Microsoft News

As we move into the era of Super Intelligence, Microsoft is entering a new phase of infrastructure. We will scale our Azure Maia family to multi-gigawatt deployments starting in 2027, making Maia the primary inference system for Microsoft products and services.

The multigenerational Azure Maia program has become a foundational element of Microsoft’s AI infrastructure strategy. Maia 200 is now serving Microsoft production workloads, including Microsoft Copilot, and supporting the MAI team as they develop next-generation models, with GitHub Copilot and Microsoft Foundry to follow by end of year. Across these workloads, Maia 200 runs OpenAI models, Microsoft’s own MAI models and leading open-source model families. Insights from the scale of Microsoft’s AI services allow for unique co-design between hardware systems and software applications to optimize real-world use cases across enterprise applications and industry scenarios.

Maia 300 is a complete intelligent system designed for the new types of AI workloads emerging today: Long-context reasoning, latency-sensitive conversational experiences and sustained, multi-step agentic tasks. By improving efficiency across the system, we expect Maia 300 to deliver 40% more tokens per megawatt compared to leading GPU-based systems. And in a full 96-accelerator configuration, it delivers up to 7x the rack-scale token throughput of Maia 200.

YouTube Video

Click here to load media Maia 200: Driving the workloads that Microsoft runs at scale

The Azure Maia system is designed around the real-world workloads Microsoft runs at scale and the models we expect to build in the years ahead. Because silicon development takes years, our software and hardware teams collaborate early to anticipate architectures, scale and capabilities, translating those requirements into greater memory capacity, higher bandwidth and more accelerators operating as a single system.

That co-engineering is already delivering measurable efficiency gains. Maia 200 is deploying across the United States, Europe and Australia and delivering up to 1.7x more tokens per watt for built-in AI capabilities in Microsoft Teams, Copilot and Cowork experiences compared with leading GPU infrastructure.

Experience running our AI models and services on Maia 200 directly informed the design of Maia 300. Together, the two generations are designed to support a broad set of additional Microsoft’s AI services: multi-step agentic work that will help GitHub deliver coding experiences more efficiently; cloud-scale model serving that will help Microsoft Foundry support diverse architectures and deployment environments; inference and data generation for the MAI team; and latency-sensitive interactions that will help Dragon Copilot scale workflows and reduce clinicians’ administrative burden.

AI is becoming a foundational layer of every application, workflow and business process. As enterprises move beyond chat experiences to systems of agents that can reason, act and collaborate, AI infrastructure must deliver intelligence securely and efficiently at scale while continuously adapting as models and workloads evolve.

As the third generation of our Azure Maia program, Maia 300 is designed with an inference-first architecture optimized for long-context, memory-intensive and decode-heavy AI workloads. This involves balancing compute, memory bandwidth, networking and power efficiency at cloud scale.

Key capabilities of the Maia 300 system include:

Less data movement, more useful compute: Moving data efficiently, rather than raw compute power, is increasingly the bottleneck in large inference systems. Maia 300’s software-defined dataflow architecture and AI-assisted kernel optimization work together to maximize locality, reduce data movement and improve utilization across compute, memory and networking resources. By treating dataflow as a first-class design principle, Maia 300 delivers a significant performance-per-watt advantage over traditional GPU architectures.

Higher memory bandwidth within every watt: As AI workloads become increasingly context and memory-bound, Maia 300 delivers industry-leading HBM bandwidth per watt. Eight HBM4 stacks provide 18.0 TB/s of peak memory bandwidth within a 950-watt TDP, enabling efficient execution of large models and long-context workloads. Specialized kernel support uses on-die SRAM as a software-managed scratchpad, reducing pressure on HBM.

Chiplet architecture for enhanced efficiency: A new multi-chiplet architecture separates compute, memory-controller and I/O functions, improving efficiency, scalability and cost structure. Compute chiplets are built on TSMC’s N2P process node to maximize performance-per-watt.

Larger context, richer intelligence: By designing for advanced low-precision formats, including enhanced MXFP4, Maia 300 reduces memory and bandwidth requirements for both computation and KV-cache storage. The result is larger context windows, higher throughput and improved model efficiency within the same power envelope.

Scale-up networking at datacenter scale: Maia 300 treats networking as a seamless extension of the software-defined dataflow architecture. Rather than moving data through separate communication layers, dataflow operations span compute, memory and network resources as a single system. High-bandwidth interconnects between Maia accelerators are optimized for tensor parallelism, while a custom AI transport delivers performant and reliable scale-up communication across accelerators and racks. The AI-optimized transport leverages Ethernet-based networking infrastructure to enable a unified topology that scales efficiently from a single rack to datacenter-scale deployments.

Higher AI throughput from every rack: Maia 300 rack systems bring together 96 accelerators within a 167-kilowatt power envelope, delivering 27.6 TB of HBM capacity and 1.7 PB/s of aggregate HBM bandwidth across a high-speed scale-up domain. That density allows Microsoft to serve more simultaneous AI interactions and larger workloads from each rack, improving datacenter capacity and the economics of expanding AI services.

Maia 300 will begin production deployments in 2027. For more information on Maia 300 architectural details, read our technical paper.

The future will not be defined by a single model or a single piece of silicon or a single innovation. Model and silicon choice are inherent in our vision for Frontier Transformation. Our focus is on developing responsible, affordable and scalable platforms that will help everyone amplify their intelligence to drive breakthrough outcomes that matter for them.

Microsoft is unique because we are able to optimize from silicon to systems to agent on an open, heterogenous and model diverse platform our customers can count on for their AI transformation, today and in the years ahead. This platform is at the heart of Frontier Transformation, and Azure Maia is a foundational component of our approach to expand access, accelerate innovation and help customers capture more value from AI.

Next month at Microsoft Ignite 2026 in San Francisco and online we’re looking forward to sharing more information on our strategy, technology adoption trends and success in Frontier Transformation.

For more details, photos and videos, visit our Maia 300 site.

Scott Guthrie is responsible for a set of hyperscale cloud computing solutions and services including Azure, Microsoft’s cloud computing platform, generative AI solutions, data platforms and information and cybersecurity. These platforms and services help organizations across the globe solve urgent challenges — and transform for the future.

The post Microsoft to expand Azure Maia system to multi-gigawatt deployments appeared first on The Official Microsoft Blog.

