Google Cloud Next 2026: all the live updates as they happen

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

Vahdat is here to talk about Google Cloud's AI Hypercomputer, its end-to-end energy, software, and infrastructure stack that powers its data centers.

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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