean, CEO of the UK Semiconductor Centre (UKSC), said semiconductors are the “foundation on which AI is built”, adding that the UK has a “significant opportunity” to play a leading role in global research and development.

“The UK’s semiconductor strengths are well aligned to opportunities across the AI stack, and we are already seeing substantial private investment in UK semiconductor companies operating at the forefront of these technologies,” he said.

“It is encouraging to see the government matching that ambition with new investment and a clear focus on accelerating knowledge, skills, scale, and deployment."

Steve Brierley, CEO and founder of Riverlane, echoed McLean’s comments, noting that the focus on skills will be critical.

“The government's commitment to supporting high-growth technology companies recognizes that breakthrough innovation alone is not enough. Alongside access to capital and specialist skills, the government can play a powerful role as a smart buyer, helping to create markets, accelerate innovation and strengthen the wider technology ecosystem,” he said.

“Ensuring the UK has the infrastructure, talent and expertise to bring these capabilities together will be critical to ensuring the UK captures the full value of its investments and strengthens its position as a leader in advanced computing.”

Looking ahead, Brierley said the UK must adopt a “holistic view of the future of computing”, focusing on how high-performance computing (HPC), AI, and quantum technologies can work together effectively.

FOLLOW US ON SOCIAL MEDIA

When we talk about wireless connectivity, often what we’re thinking about is WiFi, however mobile connectivity such as 5G is still important for many sectors.

What happens to a business when mobile connectivity is inadequate or fails completely?

On this episode of the ITPro Podcast Jane and Bobby are joined by Paul McHugh, VP of EMEA sales at Ericsson, to discuss how wireless connectivity issues can affect businesses.

Highlights

"Businesses believe that in terms of delivering the best business outcomes that the most operational efficiency that they can, they recognize now, perhaps more so than in the past, that networking underpins that on the basis that so many of the services and the processes that they run are technology based, and the majority of that technology needs to be connected."

"Historically, when we talk about resilience in networking, we just talk about more lines, okay, more more cables, more fiber going into a location. That doesn't always offer true resilience, particularly if all of those lines ... actually terminate at the same local exchange, for example. What is much more appreciated now, and this isn't just the view of networking vendors like Ericsson ... is that to achieve true resilience in a network, you actually need a variety of types of connectivity"

Footnotes

• Three and Ericsson just launched a first-of-its-kind managed 5G service for businesses
• ePrivate 5G and partner ecosystems: The blueprint for intelligent infrastructure
• Sign up to the ITPro Newsletter

Microsoft has hailed a significant breakthrough with its Majorana 2 quantum chip, which it claims is vastly more powerful than its predecessor – and AI helped make it possible.

Developed with the help of Microsoft Discovery’s agentic AI, the Majorana 2 topological quantum chip's cubits can maintain their quantum state 1,000 times longer than the first generation.

It has a mean qubit lifetime of 20 seconds, with some instances lasting as long as a minute.

“We need to make improvements each year that will get us closer to delivering a computer that we believe will have massive commercial and societal value,” said Chetan Nayak, Microsoft technical fellow.

“We’ve got to keep marching to that roadmap to accomplish that, but where are we relative to last year? We’re 1,000 times better.”

While the original Majorana superconductor was based on aluminum, Majorana 2 uses lead, which has delivered marked efficiency improvements. By using lead, this helps shield fragile qubits from cosmic disturbances that can make them unstable.

“That was actually a fairly large change, and it led to big, big improvements in device quality,” Nayak said.

AI played a key role in Majorana 2 improvements

Microsoft specifically highlighted the use of Microsoft Discovery, designed to help organizations embrace Frontier R&D, in the project.

This combines specialized AI agents for scientific research and development with a Discovery Engine that drives research and reasoning workflows, along with enterprise-level security, governance, and transparency.

The system was used to manage workflows, automate measurements, optimize fabrication, pinpoint previously unnoticed flaws, and propose new solutions.

Critical parts of the Majorana quantum devices are designed atom by atom, the company noted. To keep each atom in the right spot, another material, an impurity, can be added to the crystalline structure.

Zulfi Alam, corporate vice president for quantum at Microsoft, said the system enabled the team to spot probable targets and fine-tune development processes.

Meanwhile, it also helped handle nearly two decades’ worth of data, in many different formats.

“As you run AI agents on this data, they’re able to essentially resynthesize and make correlations that we as humans cannot see because no single individual has that much vision across that much data,” said Alam.

Supercharged development

Creating a topological state requires setting hundreds of parameters, a process that, when carried out by a human, takes weeks – but Microsoft said the use of agentic AI cut this by orders of magnitude.

Alongside this, AI’s pattern-recognition abilities helped with the difficult task of measuring what state the qubit is in and detecting whether there’s an even or odd number of billions of electrons on a semiconductor wire.

“Using agentic AI to automate the measurements was a game changer,” said Alam. “It goes through some math and starts saying, ‘Hey, where do I find the lowest point where everything sort of works?’ And it can do all these voltage adjustments in parallel, which a human cannot do. The way our minds work, we are more linear.”

Microsoft has now made Microsoft Discovery generally available, along with an app with core capabilities that individuals can download for free and run locally on their computers with a GitHub Copilot account.

“In the year since we launched, we’ve seen customers light up use cases across critical industries like life sciences, chemicals and materials, energy, manufacturing and consumer goods,” said Aseem Datar, corporate vice president, product innovation for Microsoft Discovery.

“With companies like Syensqo developing next-generation fluids for semiconductor manufacturing, the opportunities for impact are vast.”

FOLLOW US ON SOCIAL MEDIA

Nearly nine-in-10 UK business leaders reckon that quantum computing will disrupt their sector in the next four years, according to new research from EY.

A survey from the consultancy found banks and payments processors exploring the use of quantum to detect and prevent financial fraud and money laundering by 2030, while automotive manufacturers are testing it to help optimise electric vehicle battery development and traffic flow.

Across the board, businesses are taking the prospect seriously, with 35% saying they've made quantum computing a strategic priority for the next five years.

Among organizations operating in the financial services sector, the figure's as high as 67%, while it's far lower in real estate, hospitality, and construction firms, at just 17%.

“Our latest report shows that, while UK businesses are continuing to invest in and evaluate the impact of quantum computing, expectations of its maturity remain cautious," said Piers Clinton-Tarestad, technology risk partner at EY UK.

"This means that although long-term disruption from quantum computing is widely expected, including its impact on cybersecurity, the proportion of companies taking immediate steps to prepare remains relatively small."

Business leaders don't see quantum computing as an immediate issue, with 59% believing the technology is unlikely to develop sufficiently to play a significant role in core operations until from 2030 onwards.

This means that talent isn't an urgent problem either, with only 13% of UK leaders planning to recruit the talent they think they will need within the next two years.

This is despite 83% seeing loss of competitive advantage as the main risk when it comes to not adopting the technology.

‘Q-Day’ predictions are evolving

While the EY survey found quantum isn’t viewed as a near-term issue, predictions over ‘Q-Day’ - the point at which quantum computers can crack traditional encryption methods - have changed in recent months.

Most estimates on this timeline have typically placed this tipping point in the mid-2030s. However, Google recently revised its timeline to 2029 and issued a warning that most organisations aren’t prepared.

In a March blog post, the tech giant said the timeline revision came in direct response to progress in quantum hardware development and quantum error connection advances.

Quantum brings new risks for enterprises

Quantum could present businesses with an array of risks and challenges, according to the EY survey, particularly with regard to outdated IT infrastructure and compliance with future regulations.

Both of these issues were cited by around eight-in-ten business leaders, for example.

The top four barriers to quantum adoption are its complexity – picked by 80% – followed by market uncertainty at 66%, integration challenges at 60% and current talent shortages at 47%.

“For some business leaders we speak to, hesitation stems from not knowing where to begin exploration. One effective first step is to identify one or two practical ways in which quantum can be used, linked to existing pain points, before assessing which teams or processes would be most affected, and any gaps in skills, tools or data," said Clinton-Tarestad.

"In the same way that an orchestra needs a conductor, effective quantum implementation needs a senior sponsor within a business with cross-functional oversight to coordinate efforts, secure resources, and help ensure quantum readiness becomes part of broader digital transformation plans.”

Another reason organizations aren't moving faster to quantum computing is a lack of clear cut use-cases, according to recent research from QuEra Computing.

Analysis from the company found that while 44% of enterprises expect quantum budgets to increase in the year ahead, even more (46%) expect them to remain flat - and 10% are expecting budget cuts.

QuEra noted that the quantum market is shifting from “hype-driven investment” to a “proof-driven discipline”, with organizations seeking clear evidence of value before committing to investment.

FOLLOW US ON SOCIAL MEDIA

Periods of high demand, whether driven by seasonal spikes, major events, or unexpected surges in usage, are no longer just a test for end-user organizations. They are also a direct test of the partners and providers responsible for keeping critical systems running.

When things go wrong during these high-pressure periods, the impact can be significant. The 2025 cyber attack on Marks & Spencer reportedly affected its H1 2025 performance, while the effects of the Jaguar Land Rover attack rippled across its supply chain.

As IT environments become more distributed and interconnected, the responsibility for maintaining uptime, performance, and security increasingly sits with Managed Service Providers (MSPs), network providers, and other channel partners. When outages or cyber incidents occur, customers expect immediate response, rapid recovery, and minimal disruption.

From service provider to operational owner

Rapid digital transformation is sweeping across businesses, with 94% of private sector IT leaders being familiar with smart spaces. Modern environments rely on a complex mix of cloud platforms, IoT devices, remote endpoints, and real-time applications.