Satya Nadella, Chairman and CEO, and Ryan Roslansky, Executive Vice President of LinkedIn and Microsoft 365, shared the below communications with Microsoft and LinkedIn employees this morning.

After nearly 18 years at LinkedIn and Microsoft, Ryan Roslansky has decided to leave Microsoft and embark on his next chapter.

As Ryan and I have been discussing our roadmap and work ahead, we felt this was the right moment for this change. Ryan will stay through the end of the calendar year as an advisor to me and will support on the transition.

Ryan leaves the organization in a strong position. With Copilot, we are executing rapidly against our vision to build a new OS for work that spans every model, every form factor, and every task, with Office fully embedded in Copilot and vice versa! Over the past year, Ryan’s team has done critical work bringing together the product, engineering, and design foundations that have made this next phase possible. Last week’s launch was an important milestone in that journey, and the strong response from customers, partners, and the broader ecosystem reinforced both how far we’ve come and how much opportunity remains ahead.

At LinkedIn, we remain focused on our core vision of creating economic opportunity for every member of the global workforce, helping our 1.3 billion members access the knowledge, insights, and opportunities they need to navigate rapidly changing professional and labor markets.

Effective today we will make the following changes:

• Sumit Chauhan and Lan Ye will report to Charles Lamanna.

• Jon Friedman will report to Jacob Andreou.

• Dan Shapero will continue as CEO of LinkedIn and Mohak Shroff will continue in his role, both reporting to me.

Over his nearly two decades at Microsoft, Ryan has helped shape how people around the world find opportunity, build their careers, and get their work done. I’ve always appreciated his clarity of thinking and I’m deeply grateful for everything Ryan has built, for the leaders and teams he has developed, and for his friendship and counsel.

Please join me in thanking Ryan for his extraordinary contributions to LinkedIn and Microsoft, and in wishing him the very best in what comes next.

Satya

Team,

In 1996, as a college freshman, I taught myself to code so a few friends and I could help other students find a place to live around campus. I didn’t know much. I did know the internet was about to change how people found things, and I wanted to be as close to that as possible.

In 2009, Jeff Weiner made me his first hire at LinkedIn. I’ve spent the nearly 18 years since working on a much bigger version of that same idea: helping people find opportunity, and then helping them do great work once they have it.

This has (obviously) been the best professional experience of my life.

As a product manager, I got to build dozens of features that are still around today (some good, some not so good!). When I became CEO in 2020, the world was months into a pandemic that changed what work even meant. Nobody hands you a playbook for that. This team wrote one anyway. Since then, LinkedIn has passed a billion members and nearly tripled its business, and every bit of that belongs to the people who built it.

Last year, Satya asked me to help build the future of M365. Once again, the old rules didn’t apply. AI was rapidly changing how professionals work, and the world’s largest productivity suite needed to lead the way. That meant figuring out what Copilot, and the future of work, should look like as one ecosystem.

We still have a ways to go, but the progress over the past year has been inspiring to watch. A lot of talented people set aside their own roadmaps to make that happen. Thank you. I know it wasn’t easy.

Satya, thank you. When LinkedIn joined Microsoft, you gave us room to build, and then trusted many of us with a lot more. I’ve learned more from you than I can fit in an email, and I’m grateful for the friendship that came with it.

All of which made this a hard decision. I know what it would take for me to lead the next phase of this work well: being in Redmond full time, every day. But my life and my family’s life are here in the Bay Area. My wife Leigh gives her time to more local organizations than I can keep track of. Our youngest daughter just started high school, and our dog Buddy remains deeply committed to the trails on the Peninsula!

I thought about this for a long time. It’s the right call. The work now goes to leaders who are set up to take it further. I can’t wait to see what they do with it.

Over the years, people have asked me what’s kept me so energized and motivated in this work. I think the answer is that I got to spend my career working where doing good and doing well are the same thing. In technology, that’s the best place to operate.

Most companies start by focusing on doing well, so doing good becomes a compromise. It’s the last slide of the PowerPoint, or a separate .org. But when doing good is built into the product itself, it becomes a real competitive advantage. When we build valuable products, people find jobs, they hire, they learn skills, they start companies, they run companies. That creates value for members, customers and the world. And for our business too.

Not many companies get to work this way, and not many people get the chance to build something like it. We do.

That matters more now than it ever has. Technology will continue to touch every job on the planet, and the people building it will decide what kind of world it helps create. I can’t think of a better group to figure that out than the people reading this note.

Thank you for the best 18 years I could have asked for.

RR

The post Microsoft 365 and LinkedIn leadership update appeared first on The Official Microsoft Blog.

Every organization has knowledge that is uniquely its own: the way its business operates, the data it has built over time, the expertise of its people and the context behind thousands of everyday decisions.

As powerful AI models become more widely available, that unique business intelligence matters more than ever. The difference will come from how effectively organizations combine those models with the data, expertise and business context that make them unique.

Turning that intelligence into lasting advantage requires three elements to work together: human ambition to set the direction, intelligence grounded in the organization’s unique knowledge and trust to provide the foundation for operating at scale. Microsoft’s approach is designed to bring those elements together so organizations can build on what they know and create greater value over time.

That value can compound through a continuous learning loop, where AI draws on an organization’s data and expertise, improves the systems and experiences it creates and generates better outcomes over time without making the organization dependent on any single model provider.

Today at the European Microsoft Fabric + SQL Community Conference in Barcelona, we are introducing advances across Microsoft Fabric and Azure Databases that help organizations put their data estate at the heart of a successful AI strategy.

As Microsoft Copilot evolves across Home, Code and Autopilot, access to trusted business context becomes essential. Fabric IQ brings governed business context into Copilot, grounding experiences in the Power BI semantic models, metrics, relationships and business definitions organizations already trust.

With Fabric IQ in Microsoft Copilot Chat and Cowork now generally available, users can ask business questions and receive answers grounded in governed Power BI data, while Cowork can use those insights as part of broader multistep work. Integration with the new Code experience is coming through the Frontier program.

That means when someone asks Copilot for the latest sales forecast or another critical business metric, the answer can be grounded in the same semantic model that powers the organization’s reports, helping avoid confusion about which number is correct. For Power BI customers, this extends the semantic foundation they already have into Microsoft Copilot while preserving existing access controls and governance.

I experienced this myself recently as I was preparing for a sales review and asked Copilot, “What’s our latest sales forecast?” With Fabric IQ, that answer was grounded in the same governed Power BI semantic model that powers the team’s sales reporting, using the same business definitions and metrics my team already trusts. So, instead of spending the meeting uncovering and debating whose number was right, the team was able to get straight to what the numbers mean and focus on what we do next. That’s exactly the kind of experience we want to create.

For Fabric and Power BI customers, this capability is on by default in Copilot Chat and Cowork and extends the semantic foundation they already have into Microsoft Copilot while preserving existing access controls and governance.

Business knowledge does not stop at the boundary of a team or business unit. Organizations increasingly need to collaborate across customers, partners and broader ecosystems while maintaining control over the information they share.

IQ Sharing in Fabric, now in preview as part of Fabric IQ, breaks down company boundaries by enabling organizations to securely share data and the business context that gives it meaning, while maintaining governance over the underlying assets. Starting with key Fabric IQ artifacts and context, IQ Share is designed to evolve beyond preview to help customers discover, share and activate trusted context wherever it’s needed.

Together with OneLake mirroring and shortcuts, these investments create a more open foundation for sharing governed data and context across organizations, platforms and ecosystems.

As organizations put more workloads on their unified data platform, they need to understand system health and address issues before they affect business and AI outcomes.

Observability in Fabric helps teams move from reactive troubleshooting to proactive operations. Cross-workspace monitoring and an enhanced Monitor Hub provide a unified view across jobs, capacities, workspaces and workloads, while Operations Agents and Activator help teams investigate issues, identify root causes with AI and automate responses.

Together, these capabilities help organizations maintain oversight as they scale data and AI.

AI is transforming how applications are created, but enterprises still need a secure, governed path from idea to production. Fabric Apps brings application development directly to governed data in Microsoft Fabric, helping organizations build and scale AI-powered business applications faster.

With new capabilities for data connectivity, transactional workloads, backend logic, storage and policy-controlled access, developers can move from prototype to production on a unified data foundation. These experiences bring application creation closer to where enterprise data and insights already live, helping more people turn data into action.

A continuous learning loop is only as good as the data behind it, which must be prepared, refined and improved without pause. As those environments become more complex, organizations need automation that can take on more of that work without pushing people or controls out of the loop.

Earlier this year, we shared our vision for autonomous data engineering. Now, new agentic data engineering features in Fabric bring that vision to life, bringing autonomous execution to complex data engineering initiatives while keeping organizations in control. Engineers define the outcome and set the boundaries. Inside those limits, the agent plans and executes the work, validates the results and refines the approach.

Maintaining control over business knowledge also means having flexibility over where critical data and workloads run.

Now generally available, SQL Server on Azure Local brings SQL Server and Azure services into local environments, giving organizations more choice to run workloads closer to their applications and operations. SQL Server on Azure Local with disconnected operations, now in preview, extends that flexibility to environments where cloud connectivity is interrupted, restricted or unavailable.

These options give organizations operating across regulated, local, edge and disconnected environments greater choice over where their critical data and workloads reside and operate.

Intelligent applications depend on operational data. Teams need a clear view of the databases that power those applications and practical tools to manage them.

Building on its introduction at FabCon Atlanta, Database Hub in Fabric is now in public preview. It provides a central control plane for managing SQL Server, Azure SQL, PostgreSQL, Cosmos DB and SQL databases in Fabric, helping administrators monitor and govern their database estate.

Complementing this management experience, specialist database agents for SQL and PostgreSQL in Database Hub in Fabric and Visual Studio Code, coming soon to public preview, help teams investigate issues, understand performance bottlenecks and identify optimization opportunities within their development environment.

Together, these capabilities help teams understand their database estate and make more informed operational decisions.

Our goal is straightforward: help organizations use the best AI models and tools available while turning their proprietary data, expertise and workflows into an advantage that grows more valuable over time. That direction is reflected across today’s announcements as the data foundation comes into focus.

At Ignite, we’ll broaden this story across the full Microsoft Frontier ecosystem, connecting data, applications, agents, governance and infrastructure into one system that turns AI ambition into business outcomes.

Explore the complete set of announcements across Microsoft Fabric and Azure Databases in today’s blog from Arun Ulagaratchagan, Executive Vice President of Azure Data.

The post New Microsoft data innovations unlock what only your business knows appeared first on The Official Microsoft Blog.

We’re reimagining Microsoft Copilot to enable work as it evolves and to help expand what every individual and every organization can accomplish in the flow of human ambition. Today we’re introducing the new Copilot to connect the tools people rely on with the next generation of capabilities they’ll need to build, customize and scale AI across work.

The Copilot app now has three new capabilities: Home is your new starting point, where Chat and Cowork come together — and with Office in Copilot, the full power of Word, Excel and PowerPoint is now built into the experience. Code lets everyone build their own solutions with the tools to run them safely, and it’s powered by the same underlying technology as GitHub Copilot. And Autopilot is a persistent, proactive and personal agent that keeps working even when you’re not. Home and Code will start rolling out in our Frontier program in the coming weeks and Autopilot is expanding to private preview at the end of the month.

We’re also excited to share new FinOps for AI capabilities to help you manage spend and get the most value from Copilot and agents.

YouTube Video

Click here to load media A new experience for every mode of work

Home is your new starting point in Copilot, where Chat and Cowork come together in one place. As work moves constantly between meetings, chats, documents, emails, apps and agents, Home is where you get your bearings, keeping you grounded and moving forward. You can review recent activity, get suggestions on what Copilot can help with, and pick up work right where you left off. Chat is instant and conversational — ready for quick questions, lookups, draft-creation and keeping you in the loop on every turn. Cowork is for work you delegate. You define the task and it runs end-to-end to return a completed result: an RFP response, a launch kit, a customer briefing, a financial close package or any other complex work worth iterating on. And coming soon, you won’t need to choose a mode: simply state what you need to accomplish, and Copilot will route the work to Chat, Cowork or Code — whichever capability is right for the job.

We’re also bringing the full power of Word, Excel and PowerPoint to Copilot with Office in Copilot. Now you can get to real work fast: no broken formatting, no stray versions, no rebuilding the context in another app. From Home you can ask Copilot to draft a launch brief, model a budget, build a deck, or pick up a file your team is already working on, and it creates or updates a real, editable document, workbook or presentation — live for your whole team, not just you. Refine it side by side with your conversation, or @mention a teammate and their edits show up as they make them — things stay in sync between Copilot and the Office apps, so progress travels with you wherever you work.