These systems are deeply interconnected, with network performance underpinning every business function – so the need for strong, secure connectivity has grown in parallel.

MSPs are no longer just providers; they are embedded within customer operations. When something fails, they are expected to respond, coordinate, and resolve issues as part of a broader operational responsibility.

Complexity at scale

Today’s IT environments are more complex - and more fragile - than ever. During peak trading periods, network traffic can surge unpredictably, cloud services must scale in real time, distributed workforces require secure, reliable access, and connected devices continuously exchange data across environments.

These connected systems also magnify the consequences of any network failure. When connectivity falters, the impact is immediate and widespread, affecting productivity, service delivery, and customer experience.

At the same time, every endpoint and connection increases the potential attack surface. Security and networking have become interdependent challenges that must be addressed together.

For MSPs and connectivity providers, managing this complexity is now a core part of the value they deliver.

Network connectivity to power resilience

Traditional approaches to networking, which focused primarily on uptime and capacity, are no longer sufficient. Environments must now be designed with disruption and dynamic responses in mind.

Most organizations require dependable networks to keep operations running smoothly, particularly during high-demand seasons when every minute of downtime can impact revenue and customer experience.

High-capacity 5G connectivity offers the speed, reliability, and scalability needed to support IoT devices, mobile (POS) systems, digital signage, and other emerging technologies. Its adoption can also help organizations consolidate network providers, simplifying management and freeing up budgets that can be reinvested into innovation.

Hybrid working has reshaped operations across both the public and private sectors. Head office teams, customer service staff, and fulfilment managers often operate from different locations, all relying on cloud-based systems to access sensitive information. This makes secure access an essential requirement. Robust, always-on networks ensure that all parts of the business remain operational.

Investing in network infrastructure capable of supporting emerging technologies such as AI and IoT gives businesses the flexibility to scale operations and adapt quickly during trading peaks while preparing for future innovation.

Mapping your network

To survive busy periods, leaders must treat connectivity not as a technical upgrade but as a strategic investment. Infrastructure resilience begins with understanding where weaknesses exist, whether through outdated networks, insufficient bandwidth, or fragmented security policies.

Scalable connectivity technologies, such as SD-WAN and high-capacity 5G, allow networks to adjust dynamically when demand spikes or primary circuits fail.

At the same time, managing the growing number of connected devices is critical. Platforms that centralise IoT and smart devices reduce operational complexity and prevent issues being overlooked that could lead to unpatched vulnerabilities.

Security is an essential part of resilience. Secure Access Service Edge (SASE) integrates networking and security into a unified framework, providing centralised control over access, identity, and data protection.

Combined with scalable connectivity, SASE ensures that IT teams can maintain operations, safeguard sensitive data, and respond rapidly to emerging threats, keeping systems online and customer experiences seamless, even during peak trading periods.

The power of networks

Busy periods place extraordinary pressure on networks at precisely the time when cyber threats and IT faults pose the greatest risk.

What was once considered background infrastructure is now a direct driver of revenue, brand reputation, and operational continuity.

By investing in strong network foundations today, organizations of all sizes can safeguard their most critical trading periods and build the resilience needed for long-term innovation in an increasingly digital future.

In 2016, Dell Technologies completed its acquisition of EMC. While its name lived on for another four years under the Dell EMC brand, it finally disappeared entirely in 2020.

You may wonder what this wander down memory lane has to do with anything – after all, the title of this article clearly states I was at Dell Technologies 2026 but bear with me. You see, for those of us who have been in this arena long enough, there may have been the sense that the ghost of EMC was haunting proceedings at this year’s annual Dell get-together.

First, a bit of background.

While the EMC Federation comprised a variety of vendor types, EMC itself was firmly in the enterprise storage business; according to a 2014 Forbes article, about 70% of the Federation’s revenue came from the EMC part of the business. While there was undoubtedly the attraction of VMware and Pivotal, acquiring the technology behind VMAX, Isilon, Data Domain and ScaleIO would have been a prime mover for the acquisition.

According to a 2025 retrospective analysis by M&A Watch: “Dell's acquisition of EMC was primarily driven by the need to transition from a PC-centric business model to a comprehensive enterprise IT solutions provider. By 2016, Dell's traditional market--personal computers and low-margin hardware--was stagnating due to declining growth rates and intense competition from HP, Lenovo, and Apple. This acquisition represented a strategic shift into higher-margin enterprise IT markets, including storage, cloud computing, and virtualization.

“Before the acquisition, Dell had a strong foothold in servers and networking hardware but lacked a comprehensive enterprise IT infrastructure offering. EMC's dominance in high-end data storage complemented Dell's existing portfolio, enabling the company to offer a full-stack IT solution.”

In May 2016, what is now the Venetian Conference Center Las Vegas hosted the final standalone EMC World with the strapline Modernize. Across the three days, this was split into three ideas: Modernize the Industry, Modernize Your Data Center, and Modernize Your Business.

With one of the tentpoles of Dell Technologies World 2026 being “The Modern Data Center”, (finally, I get to my point) I can’t help but feel an echo of that final EMC conference.

What is the Modern Data Center, anyway?

2026’s modern data center is, of course, all about AI and having the compute and storage available to carry out large AI workloads without resorting to the public cloud. Dell Technologies launched five new products, within the confines of existing lines, at Dell Technologies World 2026:

• PowerStore Elite. A descendant of Unity, the final storage ranged launched by EMC, PowerStore Elite has been pitched as “the biggest leap forward in the platform’s history”, with built-in AI software to handle load balancing and improve efficiency.
• 18th gen PowerEdge servers, which the company says offer up to 70% better performance. Many of the new servers are advertised as being excellent for AI workloads and high performance computing.
• PowerProtect One, a cyber resilience platform that brings together Dell PowerProtect Data Domain and Dell PowerProtect Data Manager into a single offering.
• Cyber Detect, an AI-powered ransomware detection product that plugs directly into PowerSroew and PowerMax enterprise storage.
• Dell Automation Platform, which itself consists of
  • Dell Private Cloud
  • Dell Distributed PRivate Cloud (previously known as Dell NativeEdge)
  • Dell Automation Studio, and
  • Dell Automation Platform.

The modern data center range, if we can call it that, wasn’t the only chunky enterprise hardware on offer, either. Updates to Dell Technologies AI Factory with Nvidia included the introduction of PowerRack, a rack-scale storage system that comes in three flavours depending on a company’s needs.

There was also the introduction of Dell Deskside AI, which, while it runs of various specialist PCs like the Dell Pro Max with GB10 and and Pro Max with GB300, is a firmly enterprise play – not just a business one.

If Dell Technologies 2025 was the year of the AI PC, 2026 was a hard push to reassert itself the data center as use of AI agents accelerates – and costs rocket in response.

The direction of travel, for the next 12 months at least, seems fairly clear to me: Dell is pulling on its own data center heritage and in particular the heritage acquired from EMC to position itself as the answer to the burgeoning cost of AI usage. This is particularly true of anything that uses a lot of memory or compute.

It’s funny how much and how little change a decade of progress can bring.

The US Department of Commerce is handing out just over $2 billion to US quantum firms, with half going to IBM.

The aim, said the department, is to strengthen the US's position, with the sector crucial to national security, technological resilience, and long-term strategic leadership.

"With today's CHIPS research and development investments in quantum computing, the Trump administration is leading the world into a new era of American innovation," said secretary of commerce Howard Lutnick.

"These strategic quantum technology investments will build on our domestic industry, creating thousands of high-paying American jobs while advancing American quantum capabilities."

The Department of Commerce will receive a minority, non-controlling equity stake in each company.

Much of the funding will go to two quantum foundries. IBM gets the biggest slice, $1 billion, to establish a new quantum foundry subsidiary for quantum-grade superconducting wafers.

GlobalFoundries, meanwhile, will receive $375 million to establish a secure, domestic quantum foundry for leading architectures and multiple modalities – superconducting, trapped ion, photonic, topological, and silicon spin – used in large-scale quantum computers.

"IBM has pioneered quantum computing for decades. Our work in silicon wafer fabrication has been a key to IBM's success and will be critical to enable a broader quantum technology landscape that will reshape global innovation and economic competitiveness," said Arvind Krishna, chairman and CEO of IBM.

Another seven quantum computing companies are also in for funding, aimed at solving currently unresolved engineering problems in multiple quantum modalities.

"The CHIPS R&D Office is taking a portfolio approach to strengthen and accelerate US leadership across multiple quantum modalities at once, while focusing each award on discrete technological problems of genuine consequence," said Bill Frauenhofer, executive director of semiconductor investment and innovation.

"We will be providing incentives to build domestic quantum capacity, solve the hardest engineering challenges, enable multi-year acceleration of technology roadmaps, and drive continued U.S. quantum leadership."

Atom Computing, Diraq, D-Wave, Infleqtion, PsiQuantum, Quantinuum, and Rigetti will receive funding of between $38 million and $100 million.

"The award would accelerate D-Wave's ability to scale quantum innovation domestically, expedite key fabrication processes, and deliver real-world quantum applications to our global customers today," said Alan Baratz, CEO of D-Wave. "We see this as a transformative moment for not just D-Wave, but also for quantum computing and the United States."

Quantum computing is being seen as a crucial technology for national defense, advanced materials, biopharmaceutical discovery, financial modeling, and energy systems.

And this US move follows an even larger investment in quantum technologies by the British government, which recently announced £2 billion in funding for the sector.

Between them, quantum computing, quantum communication, and quantum sensing are expected to generate up to $97 billion in revenue worldwide by 2035, according to research from McKinsey. Quantum computing will represent the biggest slice, growing from $4 billion in revenue in 2024 to as much as $72 billion in 2035.