If you haven’t used Copilot in the Office apps lately, take another look. PowerPoint now keeps decks on-brand automatically and enables you to fine-tune slides and images or quickly apply a custom template. When Copilot or another collaborator makes edits in Excel, you can see exactly what changed and why, so you’re up to speed quickly, and Copilot recommends the best chart to visualize your data. And across Microsoft 365, Skills help bring specialized expertise directly into the apps, like financial skills in Excel and legal skills in Word, and enable organizations to scale best practices through reusable, role-specific workflows.

Code moves solution-building outside the realm of developers alone. To date, the unit of knowledge work has been the file: the document, the spreadsheet, the deck. Those aren’t going anywhere, but Code adds a fourth: small, purpose-built solutions anyone can create to get a job done. Learning to build them is becoming as basic a skill as writing a memo or modeling a budget.

Describe an app, tracker, dashboard, automation or workflow in natural language, and Copilot chooses an approach and builds it — from persistent desktop widgets for quick reference and interactive dashboards for exploring data and testing outcomes to cloud-hosted internal apps you can share with your team.

Powered by the same underlying technology as GitHub Copilot, Code runs in a sandboxed environment and can be hosted securely within your tenant. With Microsoft IQ, everything you create is grounded in the context of your work, and plugins sync automatically. Your software developers continue to use GitHub Copilot for their day-to-day work with more connectivity into the Copilot platform. Code is rolling out to Frontier at the end of the month, with broad availability in the coming weeks. It will be in preview for Microsoft 365 Premium and Pro subscribers later this year.

As it gets easier and faster than ever to build in Copilot, we’re introducing Microsoft Copilot Managed Runtime: hosting infrastructure that lets code run safely right inside your company’s Microsoft 365 environment. It’s governed by IT but easy for everyone else: share an app with teammates, connect it to live data and access it from anywhere. This same foundation enables apps built in Cowork, Code and Copilot Studio, and we’re opening it up to third-party and pro-code developers, too. Copilot Managed Runtime is now in preview and will also be accessible inside Code.

Autopilot, previously called Scout, is your digital teammate. And it works for you: Give it a name, a role and a goal, and it goes to work — watching channels, following up on threads, running recurring work and picking a project back up days later, without waiting for a prompt. It can do things like set up and manage a full supplier review process: building the schedule and workback plan, then handling prep, meetings and follow-ups on its own, right down to reaching out to stakeholders for updates. And Autopilot is cloud-hosted, so it keeps working while you sleep or your attention is elsewhere — no constant monitoring required.

Autopilot lives in your tenant with its own identity, memory, computer and workspace, and it’s built on Microsoft IQ so it understands how your organization actually works. It shows up where people already work — Teams, Outlook, chats, channels and documents — so you can @mention it like a colleague, with permissions, audit and governance behind it. You set the objective and boundaries; Autopilot handles the rest while keeping you informed and in control.

Copilot is grounded in Microsoft IQ, the unified intelligence platform for enterprise AI across your Microsoft stack, connecting how the business operates with the knowledge it holds into one real-time view for Copilot and your agents, governed and secure.

Today we are expanding the business context available to Copilot. Fabric IQ brings trusted context from your data, including the 20+ million semantic models in Power BI into Copilot. It grounds Chat and Cowork, and is generally available in both today, with Code integration coming through the Frontier program soon. Rolling out over the next month in public preview, Copilot is now grounded in data and workflows from Dynamics 365 and Power Platform. So, when a salesperson is working on a proposal, for instance, Copilot can incorporate data like deal history and support tickets right into the document, with no app-switching or distracting lookups required.

You also need a trusted way to connect Copilot to the actions and data that move your business forward. Plugins extend what Copilot can do, bringing together skills, connectors and other capabilities so users can complete more workflows without leaving the apps they use every day. Rolling out now and generally available across surfaces in the coming weeks, a new plugin registry will bring Microsoft, partner and custom-built plugins together in one unified catalog. IT can approve and manage plugins centrally, while developers and partners can publish once to extend Copilot across supported experiences. The result is a simpler way to introduce trusted capabilities at scale, with the control your organization needs.

For everyday AI — quick answers, first drafts, summaries and analysis — you want the best possible quality at a fixed cost. A user subscription license (USL) gives you exactly that: Copilot in Chat and across Word, Excel, PowerPoint, Outlook and Teams; model selection; and new capabilities and models. Auto is at the heart of the USL. Auto weighs accuracy, speed and cost on each request to route to the model best suited for the job. Agentic work requires a different type of spend, one that gives you the best quality possible with the most advanced models. Usage-based billing (UBB) lets you have full control to choose models and match cost to value. Cowork, Code, and Autopilot, new long-running agentic capabilities, and frontier models like Astra and Fable all run on UBB. As AI capabilities and the ways people use them continue to evolve, we’re evolving Copilot pricing alongside them. Read more here.

FinOps for AI is a shared discipline for managing AI spend and optimizing it for business value. We are committed to enabling organizations to make the most of their AI spend, and, today, we are introducing new capabilities across Agent 365, Insights and Microsoft Copilot to help do that. Cost management in Agent 365 is expanding beyond Cowork and Work IQ APIs to include Code and Copilot Managed Runtime, with support for agents built in Microsoft Copilot Studio planned for October. Admins can manage spending policies at scale with API access; route credit requests from end users into existing approval and automation workflows; and set which model families are available to different groups of users — those settings also shape which models Auto can select from. Business leaders and admins can understand which Cowork tasks are generating the strongest outcomes and determine where to sustain, expand or optimize usage. And finally, end users can view credit usage, remaining balances and usage history directly in Microsoft Copilot, so they can better manage their own AI usage within organizational guardrails. We are committed to investing in FinOps for AI to help organizations stay in control of spend, understand value and optimize their AI investments for impact.

These announcements are just the beginning. We’re excited to share future Copilot experiences that make Copilot more personalized and proactive, and help the team move forward together.

Coming soon in Home, Today is a proactive, personalized command center across mail, calendar, Teams threads, meetings and tasks. It shows you what you missed, what needs attention now and what can wait. Today doesn’t just surface the work, it gets work done. The draft is written and ready for you to edit; the schedule change is already proposed; and the forgotten follow-up is waiting for you to send it, complete with context so you can catch up quickly. Today enters private preview in October in Copilot, and will be coming to Outlook and Teams as well.

YouTube Video

Click here to load media

And now when you invoke @Copilot in Teams, your entire team gets the shared context and permissions of the channel, group chat or meeting. Ask why a product decision was made, and Copilot pulls the background from across conversations — no digging, no re-explaining. When that decision ripples into manufacturing, packaging and launch timing, Copilot goes further: it offers to prepare a cross-functional review with dependencies, owners, open questions and decisions, so everyone can see how their work connects and what still needs resolving. The team stays aligned and the project keeps moving forward. The @Copilot capability will be in private preview in Teams by the end of the month.

With the new Copilot, you now have one place to ask, delegate, build, automate and more. It’s rolling out now through the Frontier program, with new capabilities arriving continuously. Check back here for the latest.

There is so much more coming to Copilot. Join us at Microsoft Ignite — November 17–20 in San Francisco and online — to learn about the latest Copilot innovation and how it’s empowering people and organizations to reimagine work and get value from AI.

The post Introducing the new Copilot with Home, Code and Autopilot appeared first on The Official Microsoft Blog.

AI is reshaping work faster than any organization has fully mastered. Across industries, the conversation has shifted from what AI can do to how companies can use AI to create business value and expand what people are able to achieve. At Microsoft, we believe the organizations that succeed will be what we call Frontier Firms: human-led, but increasingly AI-enabled. That responsibility begins with how AI is built and continues through how it is put to work: AI should expand human capability while people retain meaningful control, judgment and accountability. We committed to being Customer Zero, learning through our own transformation so we could help others navigate their own.

Our employees have experimented with AI, while leaders have set ambitious goals and challenged teams to reimagine how we work to achieve more than was possible before. We created cross-company councils spanning corporate functions, go-to-market and engineering to share best practices and learn together. We asked everyone to challenge their fixed mindsets and embrace the growth mindset we have cultivated for more than a decade. That work is producing measurable results: for a sales team deal close rates increased by 20%1; selected supply-chain workflows cut cycle time by up to 75%2; and a nine-person engineering team shipped an initial product release in 35 days.3

As proven approaches emerged, we codified them into case studies so we could accelerate transformation, scale what worked and learn from what did not. Just as importantly, we knew that if we wanted to help customers realize the full value of AI, we had to do the work to transform ourselves first. Our own first-hand experience needed to be a source of learning we could share with others.

We have been sharing Microsoft’s Frontier Playbook with customers as a practical guide to our AI transformation journey, including what we’ve learned, what has worked so far and where we’ve grown from failures. Drawing on hundreds of AI transformation efforts across the company, the playbook captures what we are learning as we redesign work, build new capabilities, measure impact and help people grow alongside AI.

The playbook also reflects important truths: transformation is hard, and learning is the durable superpower. Among the many insights gained from our successes and our failures, five lessons consistently stand out.

We initially treated AI like a traditional technology rollout: deploy the tools, provide training, drive adoption. We learned that access and usage do not equal transformation: a tool licensed to over 200,000 people does not change how the work gets done.

Early sales usage made this clear. Despite broad deployment, usage plateaued and impact did not materialize. Rather than push adoption harder, the team started from the business goals — deliver more value to customers, win deals and improve employee experience. They mapped how account managers spent their week and identified the best tools for the moments that mattered most: an Analyst agent for pipeline, a Deal agent for deal packages and Researcher for deep customer understanding. Weekly peer-led huddles turned experimentation into habit and scaled best practices to everyone on the team. Within the group, adoption of priority use cases tripled, revenue per account manager rose 9.4% and close rates were 20% higher.4

Success still required investment in helping people build new skills, experiment with new ways of working and learn from one another. But when leaders focused on a clear business outcome and what mattered most to the person doing the job, rather than AI adoption itself, conversations shifted from using AI to creating value.

One of our biggest lessons came from reimagining workflows end to end, not applying AI to existing steps. Early efforts helped people complete familiar tasks faster but rarely transformed outcomes. Adding agents to a broken process still leaves a broken process — speeding up one step just creates a longer queue at the next.

Our cloud supply chain team simplified its processes before reimagining them with agents. Supply chain experts and engineers worked side by side, first mapping and simplifying end-to-end workflows, then created a single source of truth so every agent reasoned from the same data. With that foundation in place, they deployed more than 100 purpose-built agents across planning, sourcing, fulfillment and logistics. Those agents investigate shifts in demand and model capacity while comparing transportation options across air, land and sea on cost, timing and carbon impact — complexity few teams could manage alone. Cycle time fell by up to 75% in selected workflows.

The shift isn’t only about speed, it’s about adding new value by improving what the team can see, anticipate and act on. Within defined permissions and approval thresholds, agents have progressed from answering questions to helping planners update or cancel purchase orders directly. Planners who once spent five to seven days tracing why a demand plan changed can now get an answer in hours, and sometimes in less than 20 minutes. That makes it possible to analyze changes as planning cycles unfold, model more scenarios, build better contingency plans and identify risks earlier — helping the team make better decisions and improve the performance of the supply chain.5

We are seeing the same shift in software engineering, where the opportunity extends beyond generating code faster to redesigning how teams plan, build, test and evaluate products with agents across the workflow.

We’ve found the largest gains come when teams step back and redesign how work should flow across people, process and technology from start to finish — including what agents can access and do, how their actions are monitored and where people must review, approve or intervene. AI is most powerful when all three advance together.

The people who do the work know where processes break down, where judgment matters and where AI could help — insights that no process map can fully capture. Their expertise needs to shape transformation from the start. Leaders are responsible for setting a clear ambition, helping employees build the skills to contribute and giving them a meaningful role in deciding how the work changes.

At Microsoft, we are creating hands-on ways for employees to build those skills. An early in career development program PRAISE pairs emerging engineers with experienced preceptors and AI-assisted learning, helping newer engineers contribute to complex work while developing their craft.

Camp AIR is a multi-week AI transformation accelerator that helps cross-functional teams learn new AI capabilities while redesigning how they work together around a real business challenge. An early pilot taught us that AI transformation is a team sport and that tools and training alone were not enough. While individuals could learn new technologies independently, meaningful and lasting change occurred when teams learned, experimented and adapted together. Teams needed candid conversations about how AI would reshape roles and workflows, along with the freedom to experiment safely and build confidence in new ways of working. We could not future-proof all jobs as they exist today, but we could help employees future-proof their careers by developing the skills, adaptability and mindset needed to succeed as work evolves.