The Department of Commerce is also now looking for proposals for research, prototyping, and commercial solutions that advance microelectronics technology in the US; applications can be made under announcement 2025-NIST-CHIPS-CRDO-01 at www.grants.gov.

Research and consultancy group Wood Mackenzie has warned that the power requirements of AI data centers are growing too fast for national grids to keep up.

With grid transmission build-outs five to ten years away, the firm said, projects, markets, and consumers are all at risk. While data center operators are pursuing colocated generation and flexible interconnection models, these projects face far greater technical, regulatory, and economic hurdles than the industry understands.

"The power sector is fixated on data center flexibility, but that is not the end-game for grid operators or data center operators," said Ben Hertz-Shargel, global head of grid transformation and large loads at Wood Mackenzie.

"Firm grid service is the goal, backed by new transmission superhighways. But there is a lack of awareness throughout the power sector about the technical and regulatory risk confronting colocation projects, and the business risk of conditional interconnections."

Wood Mackenzie's figures see 16.4GW of gas capacity additions per year through 2035 to meet projected demand, despite only 4GW per year having been added between 2023 and 2025.

"Load growth and affordability are in direct opposition in the deregulated markets," said Chris Seiple, vice chairman, energy transition and power and renewables, Wood Mackenzie. "If prices rise to the level necessary to incentivize new generation, it will raise prices for all customers, prompting a political outcry."

In one example, mid-Atlantic grid operator PJM has 78 gigawatts of committed data center load against only 36 gigawatts of accredited generation capacity in its pipeline. The firm's attempting to deal with this by creating one elevated price tier for new resources contracted by large loads, alongside a lower tier for existing resources.

However, Wood Mackenzie warned, this could mean that existing gas and coal plants receiving lower capacity prices may retire, threatening reliability.

In Texas, meanwhile, current market prices of between $30 and $40 per megawatt-hour fall far below the $78-$100 needed to attract new gas generation.

Many data center operators are attempting to avoid the problem through colocation – there's currently more than 90GW of colocated generation now in US interconnection pipelines, indeed. However, the report finds that this is only achievable for the most sophisticated and well-capitalized hyperscalers.

Technical challenges include near-instantaneous changes in AI power demand, which can damage reciprocating engines and gas turbines. And while lithium-ion batteries can be used as a buffer to prevent this, they have a short useful life.

Battery response times need to be extremely short, relying on technology that has not been widely commercialized, and the irregular way in which AI cooling and GPU loads consume power introduces power harmonics, which, if unfiltered, cause equipment to overheat and degrade. Sub-synchronous oscillations, meanwhile, pose a fundamental stability risk to local generators and to distant ones on the transmission system.

And on top of this, the need to protect against downtime exposes data center companies to significant regulatory risk.

There's also, said the firm, a fundamental problem with cost allocation. While grid operators are planning close to $100 billion in transmission investments, much of this could be spread across all existing ratepayers, rather than assigned specifically to a data center.

"Grid operators are positioning flexible interconnections as a stopgap, not a long-term solution," said Seiple.

"The expectation – and often the requirement – is that transmission will eventually provide complete, firm service to large loads. That could cause costs to rise for existing customers if cost allocation methodologies aren't changed and if the data center demand doesn't materialize as forecast."

Data center operators are all too aware of the problem. Research from Savills late last year found that only 850MW of power capacity was delivered across EMEA during the first three quarters of 2025, down by 11% compared with the same period last year. They've been attempting to deal with this by concentrating facilities in the most favorable regions, with some pivoting to emerging markets offering more accessible land and power.

Dell Technologies has launched a new storage platform in its PowerStore range that it claims offers "breakthrough" performance and efficiency.

Dell PowerStore Elite, as the new product is called, comes in three new models – Dell PowerStore 1500, Dell PowerStore 5500, and Dell PowerStore 9500. Depending on the model, Dell PowerStore Elite offers up to three times more performance compared to the PowerStore 1200T model and three times more throughput than the 9200T models, according to preliminary internal testing.

Varun Chhabra, SVP of infrastructure (ISG) and telecom marketing at Dell Technologies, said PowerStore Elite is the "most substantial update for the [PowerStore] platform in the last two years" and that it "represents the biggest leap forward in the platform's history".

"It's a full refresh of hardware and software, and builds on the success of PowerStore Prime from two years ago," he told press ahead of the announcement. "[It's] designed for what customers are facing next, and its timing is really, really important as... infrastructure planning is becoming far less predictable."

Dell PowerStore Elite specs and features

In terms of specs and features, Dell PowerStore Elite offers:

• Xeon Scalable processors, with up to 50% more CPU cores than PowerStore 3200T and PowerStore 5500
• DDR5 memory
• PCIe Gen 5 support
• 200Gb RDMA node interconnect for better load balancing and failover
• Up to 40 drives and 5.8PB effective capacity in a single 3U chassis
• Up to 3x density than prior generations on low-profile E3 NVMe flash
• Up to 40 network ports per appliance with 64Gb FC and 200/400Gb Ethernet-ready connectivity.

It also offers a guaranteed 6:1 data compression.

AI-based efficiency improvements

Continuing the theme of Dell Technologies World, AI is a key feature in PowerStore Elite. According to the company, built-in AI continuously balances workloads, tunes performance, and improves efficiency, which, when put together, reduces manual effort by 95% compared to traditional array management. It also offers Dell Private Cloud integration.

"PowerStore Elite is not just a faster array, it is a brand new class of a modern data platform that is built to help customers lead through change, not just react to it," Chhabra said.

PowerStore Elite will be available worldwide from July 2026.

Dell Technologies has unveiled a new, integrated rack-scale product for AI workloads that it claims can go from delivery to running live AI or HPC workloads in as little as six-and-a-half hours.

Dell PowerRack was announced at the company’s annual Dell Technologies World conference in Las Vegas as one of several updates to Dell AI Factory with Nvidia that were announced on the opening day keynote.

“One of the most common challenges customers have with AI infrastructure is dealing with the complexity of bringing all the pieces together,” Varun Chhabra, SVP of infrastructure (ISG) and telecom marketing at Dell, told journalists and analysts ahead of the conference. Dell PowerRack, he said, brings together AI compute, AI networking, and AI storage as scalable units designed, tested, and validated as one system.

“With Dell PowerRack, customers no longer need to buy components and hope they work together. They can deploy a fully integrated system in which the thermal design, the power management, and the software for infrastructure management have all been engineered to work as one unit,” he said.

Dell PowerRack is available in compute, networking, and storage configurations.

PowerRack for Networking has over 800 Tb/sec of switching capacity with eight Dell PowerSwitch SN6600 LG Ethernet switches per rack, which Chhabra said enables “the high-volume, low-latency east-west traffic required by GPU-dense environments”.

PowerRack for Storage, meanwhile, is a software-defined system based on Dell Exascale Storage and incorporates Dell PowerFlex. The addition of PowerFlex brings block storage into the equation. This means it can be used not only for AI and HPC workloads but also for very demanding but more traditional enterprise workloads.

Dell PowerRack for storage is, according to Chhabra, “the only 4-in-1 storage that's built for extreme-scale HPC, HPC AI, and demanding enterprise workloads”.

ITPro’s managing editor Jane McCallion is at Dell Technologies World all week. Click here to follow her live blog, or here for all our coverage of the event, including news and analysis.

Construction is the world's largest industry by output and among its least digitized. McKinsey estimates that global construction output could hit $22 trillion in 2040. In order to reach this, it will have to meet a compound annual growth rate of 3.2% – a target that requires an industry-wide productivity transformation.

The gap between potential and reality is stark: McKinsey also found that 98% of megaprojects finish behind schedule and over budget. It’s also a naturally risk-averse industry, as when you’re pouring concrete or hanging steel, there are no “undo” buttons.

Intelligent technological adoption can help resolve some of these issues – and when you zoom out, it becomes clear that retail giants such as McDonald’s, 7-Eleven or Dollar General know this better than most. These are some of the largest builders in the world: for example, McDonald’s operates tens of thousands of locations globally. Supporting that kind of footprint is an army of project managers, coordinators, architects and vendors that ensure every store is built to a brand standard.

Building and renovating stores is a core skill, but not their core business. Their business is serving customers and growing the brand – but they cannot grow without building.

Notably, these companies already have some of the most structured data in the industry: playbooks and design manuals, brand standards and prototypes, checklists and approval flows, typical schedules and budgets, which create repeatable store types.

“If you compare a skyscraper to another skyscraper, there might be only 20% overlap in how they’re built,” says Alim Uderbekov, founder and CEO at Surfaice Corp. “But if you compare a McDonald’s or a 7-Eleven store to the next one, 80% of the process looks the same,”

“That’s why it’s far more practical to train AI to build the next McDonald’s than the next Empire State Building.”

There’s evidence the construction industry is already taking note. Use of AI in the construction market is growing rapidly regardless — from an estimated $11 billion in 2025 to a projected US$27.92 billion by 2031, a compound annual growth rate exceeding 16.6%, according to Mordor Intelligence. Surfaice is one player in an increasingly crowded field.

Uderbekov grew up in Kazakhstan, the son of a construction engineer and an architect.

"Growing up around construction sites, I was always struck by the contrast. I’d see incredibly advanced CAD software, and then I’d walk onto a jobsite with [my] dad where there was far less intelligence built into how work was coordinated. That difference confused me early on, and it stayed with me. I kept thinking how this can be improved,” said Uderbekov.

Aggregating data

While the team Surfaice focus on the retail developer's end-to-end workflow, nPlan is attacking the fundamental planning lie that bedevils every large capital project: the optimistic schedule. The company has assembled what it claims is the world's largest dataset of completed construction programs – more than 750,000 project schedules representing over US$2.5 trillion in capital spend. That archive is the foundation for machine learning models that forecast how individual schedule activities will actually perform, rather than how planners hoped they would.

The distinction matters because traditional schedule risk analysis relies on human estimators applying subjective probability ranges to activities –an approach that academic research has consistently shown to be infected by optimism bias. nPlan's AI-driven Schedule Risk Analysis (AI-SRA) instead learns from how comparable tasks have actually played out across thousands of similar projects, producing probability distributions that are, in the company's assessment, "unattainable by human operators".

The distinction matters because traditional schedule risk analysis relies on human estimators applying subjective probability ranges to activities — an approach that can be affected by optimism bias. Instead, nPlan's approach is to learn from how comparable tasks have actually played out across thousands of similar projects, producing probability distributions that are, in the company's assessment, "unattainable by human operators".

It’s clear that standard large language models (LLMs), to which users send a prompt and wait for a response, aren’t the primary route forward for AI in construction.

When you look closely at retail development, there are some basic demands:

• Context and memory across many months of a project
• Multi-step reasoning and planning
• Operation of third-party tools
• Deep reliance on a brand’s documented best practices and standards
• Coordination across many data sources

Uderbekov explains that he pushed his team to consider how to train AI on construction manuals, playbooks, and real projects of repeatable retail development. The idea was to build an agent that knows how to build a store.

“Most construction software treats the work as a series of isolated problems: one tool for scheduling; another for budget tracking; another for drawings; another for permitting; another for field surveys,” Uderbekov explains.

“Yes, that’s all very useful, but it’s still fundamentally horizontal. No software takes responsibility for the whole outcome. The real shift for me was realizing that AI shouldn’t optimize a task; it should be able to help build the entire store,”.

A project manager understands the entire project vertically: the sequence of tasks, dependencies, brand rules, city constraints, vendor reliability and the risk profile of a site.

This matters because in large retail programs, the engineering challenge grows exponentially with scale. A single store requires hundreds of milestones spanning design, surveying, permitting, procurement, inspections and turnover. Multiply that across 100 stores, and the workload quickly becomes more than 10,000 interdependent tasks, each one a potential source of delay.

Another company in the space is Buildots, which provides customers with 360-degree cameras mounted on their hard hats during routine site walks. These images are then processed by AI models that automatically compare physical site conditions against the BIM model and project schedule.

The commercial results reported by clients are notable. UK contractor Sir Robert McAlpine deployed the platform across more than 260,000 square meters of live projects, using it not just for progress tracking but for subcontractor billing verification and quality audit trails. Intel is now using Buildots to support its global fab expansion, with the company claiming the platform has helped Intel avoid roughly four weeks of construction delays per fabrication facility.

Computer vision models also have clear potential in construction. Cranes are among the most capital-intensive assets on any vertical construction site, and their utilization is notoriously difficult to measure.

Turner Construction deployed CraneView, an Internet of Things (IoT) based tool by construction tech firm Versatile, on two tower cranes at its Manchester Pacific Gateway project in San Diego. The firm was able to use the tool to analyze each crane lift and classify them using machine learning, to build a log of idle periods and surface patterns that superintendents can act on in real time.

Analysis revealed when cranes were sitting idle or handling uneven load distributions, which allowed the project team to rebalance crew schedules and crane assignments. The result was the early demobilization of one of the two tower cranes, generating both time and cost savings that tower crane rental rates – which can run to tens of thousands of dollars per month.

Versatile's approach reflects a growing recognition that the construction site's data problem is not just about integrating existing software systems — it is about creating instrumentation where none existed before. Cranes, excavators, and scaffolding systems that were previously invisible to any digital layer are becoming active data sources, producing the raw material for the next generation of construction AI.

Quantifying results

To quantify the effects of digital transformation on construction sites, Surfaice’s engineers have been evaluating impact using a “total attributable value” model with three components. The first is direct cost reduction, where workflow automation and earlier risk detection have cut delivery costs by as much as 30% in conservative scenarios.

The second component is error-related cost reduction. In many retail projects, design and coordination mistakes can account for 10–15% of total cost; by enforcing guardrails and maintaining a unified source of truth, agentic AI systems have reduced error rates to 2–3%, significantly lowering rework. The third component is redeployed labor value. When AI handles routine administrative work, organizations can redirect roughly 10% of monthly project-management capacity toward higher-leverage engineering and coordination tasks.

As AI evolved, firms must evolve with it. Today, Surfaice's core architecture is built around autonomous agents running on top of state-of-the-art models (including Google’s Agents Development Kit and Gemini family), but the design is model-agnostic by intent. Companies like OpenAI, Anthropic, and Google will keep making their models better. What if every new version of GPT, Claude, or Gemini can make the product better?

Uderbekov added, “No customer has ever asked, ‘Are you using Model X or Model Y?’ They asked this question: ‘Can your agent do this process better and faster than our current way?’ That’s the bar.”

Equinix has announced a major expansion of its Fabric Geo Zones service as the data center provider aims to shore up customer data sovereignty capabilities.

Fabric Geo Zones is designed to ensure that enterprise data remains within defined jurisdictions. According to the firm, unlike solutions built within a single cloud or delivered as software overlays, Fabric Geo Zones enforces sovereignty at the network layer.

"Sovereignty can't be a setting you configure inside a single cloud. Global enterprises must enforce sovereignty at the network layer, across every cloud, provider and path simultaneously," said Arun Dev, vice president of digital interconnection at Equinix.

"Equinix Fabric Geo Zones is the only solution that enforces geographic boundaries as a property of the network itself. Traffic either flows along compliant paths or it's blocked. That's why enterprises across industries trust Equinix to move data across clouds without compromising sovereignty."

Fabric Geo Zones is aimed primarily at workloads requiring a high level of compliance – financial institutions, healthcare organizations, government agencies, and the like.

Global companies, meanwhile, can automatically apply jurisdiction‑specific routing rules to meet GDPR, LGPD, APRA and other regional requirements across their operations.

Data sovereignty in the spotlight

Inadvertently moving sovereign data across borders that organizations are legally required to respect is a growing risk when working in hybrid multi-cloud environments, said Courtney Munroe, founder of Apex Research.

"Businesses are facing one of the most complex global regulatory environments in history while at the same time facing huge pressure to deploy new technologies," he said.

"A global enterprise operating under GDPR in Europe, LGPD in Brazil, and APRA in Australia simultaneously needs different data routing rules for each jurisdiction, with every outage, failover, or congestion event a potential compliance violation."

Recent research from Kiteworks found that one-in-three organizations has experienced a sovereignty-related incident in the past 12 months. One-in-eight of these incidents consisted of unauthorized cross-border transfers.

In Europe, 44% cited provider sovereignty guarantees as their top barrier to cloud adoption – the highest of any region – despite near-universal GDPR compliance.

“Organizations across every region we surveyed are spending millions on sovereignty compliance, scoring high on awareness, and still getting hit by breaches, unauthorised transfers, and government access requests,” said Dario Perfettibile, EMEA GM of GTM and customer operations at Kiteworks.

Fabric Geo Zones is available in preview in countries including Australia, Brazil, Canada, Japan, Switzerland, the UK, and the US, with European Union availability expected in June.

FOLLOW US ON SOCIAL MEDIA

Chinese manufacturer LincPlus started life producing budget-priced and aesthetically pleasing laptops and tablets, and moved its focus onto the NAS market a couple of years ago. Its first product in 2024 was the LincStation N1, which boldly claimed to be the world's first six-bay all SSD appliance, followed by the N2 the following year, which employed the same compact chassis and delivered a big performance boost.

In this review, we look at the LincStation E1 – a table-top NAS which teams up two LFF/SFF SATA hot-plug drive bays with a pair of M.2 NVMe SSD slots. Recently launched on the Kickstarter crowdfunding platform, its biggest claim to fame is its very low price, with the first 200 units going at a super-early-bird price of €129, followed by an early-bird price of €149 and then a regular price of €219, which pans out to around £190.

Perhaps the biggest change is in software; unlike the N1 and N2, which both employ Lime Technology's Unraid software, the E1 introduces LincPlus' very own LincOS operating system. Unraid is undoubtedly a powerful OS with plenty of features, but it isn't the easiest to configure, and LincOS aims to deliver a more user-friendly experience.

LincPlus LincStation E1 review: Design and hardware features

The E1 employs a solidly constructed plastic chassis with the two telescopic tool-free hot-swap drive carriers accessed by removing the magnetic upper cover. In the base, you'll find a non-magnetic cover with the two M.2 NVMe slots behind it and a single fan sitting directly on top of them.

It's all change in the CPU department as LincPlus moves away from Intel, with the E1 receiving a quad-core 2GHz Rockchip RK3568 SoC sporting an ARM Mali G52 GPU and a 1TOPs NPU (neural processing unit). Memory is the DDR4 variety with the 4GB soldered on the motherboard and not upgradeable.

There was some confusion when the appliance was first launched, as its KickStarter project page showed the network port as Gigabit, and LincPlus initially confirmed that the 2.5GbE port in our early pre-launch model had been dropped down to Gigabit on production models. We were subsequently advised that this was incorrect and that production appliances will have a 2.5GbE port.

There are plenty more hardware features, along with integrated dual-band Wi-Fi support, you have a USB-A 3.2 Gen 2 USB port at the front and two more USB-A 2.0 ports at the rear. The HDMI 2.1 port alongside receives 4K 60Hz video from the SoC's GPU.