That lesson became a core design principle of Camp AIR and has helped the program scale to more than 3,000 engineers across that organization.

The team behind Copilot Cowork shows what this can look like in practice. The nine-person team of engineers, designers and product managers was given the freedom to rethink how a product gets built, with AI embedded from day one. Working alongside agents, their roles expanded into what they called meta-engineers, meta-designers and meta-PMs. Together, they shipped an initial release in 35 days and documented what they learned so other teams could build on it.6

In our experience, leaders set outcomes and accountability, while the people closest to the work see where AI adds value and where human judgment must stay central. Managers connect the two, and their role-modeling and support has been one of the strongest predictors of success. Our research bears this out: when managers actively model AI use, reported value from agentic AI rises 17 points and trust in it rises 30 points — and employees on teams where managers create psychological safety are 1.4 times as likely to be high-frequency users of agentic AI. Where employees have context, capability and agency, they can become the engine of transformation.

We started where many companies start: automating tasks to increase efficiency. That value is real, but only the beginning. We realized over time the larger opportunity is “Capability Add”: combining human and AI strengths to achieve outcomes that were previously impractical or impossible. Think of it as an equation: CI + AI = CA. Continuous improvement takes waste out and AI adds capability in. Together, they produce Capability Add — output with higher strategic value. Continuous improvement alone results in a leaner version of the old company; Capability Add creates a different one. This equals transformation.

In our latest Work Trend Index, 58% of AI users said AI helps them do work they could not do before — it was 80% among advanced users. In our case examples, it looks like predicting a quality failure instead of catching it or exploring twenty options where a team had time for three.

This broader view shapes how we measure ROI. We like to say efficiency is the floor; capability is the ceiling. Usage and adoption, along with improvements in speed, quality and cost, are important signals. But the ultimate measure is whether AI improves customer and employee experiences, drives growth and innovation, reduces risk and expands what the organization can accomplish. Because those outcomes can take time to emerge, we also track leading indicators. For account managers, that might mean more time with customers, a stronger pipeline or better win rates before revenue gains fully materialize. For an engineer, it is not how much code AI produces, but whether the team is building better products faster.

We encourage teams to look beyond efficiency and ask, “What could our people accomplish that they couldn’t before?”

At Microsoft, we have long believed that a growth mindset is a source of competitive advantage. Our willingness to challenge assumptions, experiment, learn and adapt has shaped our culture for more than a decade — and provides an even more important foundation for the AI era.

AI is giving that idea new meaning. As a thought partner, it can help employees explore new ideas, build skills and learn in the flow of work. It can make learning faster, more personalized and more accessible, helping people take on challenges that once felt out of reach.

But we have also discovered that learning moves in both directions. People help AI become more useful by bringing the ambition, context, judgment and feedback that teach it how work gets done and what good looks like. They set the requirements, correct what falls short, determine where human judgement is essential and remain accountable for how AI is applied.

In our People organization, for example, teams are continuously improving agents to support the many cross-company activities involved in onboarding a new employee. Our team guides the process and sets the bar for what good looks like, determines where human judgment is needed, addresses issues and governs the overall system.

Over time, this learning loop turns individual insights into organizational capability. AI helps people learn faster, while people help make AI more useful within their organization. We recognized that the continuous learning loop we were creating was also generating unique institutional knowledge, and that retaining control of it would be critical to our long-term advantage. And that every organization should be able to build its own continuous learning loop without becoming dependent on any one model provider, while retaining control of its knowledge and how it is embedded in the models and systems it uses.

Having access to the best models will not be enough. The organizations that lead will be those that build the capacity to accelerate learning — where people and AI improve together, and where the knowledge they create becomes a source of lasting advantage.

***

These lessons come together in three practical recipes: Persona Acceleration for a role, AI-Powered Process Redesign for a workflow and AI-First Possibility for a greenfield opportunity. Distilled from hundreds of transformation efforts, they are patterns teams can adapt — often in combination — to move faster and avoid reinventing the journey.

Access to AI will not be the differentiator. The advantage will come from an organization’s ability to empower and engage employees, redesign work, govern AI responsibly and operationalize what works. The results we are seeing — from increased revenue and a more efficient supply chain to faster product development and better employee experiences — show what is possible when those pieces come together. But achieving these results requires more than deploying AI. It requires the intentional alignment of people, processes and technology. That is the hard work of becoming a Frontier Firm — and why we created Microsoft Frontier Company: to bring what we are learning from our own transformation to customers and work alongside them as they navigate their transformations.

Read Becoming a Frontier Firm: Our Frontier Playbook for the frameworks, transformation recipes and lessons from Microsoft’s Customer Zero journey.

Kathleen Hogan is Executive Vice President and Chief Strategy and Transformation Officer at Microsoft, where she leads the company’s enterprise-wide strategy and transformation agenda, accelerating Microsoft’s evolution into a Frontier Firm. Previously, she served as Chief Human Resources Officer and Corporate Vice President of Microsoft Services.

NOTES

1 Internal Microsoft sales team data based on 687 sellers of Microsoft 365 Copilot from Jan. – June 2024, as compared with sellers with low usage of Copilot. Regular usage of Copilot means sellers who use Copilot daily at least 50% of the time during the testing period.

2 Based on Microsoft internal analysis of work led by a 150+ person cross-functional team between September 2025 and August 2026. As of September 2026, more than 111 agents had been deployed across cloud supply-chain workflows. Across 5 monthly planning cycles measured between April 2026 and August 2026, average cycle time declined from approximately 10 to less than 2.5 business days. Separately, across 20+ demand-plan investigations each month, average time to produce a human-validated explanation declined from five to seven days to less than a few hours, with some completed in less than 20 minutes. Results are specific to these workflows and measurement periods.

3 Based on internal project records. The 35-day period was measured from the team’s formal project kickoff to delivery of the initial release in Spring 2026. This result reflects one project undertaken by a dedicated cross-functional team and is not a companywide product-development benchmark.

4 Internal Microsoft sales team data based on 687 sellers of Microsoft 365 Copilot from Jan. – June 2024, as compared with sellers with low usage of Copilot. Regular usage of Copilot means sellers who use Copilot daily at least 50% of the time during the testing period.