LincPlus LincStation E1 review: Swift deployment

For testing, we fitted two 4TB Seagate IronWolf NAS HDDs, added a couple of Kingston M.2 NVMe SSDs, and made sure the NVMe bay cover had been pushed forward until it was locked in place. The appliance was then hooked up to the lab's Zyxel XMG1930-30HP multi-Gig PoE++ switch, which confirmed a 2.5GbE connection speed.

Multiple deployment options are available, as you can attach an HD monitor, keyboard, and mouse to the appliance and use the desktop user interface. The easiest method is to install the LincStation Windows, iOS, or Android client apps, which run a wizard for registering the appliance and setting up an admin account.

To create storage spaces, you select the Profile/My Device/Storage Management option from the client and choose your drives with RAID0, 1, 10, 5, and 6 arrays supported. We created separate RAID1 mirrored storage spaces for the HDDs and NVMe SSDs. The appliance doesn't support SSD caching in the traditional sense, but you can use NVMe SSDs as a fast repository for file sharing and downloads.

LincPlus LincStation E1 review: LincOS features

LincOS is easy to use, and we found all features can be accessed directly from the client apps, which open with quick access links to personal storage spaces and a collection of apps for file sharing, multimedia services, remote downloads, and more. The client Profile menu offers a desktop mode, which presents the same graphical interface as a locally attached monitor.

The admin user can load a basic web UI from the client with access to personal spaces for viewing, creating, and sharing folders, monitoring upload tasks, loading a recycle bin, creating encrypted spaces, and managing share spaces. New users are created from the client, and they can log in from their choice of client and use the web UI for accessing their personal space, task monitoring, and a recycle bin.

The Backup Disk tool secures PC and mobile data to the appliance, where you choose local folders, an appliance storage space, the file types to be backed up, and whether to use WiFi only or any network. The Sync Disk tool keeps selected folders on the local device and NAS in step, and in both cases, the processes are automated with an option to run jobs on demand as well.

The Secure Space feature allows users to create hardware-encrypted, password-protected areas in storage spaces for private data. Folders in this space cannot be shared, and the client provides an option for admins to back up encryption passwords for all user accounts for disaster recovery purposes.

Multimedia services are extensive as LincOS has music and video centre apps, a smart album for photos, and Photo Frame for creating slideshows. The smart album's local AI-powered photo image and facial recognition work well, while the smart library AI document centre provides full text search facilities for all file content.

LincPlus LincStation E1 review: Gigabit performance

With a share on the appliance's HDD RAID1 storage space mapped to a Dell PowerEdge Windows Server 2025 host on our 10GbE LAN, Iometer recorded uninspiring sequential read and write rates of 216MB/sec and 131MB/sec with the Windows client app showing CPU utilisation of up to 80%.

Real world numbers also failed to impress with 25GB file copies between the NAS and server delivering average read and write rates 170MB/sec and 112MB/sec while our 22.4GB backup test folder with 10,500 small files was secured using drag and drop at 66MB/sec.

Moving to a share on the RAID1 NVMe SSD mirror saw minor performance improvements with Iometer recording sequential read and write rates of 225MB/sec and 139MB/sec. Read and write speeds for our large file copies stepped up slightly to 174MB/sec and 117MB/sec while our backup test mustered 71MB/sec.

LincPlus LincStation E1 review: Is it worth it?

The KickStarter campaign ends at the start of June, and with a projected regular price of around £190, the LincStation E1 is probably the lowest-cost 4-bay NAS on the market. Performance over 2.5GbE is below expectations, but it should satisfy the target market of home and small business users.

A standout feature is the new LincOS software, as even though it's still in development, it already offers a good range of storage features with the smart LincStation clients making light work of configuration and management. LincPlus advised us that the next LincOS update will add an improved web GUI, a new task manager, drag and drop support for the cloud drive module, a redesigned music module, and the LincClaw AI assistant.

LincPlus LincStation E1 specifications

The road to hell is paved with good intentions and, it turns out, the road to getting approval for an AI data center can prove hellish. With an ever growing need for more compute, as well as growing grid constraints and distrust over their climate impact, in the name of generative AI, and a construction industry already under tremendous strain. Just how data centers get built is just as interesting as why they’re being built in the first place.

Rob Laudati is chief product and partnership officer at Render Networks, a critical infrastructure specialist. He tells ITPro that laying out the typical process for getting approval for a data center, from initial consultations, to permitting, to construction and delivery is like “painting a fairly ugly picture”. According to him, the bottleneck tends to be in the permitting approval process that must go on before the lion’s share of the work gets done.

"It's turned out to be the long pole in the tent in almost all of these projects, because those permits will go through various jurisdictions of local, county, state, [and] private landowners,” he explains. “Typically the land is never purchased, but there are easements created and permits approved, and that is an enormously difficult process, highly variable.”

That permitting isn’t just for land use. It can be for zoning, but it can also require approvals based on power, water, pollutants, and other environmental factors – like whether building the project will disrupt the patterns of migratory birds or local air pollution. Then there are logistical headaches: who is going to build it? When is it going to be built by? Will the public be happy? When it comes to permitting, sources who spoke to ITPro said that while there are exceptions, these processes tend to last months rather than days or weeks.

Data center emissions standards

Kaitlyn Bencosme is senior managing consultant in the air and climate group at the engineering and consultancy firm Ramboll. She’s one of the people who helps steward clients through the permitting process, which in her case means making sure the permits surrounding air and climate are squared away. For example, do the gas turbines you may be using meet the mandated emissions standards under local, state, and federal laws and does what you’re permitting match your use case?

“Step one is figuring out what air emissions are happening so your combustion equipment, any kind of diesel storage, that could have any off gassing, that sort of thing, and then you have to permit those activities with the local authority having jurisdiction (AHJ),” Bencosme explains.

Without air permits, construction can be severely limited, depending on where the data center is being built. In some areas of the US, you can still pour foundations or grade the area in this phase. However, Bencosme says there is one fairly clear rule:

“Every state kind of has their own variation on what they will allow you to construct. A pretty hard line is that you can't turn on any emissions equipment prior to having your permit in hand.”

Bencosme adds that when companies set out minimum uptime requirements, such as five nines, it can necessitate a large amount of emission generating equipment. However, often the permitting environment or other constraints won’t allow for it.

“They want to put in this pie in the sky image of a huge AI data center, and then I have to bring them back to reality where I'm like, ‘Well, in this specific part, because of the rules that you have to comply with, you'll only be able to do X.”

Another problem, what Bencosme calls a fatal flaw, is when a company is looking for a permit that requires the purchasing of emissions offsets but there aren’t any available in a given region.

“In the Dallas Fort Worth area, for example, it's nearly impossible to find any credits. So facilities in that area, even if they wanted to exceed the major source threshold which right now is 25 tons per year. They feasibly could not do it… Everything else is fluid just based on facility, goals, and timelines. And it's usually timelines.”

The right labor for building data centers

Emissions standards aren’t the only barrier. Finding the right labor for the job can also be a hurdle. Tony Qorri is vice president of construction at Data Bank. He says that while his focus is always on hiring talent in the local market, sometimes those people aren’t available.

“You might have a national general contractor that has a presence in multiple markets, but then when you get to the subcontractor perspective, you look at places like Texas, Georgia, Northern Virginia, the tradesmen are very hard to secure.”

Qorri says that these issues, whether they’re from a regulator or a workforce standpoint, can be complicated when it comes to projects in rural areas where regulatory meetings may happen less frequently and the labor force just isn’t available.

And then there’s the large elephant in the room: a growing discontent in the communities where data centers are being built with concerns over water consumption, power usage, and emissions all being front of mind. Laudati sees part of the issue as being rooted in a lack of coordination that leads to repeated disruptions to people’s everyday lives.

“I think it is a maturity issue. That people are running and trying to deliver these systems as soon as possible, and they're not taking a moment to really think…’How could I actually save time if I did these projects in concert with each other?”

Qorri points to the need for continued education and for more companies to see their primary role as being a good steward while completing their project:

“How do we continue to help educate the folks that really don't know what it is that we do or have this general idea of a data center?”

Changes are underway in the market. More advanced liquid-cooled infrastructure, such as closed-loop systems, direct-to-chip cooling, and immersion cooling. In August 2025, TechUK found the average data center in England uses less water than the "typical leisure center".

But data center operators will have to balance efforts to disrupt data center designs for sustainability with construction targets. The road ahead for all involved requires long-term planning, a per-region strategy, and flexibility in the face of a highly dynamic market.

Amazon Web Services (AWS) has confirmed a recent outage that impacted customers was caused by overheating at a North Virginia data center.

The disruption affected one of AWS' six Availability Zones, use1-az4 in the AWS US-EAST-1 region. This is one of the company's most heavily used regions globally.

Notably, the incident hit platforms including cryptocurrency exchange Coinbase, disrupting core exchange functions for more than five hours. Other reported victims include the CME Group trading platform and major gambling company FanDuel.

Coinbase last night warned that some users might experience delayed sends and receives on the Solana network and for ALEO, but said it was working on the issue.

The crypto trading platform has since resumed operations. In a statement, Coinbase said: "All markets have been re-enabled for trading on coinbase.com and in the Coinbase iOS and Android apps. Coinbase customers can log in to trade."

FanDuel, meanwhile, posted a statement on X, saying it was working to troubleshoot the issue. "Our team is aware and investigating the current technical difficulties prohibiting users from accessing our platform," it said.

Overheating behind AWS outage

In an update to customers, AWS attributed the cause of the outage to overheating. The hyperscaler is yet to confirm how the overheating occurred.

"We have experienced an increase in temperatures within a single data center, which in some cases has caused impairments for instances in the Availability Zone," AWS said in a status report.

"EC2 instances and EBS volumes hosted on impacted hardware are affected by the loss of power during the thermal event."

In its latest update, AWS said it had shifted traffic away from the impacted zone. The hyperscaler said it was still carrying out mitigation efforts.

These are taking longer than expected to bring additional cooling system capacity online and recover the remaining affected infrastructure safely and in a controlled manner.

AWS warned some customers will continue to see their affected EC2 instances and EBS volumes as impaired until it can achieve full recovery. It said it currently didn't have an ETA for this.

What services are back online?

A number of services are back online following the outage, according to the hyperscaler. This includes:

• AWS IoT Core
• AWS NAT Gateway
• Amazon Elastic Kubernetes Service
• Amazon Elastic Load Balancing
• Amazon Redshift

Some services are still impacted at time of writing, including:

• Amazon ElastiCache
• Amazon Managed Streaming for Apache Kafka
• Amazon OpenSearch Service
• Amazon SageMaker

Yet another AWS outage

It's not the first time that a major AWS outage has caused chaos. Last year, hundreds of apps and websites including Slack, Zoom, Coinbase, Snapchat, and Signal were taken down in a global outage.

Banking applications including Lloyds and Halifax also saw customers unable to access services. On that occasion, AWS attributed the outage to a DNS issue.

The incident highlights the extent to which major websites and apps are dependent on just a few tech giants.

In 2024, for example, issues with CrowdStrike saw hospitals, banks, and airports in Australia, New Zealand, India, Japan, the US, Germany, and the UK seriously affected.

FOLLOW US ON SOCIAL MEDIA

Argyll Data Development has launched a sovereign AI inference cloud, built in partnership with AI infrastructure firm SambaNova.

The UK-owned facility will be based on SambaNova’s AI hardware and software stack, and will keep all data, models, and operations fully under local control.

According to the duo, inferencing requires significantly less power than training, reducing costs for developers and avoiding energy waste, while minimizing latency for time-sensitive sectors like healthcare and banking.

Argyll said it expects early adoption in life sciences, medicine, and banking, thanks to their rapid AI integration. Other applications are likely to include logistics and customs management, especially given post-Brexit paperwork demands, along with education and agriculture.

Rather than using standard GPU-based systems, the platform is based on SambaNova’s Reconfigurable Data Unit (RDU) architecture, running its SambaManaged solution.

Argyll said this gives customers high-performance AI inference with lower power and cooling requirements, operating within a power envelope of approximately 10kW per rack.

"As organizations scale AI, many are defaulting to GPU infrastructure without fully accounting for long-term cost, energy and operational complexity," said Jude Sheeran, managing director, EMEA at SambaNova.

"Our work with Argyll provides an alternative, enabling high-performance AI inference that is more efficient, deployable and aligned with sovereignty requirements."

The facility, powered by 100% renewable Scottish energy, is housed in the Killellan AI Growth Zone, a 184-acre digital campus in Scotland on the Cowal Peninsula, and is already live.

Operating within a power envelope of around 10kW per rack, SambaNova said its technology allows sovereign AI infrastructure to be deployed within existing UK data centers.

Argyll’s inference cloud has been designed as a disaggregated platform, allowing compute, storage, and networking to be deployed across multiple UK locations, while operating as a single, unified inference layer.

This, Argyll said, increases resilience and flexibility for organizations operating in regulated or security-sensitive sectors.

Customers will be able to run AI applications in real time - from customer operations to fraud detection - at low cost, the duo claimed. The inference cloud hosts open source models such as Minimax, running at speeds of up to 400 tokens per second, delivered entirely within a UK-resident environment.

"Sovereignty in AI is not a label you can apply to a contract or a colocation agreement," said Peter Griffiths, chairman at Argyll Data Development.

"It is a condition that has to be demonstrated - who is accountable, where the infrastructure sits, who controls the intelligence layer, and whether all of that aligns with the expectations of the society being served. Our platform satisfies those conditions. We are building the standard that others should be measured against."

FOLLOW US ON SOCIAL MEDIA

Organizations are reluctant to invest in quantum computing due to an apparent lack of clear cut use-cases, according to new research.

Analysis from QuEra Computing suggests that the market is shifting from “hype-driven investment” to a “proof-driven discipline”, with organizations now demanding clear evidence of value before committing to investment.

This market shift is showcased by current spending habits, the study noted. While nearly half (44%) of enterprises expect quantum budgets to increase in the year ahead, a larger number (46%) expect them to remain flat.

Elsewhere, 10% also anticipate budget decreases as hype falters.

"Buyers want proof, not glossy brochures," said Yuval Boger, chief commercial officer at QuEra Computing.

"Organizations are moving from early experimentation to disciplined investment decisions, where budgets are scrutinized, use cases must be justified, and procurement plays a central role."

A quantum "reality check"

A contributing factor to the budget tightening lies in optimism among leadership figures, the study noted, pointing to a growing divide between executives and technical staff.

Senior decision-makers, for example, are far less optimistic about the near-term prospects of quantum computing and are instead adopting a more cautious, cost-conscious approach.

QuEra described this trend as a “C-suite reality check” and noted that “enthusiasm alone is no longer sufficient to unlock funding”.

Simply put, businesses need to have a clear understanding of the potential value of quantum computing, as well as a clearer route to ROI.

The current state of affairs in quantum bears similarities to that in the generative AI market over the last three years, with C-suite leaders growing increasingly concerned about returns on investment.

"With nearly 100 quantum companies competing today, executives are asking two questions before they commit," Boger commented.

"Who has the funding to be here for the long haul, and who has clear scientific proof that their approach works and a credible path to larger machines that deliver real enterprise value?"

Quantum FOMO isn’t enough to convince execs

According to QuEra, earlier investment cycles in quantum were largely driven by a “fear of missing out”, with many organizations getting caught up in the hype cycle.

This trend has dissipated over the last year, however. Just 9% of respondents cited successful pilot results as their primary driver of increased spending in this domain.

Indeed, the main drivers of investment now vary wildly, the study noted. Early-stage organizations still cite competitive pressure and FOMO as leading motivators.

Their more advanced counterparts, meanwhile, typically point to the “classical wall” - the point where conventional computing cannot handle certain workloads - as a leading driver of investment.

Public sector investment also plays a key role in decision-making when it comes to quantum computing, according to QuEra.

Government-related investment schemes were cited as a key driver of budget increases by 28% of respondents, more than any other factor, the company noted.

"Quantum computing is still pre-commercial, and public funding is underwriting the risk that private capital is not yet willing to take on alone,” the company said.

FOLLOW US ON SOCIAL MEDIA

Data center operator Stellanor has dramatically boosted its UK presence by taking over eight data centers from UK managed services provider Redcentric in a £123 million deal.

The new facilities, serving around 450 enterprise customers, mean the company now has eleven data centers across the UK, with a total of 39MVA of secured grid capacity.

The firm said the deal will enable it to deliver high-density colocation capabilities, along with advanced cooling systems, robust security, and sustainable operations powered through renewable energy procurement across its network.

“This acquisition represents a transformative step in building the UK’s leading urban data center platform,” said Michael Tobin CBE, Chairman of the Board, Stellanor.

“We’ve scaled from two to eleven facilities in nine months, backed by the infrastructure fund managed by DWS Group. This enables Stellanor to serve the accelerating demand for enterprise colocation and AI-ready infrastructure across the UK, with further expansion into Ireland and the Nordic regions planned."

With the deal complete, Stellanor is now carrying out a series of coordinated infrastructure upgrades across the network, including high-density power capabilities, advanced cooling systems, and enhanced fiber interconnection.

This, the company said, will enable the platform to support AI inference, machine learning and real-time analytics workloads, without any service disruption for existing clients.

Redcentric, meanwhile, said the divestment will allow the company to solely focus on its core managed services business, strengthening the group's balance sheet by significantly reducing debt.

"As a dedicated UK MSP with vast experience in supporting organisations in highly regulated markets, we’re perfectly positioned to use this expertise to help business navigate an increasingly complex technology landscape,” said CEO Michelle Senecal De Fonseca.

“We’re confident our data center clients will benefit from Stellanor’s platform approach and DWS backing. We look forward to continuing to develop our partnership together and building on the strong ties we have built through this sales process.”

Stellanor has grown rapidly since its formation a year ago, backed by German investment firm DWS with the acquisition of two data centers in London from Colt Technology Services.

It acquired its Hemel Hempstead site, supporting GPU and AI chip workloads, from Imagination Technologies just last month in a sale-and-service-back deal that sees Imagination continuing to operate from the site as a fully managed client of Stellanor.

FOLLOW US ON SOCIAL MEDIA

HPE has announced an expansion to its ProLiant edge portfolio, with a particular focus placed on supporting organizations operating in “harsh environments”.

The new HPE ProLiant Compute EL2000 chassis, which supports up to two HPE ProLiant Compute EL220 Gen12 servers or one EL240 Gen12 server, is designed specifically for applications in manufacturing, retail, telecommunications, and national security.

According to HPE, the new servers boast a range of performance improvements compared to prior generations, and can scale from eight up to 144 Intel Xeon 6 cores.

The EL240 Gen12 server is also available with Nvidia RTX Pro 4500 or Nvidia RTX Pro 6000 Blackwell server edition GPUs.

Durability is a key focus here, according to HPE. The new server range can operate in extreme temperatures ranging from -40 degrees Celsius to 55 degrees Celsius, as well as environments with up to 95% humidity.

They’re also designed to compensate for heavy vibration from aircraft or ground vehicles, as well as electromagnetic interference (EMI) and environmental contaminants.