 5 Based on Microsoft internal analysis of work led by a 150+ person cross-functional team between September 2025 and August 2026. As of September 2026, more than 111 agents had been deployed across cloud supply-chain workflows. Across 5 monthly planning cycles measured between April 2026 and August 2026, average cycle time declined from approximately 10 to less than 2.5 business days. Separately, across 20+ demand-plan investigations each month, average time to produce a human-validated explanation declined from five to seven days to less than a few hours, with some completed in less than 20 minutes. Results are specific to these workflows and measurement periods.

6 Based on internal project records. The 35-day period was measured from the team’s formal project kickoff to delivery of the initial release in Spring 2026. This result reflects one project undertaken by a dedicated cross-functional team and is not a companywide product-development benchmark.

The post What we’ve learned from Microsoft’s own AI transformation appeared first on The Official Microsoft Blog.

Microsoft has been in the business of education for five decades. Since the earliest days of this company, we have worked alongside educators, students, families and educational leaders through each new wave of technology — and we have learned what helps learning and what gets in the way.

That commitment continues today, in a technological moment unlike any we have seen. AI can personalize learning, reduce administrative burden, accelerate research and expand access to high-quality instruction. It also brings real risk. Educators and parents are asking whether students are still doing the thinking, and whether the tools their children use are safe. Those are the right questions.

At its core, teaching and inspiring students to become their best selves is a deeply human endeavor. Education is built on a meaningful partnership among educators, students and families that helps unlock human potential and shape brighter futures.

No one company understands entirely how AI will reshape education. But uncertainty calls for more care, not less. That is why Microsoft is taking a holistic approach, grounded in five principles.

1. Safety, privacy, security and transparency by design
2. Educators remain in control and at the center of AI use in the classroom
3. AI designed to support students’ learning, not replace their thinking
4. AI strengthens education systems, from learning to discovery
5. Every student prepared for an AI powered future

1. Safety, privacy, security and transparency by design

Every AI system used in education should be designed with robust safeguards that protect student privacy, enhance safety, reduce risk – including safeguards that address bias – and provide the transparency that students, educators and parents deserve.

We believe institutions should never have to choose between innovation and trust.

Security, privacy and transparency are foundational to Microsoft’s approach in education.

That is why last week, we signed a landmark agreement with the American Federation of Teachers (AFT) and introduced a new Privacy & Safety Standard for Schools covering Microsoft Education products. The agreement represents the first time a major technology company and one of America’s largest teachers’ unions have come together to define what trusted AI should look like within these products. The Standard limits how Microsoft Education products can use student and educator data, requires human oversight for consequential decisions, provides meaningful transparency for families and holds Microsoft accountable when we fall short. It also affirms that schools retain ownership of the knowledge and ideas they create.

We did not build this Standard to keep it to ourselves. Students and institutions deserve these protections, whichever technology their school chooses. We invite others across our industry to meet this bar with us.

2. Educators remain in control and at the center of AI use in the classroom

Educators should remain at the center of teaching and learning. AI should augment educator expertise, not replace professional judgment. Educators should help shape the design of AI systems and be equipped with the skills and support needed to use AI in ways that support student outcomes.

Technology alone does not transform education. Educators do. At a school in the Bronx, special education teacher Ashley Hernandez is using Copilot to develop differentiated, sensory-aware materials for students with complex communication and medical needs. Technology helps her adapt learning experiences more efficiently and creates more time to work directly with students. Just as importantly, her experience has helped shape the future of the technology itself.

Educators know best what works in the classroom, and their voices are reflected in the products we create. Through our Education Insiders Program, educators and school leaders participate in private product previews and feedback sessions to help us improve tools before they’re released.

That is why we build products like Teach in Microsoft 365 Copilot. Teach is our education-first AI experience for educators. Designed around real instructional workflows, it helps educators adapt materials for different learners, build lessons that match their standards and spend less time planning. Our goal is not to automate teaching; it is to lift the administrative burden that gets in the way.

Customers tell us this works. Brisbane Catholic Education achieved a 275% increase in learner agency among at-risk cohorts. Since deploying Microsoft 365 Copilot, Miami Dade College has seen a 15% increase in student pass rates and a 12% drop in course dropout rates.

Educators need practical training, communities of practice and in-demand credentials. Through Microsoft Elevate for Educators, we are connecting educators with those resources and pathways. Over the last year, more than 17 million people have completed an in-demand AI skills credential. Through the Elevate for Educators program, we help educators build the understanding and confidence they need to teach about AI effectively and responsibly. We also offer practical materials that help educators introduce generative AI safely and responsibly to students. 

3. AI designed to support students’ learning, not replace their thinking

AI systems in education should be grounded in learning science and designed to promote healthy student engagement, critical thinking, student well-being and improved learning outcomes.

According to our recent research on learning science, a significant risk of AI in education is cognitive offloading that creates the appearance of learning without the cognitive work that makes learning durable. This is a real problem. Students think they are learning when they are not.

Many commonly used tools were not designed for education. They prioritize speed over understanding, and many institutions have responded by limiting or blocking access entirely, even as students continue to use AI on their own. Without a purpose-built learning experience, trust becomes the barrier to enabling AI for students at all.

Education needs AI designed to support the purposes of education: helping people build knowledge, critical thinking, judgement and confidence.

Some of the most important learning happens when students are stuck. Difficulty builds understanding that lasts. AI that removes that struggle can quietly move the thinking from the learner to the machine. We must protect the work that helps learners grow.

Microsoft’s AI experiences in education are age-appropriate by design — including default-off access to Copilot Chat for K-12 students with administrator controls for age-based enablement as learners progress. In earlier grades, Minecraft Education builds foundational AI literacy while Learning Accelerators personalize support and provide timely feedback. As students advance, the Study and Learn Agent uses scaffolded questions and interactive practice to help learners work through concepts without doing the work for them. For older learners, Microsoft 365 Copilot introduces the same enterprise-grade tools they will use in their careers.

4. AI strengthens education systems, from learning to discovery

Strong education institutions create opportunity by identifying needs early, acting quickly and directing limited resources where they matter most.

AI can help institutions support students sooner, reduce paperwork, use information more effectively, strengthen security and accelerate research and discovery.

Today, important information is often scattered across different systems and teams. Used responsibly, AI can bring it together so schools, colleges and universities can make better decisions and take coordinated action.

Broward County Public Schools confronted a significant budget shortfall requiring operational efficiency across 235 schools. Despite being data-rich, the district lacked real-time insights. They deployed Microsoft 365 Copilot integrated with its Microsoft 365 ecosystem and identified significant operational inefficiencies that helped them achieve projected facilities savings of $40 to $50 million over five years. That savings could be reinvested in supporting their educators and students.