During a press briefing ahead of the announcement, Krista Satterthwaite, senior vice president and GM of HPE’s compute group, said the focus on durability and ruggedness comes in direct response to changing edge requirements.

“Every edge is different, and edge is hard,” she said. “Servers at the edge need to be tough across industries like retail, manufacturing, [and] telecommunications.”

“Workloads are running at edge locations that are understaffed, constrained by space power, connectivity, and security becomes even more critical because these environments are much more exposed, which means protection has to be built in from the start.”

HPE touts telecoms, defense potential

HPE revealed that each platform is now available with an “Environmental Ruggedization” kit, enabling organizations operating in harsh conditions to compensate for “extreme stressors”.

“We're introducing an environmental ruggedized kit that enables these servers to handle extreme stressors,” Satterthwaite told assembled media.

“These new offerings satisfy customer needs for purpose built systems that deliver highly reliable performance in harsh and highly distributed locations that are designed to operate remotely with end to end visibility and control.

According to HPE, the ruggedization features adhere to a range of standards, including US national security standards, which require validated server survivability capabilities in extreme temperatures or high or low altitudes.

Satterthwaite noted that growing demands for edge capabilities in the defense industry and the public sector have played a key role in the durability focus.

“That is one of the main focus areas, which is why we have so many things that we’ve pursued in terms of certifications for more ruggedized environments,” she commented.

“So big, big demand coming from the defense space, in particular in the public sector.”

John Carter, VP for mainstream compute at HPE, echoed Satterthwaite’s comments, noting that the systems will “go anywhere” and “withstand any environment”.

“We’ve seen demand and had customer conversations from everything from space stations to battleships, to Jeeps, to anything else you can think of,” he said.

Elsewhere, the new range also adheres to telecoms network equipment standards that support 5G and radio access network (RAN) availability in unattended environments.

On the telecommunications front, Carter told assembled media this is an industry where the company is seeing surging demand for edge capabilities.

“Telcos have really been embracing the edge for quite a while,” he said. “They’ve been expanding the 5G roll-outs, now they’re starting to look at 6G in the future, and that is requiring a new level of performance and compute.”

“But we have seen numerous different environments from different customers. It just depends on the provider. And we do work with those telco providers from everything from the extreme edge all the way back to the packet core,” Carter added.

FOLLOW US ON SOCIAL MEDIA

Ugreen is rapidly gaining a strong reputation in the NAS market with its ever-expanding range of desktop appliances. And the latest DXP4800 Pro focuses on demanding home and small office users who want plenty of power and performance for their money. Stepping up as Ugreen's flagship 4-bay appliance, it takes everything we liked about the two-year-old DXP4800 Plus and gives it a performance boost.

The Gen12 5-core Intel Pentium Gold 8505 CPU in the Plus model gets replaced with a more modern Gen13 6-core Core i3-1315U offering four efficient plus two multi-thread performance cores and a top Turbo frequency of 4.5GHz. Memory stays the same at DDR5, but those with deep pockets can upgrade the base 8GB to 96GB – the maximum supported by this CPU.

Along with four front drive bays, you have a pair of M.2 NVMe SSD slots lurking in the base, and it's worth noting these are the faster PCIe Gen4 variety. Network port choices remain the same, although these are superior to a lot of the competition as the appliance sports 2.5GbE and 10GbE multi-Gigabit ports.

Ugreen NASync DXP4800 Pro review: Build quality and hardware features

No cheap plastic chassis here as the DXP4800 Pro is clothed in a sleek and sturdy aluminium alloy shell. SATA hard disks slip easily into the clever telescopic tool-free carriers, which also have mounting holes for SFF SATA SSDs.

The two M.2 slots are easy to access as the cover is spring-loaded, and you'll find two SO-DIMM slots next to them with the base 8GB of memory provided on one stick. General cooling is handled efficiently by a big rear 14 cm diameter fan behind a magnetic dust cover, and we found the appliance to be pleasingly quiet.

The appliance offers plenty of other interfaces with 10Gbps USB-C 3.2 and USB-A 3.2 at the front next to an SD Card reader slot. A peek around the back reveals one 5Gbps USB-A 3.2 and two USB-A 2.0 ports, while 4K 60Hz video from the CPU's embedded Intel UHD Graphics GPU is piped through to an HDMI 2.0 port.

We don't often comment on packaging, but Ugreen gets praise for all the extra bits and prices it includes in the box. The external PSU comes with a UK plug lead, and the accessory box contains two Ethernet cables, a screwdriver, and a full set of screws for SFF drives, two carrier locking keys, two adhesive silicon heat pads for M.2 devices, and even a Ugreen-branded HDMI cable.

Ugreen NASync DXP4800 Pro review: Deployment and UGOS apps

For testing, we fitted four 4TB Seagate IronWolf NAS HDDs, which were accepted without any problems, as unlike Synology, Ugreen doesn't have any drive restriction policies. Installation is a breeze as Ugreen's discovery web portal found the appliance on our network, initialised it, and downloaded the latest UGOS software.

The UGOS web console is nicely designed and aids storage pool creation by offering sage advice on the various RAID options. We chose the recommended RAID5 volume and were prompted to create personal and shared folders and configure remote access, which creates a unique Ugreenlink ID and secure web link.

During volume creation, you can choose from the Ext4 and Btrfs file systems, with the latter the preferred choice if you want snapshots, storage space quotas, and higher levels of data integrity protection. We also noted that the Samba file sharing service is disabled by default, which we think is a good security measure.

UGOS is light on apps as it only offers 29 for the DXP4800 Pro – well short of the huge range offered by Qnap and Synology. Data protection features start with the Snapshot app, which supports manual and scheduled snapshots on user and shared network folders stored in Btrfs volumes, but Ugreen has yet to update this to support iSCSI LUNs.

The Sync & Backup app secures local data to a remote Ugreen NAS and vice versa, and also supports Rsync-compliant servers. The app has a download link for Ugreen's NAS Windows and macOS desktop tool for creating sync tasks that secure selected data to a shared folder and update it in real time, while the Cloud Drives app provides file syncing for seven cloud storage services, including OneDrive and Google Drive.

The Security Manager app provides real-time and scheduled anti-malware protection with quarantining. Virtualization is on the menu with the Docker app running lightweight containers and the Virtual Machine app providing full hypervisor services.

Ugreen NASync DXP4800 Pro review: 10GbE performance

For performance testing, we hooked the NAS up to the lab's 10GbE network and mapped a share to a Dell PowerEdge Windows Server 2025 host. UGOS now supports 9000-byte Jumbo frames, and with these enabled, we saw good NAS speeds with Iometer reporting sequential read and write rates of 9.2Gbits/sec and 8.1Gbits/sec

For our real-world tests, copies of a large 25GB test file between the appliance and server delivered sustained read and write rates of 3.2Gbit/sec and 2.7Gbits/sec while backing up a 22.4GB folder with 10,500 small files averaged a reasonable 1.5Gbits/sec. Ugreen added the SAN Manager app a while ago, and moving to IP SANs saw a 500GB iSCSI target deliver Iometer read and write rates of 9.2Gbits/sec and 7.5Gbits/sec.

For cache testing, we fitted two 480GB Kingston M.2 NVMe SSDs and assigned them to our data volume as a mirrored read/write cache. This made noticeable improvements to random write operations with our NAS share and iSCSI LUN, seeing respective speed increases of 123% and 120%.

Ugreen NASync DXP4800 Pro review: Is it worth it?

With Ugreen currently offering the NASync DXP4800 Pro for £620 including VAT, it compares very well on value with the competition. Synology's DS925+, for example, costs a shade under £600 but only supports DDR4 memory, has dual 2.5GbE ports, is powered by a pensionable AMD Ryzen 1500B CPU, and doesn't offer a 10GbE upgrade option.

Ugreen's build quality is also vastly superior, but there's a lot of work to be done with the UGOS software before it can match the incredible range of apps, services, security, and data protection features offered by Synology's DSM and Qnap's QuTS Hero. The DXP4800 does deliver an impressive hardware package for the price, though, and if you can cope with the basic range of apps, is a great choice for home and office users seeking a powerful NAS appliance with great 10GbE performance.

Ugreen NASync DXP4800 Pro specifications

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

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.”

Fast-growing businesses seeking a tower server that can keep up with them will find HPE's ProLiant Compute ML350 Gen12 a worthy partner. Stepping up as HPE's most powerful tower yet, it supports dual Xeon 6 P-Core CPUs up to 86 cores plus a massive 8TB of DDR5 memory and presents a wealth of storage permutations.

The ML350 Gen12 looks capable of turning its hand to a multitude of tasks as its modular chassis offers an expansion potential that defies belief. Along with SMBs, HPE is targeting this server at a range of other scenarios, including ROBOs, edge deployments, and retail outlets, along with enterprise workloads such as IT infrastructure, data management, VDI, and ERP/CRM.

There's more, as it can handle up to eight single-width or four double-width Nvidia GPUs. These attributes make this server a fine choice for demanding AI, machine learning (ML), and database analytics tasks.

HPE ProLiant Compute ML350 Gen12 review: Design and expansion

Make no mistake, this server is big with the chassis measuring 46.2cms high, 71.2cms deep and 17.4cms wide, which equates to an internal volume of just over 57 litres. Build quality is impressively sturdy, with even the base model tipping the scales at a hefty 24kgs. HPE also offers a conversion kit that transforms it into a 4U rack server, which occupies 5U of rack height.

Slipping the lockable side panel off shows HPE had made excellent use of the server's voluminous interior. Behind the front storage bays is a full-height cooling fan array with base systems coming with three NHP (non-hot-pluggable) modules included.

The assembly has room for up to eight fans and the dual 2.5GHz Xeon 6730P CPUs in our system, and their 250W TDPs dictate that we need a full house of hot-plug modules. CPUs over 225W also require the high-performance heatsink kits, which extend large radiators into the fan array for maximum cooling and work extremely well, as our system was whisper-quiet.

Expansion potential is impressive as further back are two riser cages, each with up to four full-height, full-length PCIe Gen5 x16 slots. The quad-slot riser has on-board connectors providing backup power for HPE's storage controllers, while the dual-slot risers also have GPU sideband and auxiliary power connectors.

HPE's tool-free design comes into play as the two main riser cages can be easily released with just a flick of the wrist. The motherboard has two edge connectors for PCIe Gen5 OCP cards, and the standard PCIe slot count can be pushed to ten, as there's even space for a tertiary dual-slot riser in front of the upper PSU bays.

HPE ProLiant Compute ML350 Gen12 review: Storage features

As you'd expect, there's plenty of room at the front for storage, and the smart triple box design makes the server very flexible. Depending on your choice of backplane, each box supports four LFF SAS/SATA, eight SFF SAS/SATA/NVMe, or twelve EDSFF E3.S NVMe storage devices, and you can mix and match them as required.

The motherboard has Intel's Virtual RAID on CPU (VROC) for HPE ProLiant Servers Gen12 embedded, which is specifically for NVMe SSDs with direct to CPU connections and supports software-managed stripes, mirrors, and RAID5 arrays. For all other storage devices, there are HPE's Gen11 hardware RAID controllers with a choice of three OCP and three PCIe versions – we have the MR408i-o OCP card, which supports all the usual RAID array suspects and comes with 4GB of Flash-backed cache.

We also have HPE's NS204i-u V2 boot optimised storage device accessible at the rear and fitted with dual 480GB mirrored M.2 NVMe SSDs. If you want it hidden away from wandering fingers, you can have it front-mounted so it's behind the lockable bezel.

HPE ProLiant Compute ML350 Gen12 review: Remote management

The server comes with HPE's excellent iLO7 remote management controller, although, unlike most of the Gen12 rack servers, this is not implemented as an OCP-compliant DC-SCM (data center secure control module) board but embedded on the motherboard. No matter, as it provides all the same top-notch security features.

Platform security and supply chain integrity are assured, and the iLO7 silicon root of trust (RoT) digital firmware fingerprinting is future-proofed against next-generation quantum computing threats. It also includes HPE's patented Secure Enclave – a physically tamper-resistant security processor for storing sensitive data such as encryption keys, passwords, and security configurations.

The iLO7 web interface is designed around workflows, with the customisable home page dashboard presenting up to six cards for views of key areas such as host health and security issues. Colour-coded icons are provided on each card for at-a-glance status views, and you drill down into each one for more detailed information.

We run HPE's OneView in the lab on a Hyper-V host and had no problems importing the ML350 Gen12. The free OneView Standard licence supports server inventory and monitoring, with the Advanced licence offering OS deployment tools plus server, firmware, and power management services.

Enterprises will love HPE's GreenLake Compute Ops Management (COM) cloud service as it connects directly to the iLO7 controller and, unlike Dell's AIOps, doesn't require an intermediary app running on the local network. COM presents a complete overview of your server estate and provides tools for server deployment and management, automated system lifecycle compliance, and AI-powered insights.

HPE ProLiant Compute ML350 Gen12 review: Is it worth it?

Yes, very much so. We can see from HPE's One Config Simple portal that prices for base systems start at around £6,800, which gets you a 12-core Xeon 6505P CPU, 32GB of DDR5 Smart Memory, a four LFF drive box with two 8TB SATA drives, an MR408i-o OCP RAID card, and dual 800W PSUs.

The ProLiant Compute ML350 Gen12 offers a lot of tower server for your money and delivers an expansion potential that makes it incredibly versatile. Clearly capable of handling a diverse range of workloads, it's an ideal choice for businesses that want a high expansion and plenty of growth potential.

HPE ProLiant Compute ML350 Gen12 specifications

Energy costs are already holding back AI, with one-in-five UK firms moving their AI operations overseas to dodge prices.

That's according to a survey from CUDO Compute, which found 20% of British companies polled had moved some workloads out of the country over rising energy costs.

Data center energy costs have long been a concern when it comes to the AI, in part due to aging grid infrastructure constraining capacity for a technology that has heavy power demands.

Last month, Google's president said the US needed to step up power generation to meet skyrocketing demands.

Energy costs in the UK specifically have become a key talking point amid infrastructure expansion plans. Earlier this month, OpenAI cited energy costs as a key factor behind its decision to pause the Stargate UK infrastructure scheme.

CUDO Compute's survey of 700 AI decision makers found a third of respondents said energy costs are limiting their ability to scale AI operations, with 42% adding that high energy and infrastructure costs are fuelling concerns around an AI bubble.

"AI is not abstract software. It is physical infrastructure that depends on power, land, cooling and grid access," said Matt Hawkins, CEO of CUDO Compute.

"When those constraints tighten, economics take over. If it is cheaper or easier to run workloads elsewhere, they will move, regardless of sovereignty ambitions."

The report notes that infrastructure is the key to enabling AI growth, with access to land, affordable energy, and scalable compute all potentially constraining growth.

"Right now, every UK boardroom is talking about AI, but almost nobody is talking about the infrastructure needed to power it," said Hawkins.

"Until we close that gap, there will continue to be a disconnect between policy, ambition and reality. The countries that solve this first will shape the future of AI, and the UK still has a window to lead, but it needs to act quickly."

The report predicts $600 billion in capex spend by hyperscalers this year, up from $400bn last year. But as data centers are quickly built to enable AI, campaigners and local residents continue to push back against data centres in their neighbourhoods, criticizing pollution, emissions and water use.

Sovereign AI ambitions

Energy costs aren't the only reason companies are thinking about shifting their AI workloads: AI sovereignty and data center sovereignty are also major concerns.

The European Commission yesterday backed €180 million to procure sovereign cloud services for EU institutions for the next six years, amid a wider push for digital sovereignty.

In the CUDO Compute survey, a third of respondents said they were "actively considering" moving workloads over geopolitical pressures, with 45% of those in the UK saying data or digital sovereignty, regulatory compliance, and national security were driving decisions on deployment strategy, higher than 33% in Europe.

One-in-three Brits surveyed say they are prioritizing regionally controlled compute — even if it costs the company more.

"AI sovereignty is being hotly discussed as a priority for UK organisations, but it only works if the infrastructure exists to support it," said Hawkins. "What we are seeing is a growing tension between where businesses want to run AI and where they actually can."

FOLLOW US ON SOCIAL MEDIA

Cisco is rolling out its Sovereign Critical Infrastructure (SCI) portfolio across the EMEA region, allowing customers to build and operate their own sovereign infrastructure.

The announcement covers the company's core product lines, including networking, security, compute, collaboration, network management, AI, and Splunk.

With the move, customers can now configure and operate the products in their own air-gapped, on-premises physical environments.

"At Cisco, we understand that true sovereignty means having the freedom to innovate with choice and control remaining crucial to businesses in Europe, the Middle East and Africa,” said Gordon Thomson, President Cisco EMEA.

"Our Sovereign Critical Infrastructure portfolio meets organizations where they are on their sovereignty journey. The availability of this portfolio reinforces our commitment to being a trusted technology provider for the region’s critical infrastructure."

What to expect with Cisco SCI

Cisco said customers will have control over their data, via on-premises and air-gapped deployments where they retain full authority over location and access to data.

They get operational autonomy, with product capabilities and licensing models that allow products to function without internet connectivity or remote intervention.

A key appeal, according to Cisco, lies in freedom from dependency through continued legal use rights even in extraordinary disruption circumstances.

"A key strength in delivering on our promise of sovereignty is Cisco’s strong partner ecosystem," said Thomson.

"Together, we offer unmatched breadth and choice. Whether customers require a fully sovereign environment or the ability to deploy on-prem, air-gapped Sovereign Critical Infrastructure alongside cloud-based services in a hybrid environment, we provide the choice and control they are asking for."

Meanwhile, the Cisco Customer Experience (CX) organization now offers support and services with air-gapped, on-premises, or hybrid setups.

As part of this, Cisco CX is expanding its Critical National Services Centers (CNSCs) across EMEA - including UK, France, Spain, and Italy – along the lines of its existing center in Germany.

These centers will provide support reflecting strict data sovereignty strategies, including dedicated facilities, controlled access, and cleared personnel.

The company said its on-premises portfolio conforms to IPv6 Ready standards and Common Criteria certification.

Cisco is also working towards achieving the new European Union Cybersecurity Certification (EUCC) across its product lines, already in place for its Cisco Nexus 9K Series family of high-performance data center switches, as well as Intersight Appliance on UCS.

Cisco doubling down on sovereignty

Just last week, Cisco broadened its offer of Webex data residency to the UK, after having rolled it out across the EU five years ago.

Gartner predicts that spending on sovereign cloud infrastructure in Europe will triple from 2025 to 2027, with around a fifth of existing workloads set to shift from global cloud providers to local ones.

"As we navigate an era of digital transformation and AI adoption at a speed that we have never experienced before, the ability to maintain control and autonomy over the most critical data and digital infrastructure has become a strategic requirement for organizations managing complex digital environments," said Thomson.

"Our role is not to define sovereignty for our customers, but to listen to their needs and help them have control over their infrastructure and data. This is our long-term commitment to the region."

FOLLOW US ON SOCIAL MEDIA