At University of North Carolina, medical researchers are using a secure research environment built on Microsoft Azure to conduct data-driven clinical research while protecting sensitive patient information. It’s now supporting 25 active studies and expanding access to advanced analytics and AI capabilities within a controlled research environment.

Innovative educational systems do more than adopt new technology. They drive coordinated action that results in better outcomes for learners, educators, researchers and communities.

5. Every student prepared for an AI powered future

Every student should have the opportunity to develop the knowledge and judgment needed to use AI responsibly and the skills to prosper from the emerging AI economy.

The most consequential AI divide of the next decade may not be who has access to AI. It may be who has access to AI skills and the pathways that turn those skills into economic opportunity.

The AI Economy Institute, Microsoft’s flagship think tank, recently released a book exploring how AI technologies shape the education experience and found that AI-related skills command wage premiums up to 40%, surpassing returns from advanced degrees.

AI Literacy is now nonnegotiable across every discipline. The importance of understanding AI fundamentals, limitations and ethical implications — including healthy skepticism toward outputs — is universal.

Offerings like Minecraft Education AI Ready Skills help students build that foundation through a game-based pathway culminating in a credential aligned to CSTA and ISTE standards

Our partnerships with the American Federation of Teachers, the National Education Association, National Applied AI Consortium and North America’s Building Trades Unions help to ensure this work is informed by educators, workers and community members alike. Their insights help guide the tools we build, the skills we teach and the pathways we create so that students can succeed in an AI-powered economy.

Expanding our NABTU partnership, we recently introduced the Jobs Machine to help parents, students and high school counselors explore the skilled trades as a valuable education and career pathway.

In higher education, institutions are turning AI literacy into curriculum. Auburn University created a “Teaching with Artificial Intelligence” course now available to nearly 10,000 faculty at more than 90 institutions. Babson College built an interdisciplinary AI lab — The Generator — integrating Azure OpenAI tools and Copilot assistants across every faculty member and student. And the University of South Florida built a student ambassador program that has driven 55 to 60% time savings on some tasks and now demonstrates practical AI use in real department workflows.

The aim is not to train learners for one fixed set of tools. It is to help them develop the knowledge, agency and judgment to keep adapting throughout their lives.

The future of education will not be defined by technology alone. It will be defined by how effectively we bring together innovation, trust and human potential.

Building the future of education, together

Microsoft will continue investing across all five principles and, more importantly, alongside the educators, learners, institutions, governments, labor organizations and communities shaping education’s next chapter.

AI is already changing how people learn and work. The responsibility and the opportunity before all of us is to ensure it strengthens learning, expands opportunity and keeps people at the center.

That future is already taking shape, and we are committed to building it together.

Justin Spelhaug leads Microsoft Elevate, a global initiative helping education, workforce and nonprofit organizations expand opportunity and help prepare people for the AI economy. During his 28 years at Microsoft, he has held leadership roles advancing technology’s impact for people and communities. He began his career serving in the U.S. Marine Corps.

The post Microsoft’s commitment for AI in education appeared first on The Official Microsoft Blog.

As we enter the next era, what will be the defining measure of our progress?

Every industry has a word that shapes how it thinks. For pilots, it’s safety. For insurers, it’s risk.

For the semiconductor industry, it’s yield.

Yield does not ask how elegant the solution is, how many years it took or what the roadmap promised. Rather, it asks one simple question: What useful output did we produce?

For more than 60 years, the semiconductor industry has asked that question, relentlessly maximizing the number of usable chips produced from every wafer. Generation after generation, wafer after wafer, it is precisely that discipline that turned the transistor from a laboratory curiosity into the foundation of modern life.

Today, we need to apply the same principle to the unprecedented resources that the world is pouring into AI: capital on a scale once reserved for nations, gigawatts of power and record-breaking fabs and datacenters.

The question that will define this decade is the same one this industry has always asked: What actually comes out? Not just chips and tokens, but as affordable intelligence, as work that matters and as outcomes that improve lives.‌

That is the yield imperative.

YouTube Video

Click here to load media The limit of more 

AI has proliferated with remarkable speed, at a rate of adoption faster than the internet, the PC or even the smartphone. Yet, global penetration still stands at just 18% of the working population, and the vast majority of that usage is chat-based. As systems move from answering individual prompts to reasoning, planning, using tools and executing longer agentic workflows, the infrastructure equation changes dramatically. A single agentic task can use more than 3,400 times as many tokens as a typical chat interaction.

We are only in the early innings of agentic adoption, and the infrastructure is already strained. Power is setting the limits on what we can build and when. Packages and racks are growing larger and denser. Memory is becoming an even tighter constraint.

For years, the industry’s rational answer to each new requirement resulted in more: more silicon in the package, more memory beside it, more power to feed it and more fiber to connect it. Each generation delivered meaningful progress. But when each new gain requires more input than the one before it, we are on a treadmill. It moves only as long as we keep adding to it.

I believe we need to pursue two paths forward. The first is evolutionary: we continue improving the architectures we have today, driving incremental efficiency, utilization and economics within each generation. The second is transformational: changing the curve itself with innovation in new architectures, new materials and new approaches to system and model design.

The history of our industry is defined by transformations like these. When increasing CPU clock speeds ran into the power wall, we moved to multicore processors. When planar NAND reached its limits, memory went vertical.

And now, once again, we have an opportunity to challenge our assumptions and rethink the fundamentals. Because the next chapter of AI won’t be defined simply by how much infrastructure we build, it will be defined by how much intelligence we can create from it.

For decades, the computing industry has optimized yield in the context of manufacturing. Today, that discipline has to extend across layers, from datacenters and silicon through models and the agentic harnesses that orchestrate them. And the work does not stop once the technology is built. We must then deploy and scale it faster, while developing new tools and systems to maximize utilization across our fleet.

From development through execution, each layer has a yield of its own, and losses and gains compound across them. Capacity at any one layer is only a starting point. The real measure is how effectively those layers work together to produce useful output from the system as a whole.

Our experience at Microsoft building and operating AI infrastructure at scale has reinforced two key lessons. First, the biggest constraints are rarely solved in the layer where they appear. Second, when we attack a constraint across the whole stack, tradeoffs that seemed inherent to the problem often turn out to be artifacts of the architecture.

The greatest advances often come when we apply these learnings through co-design, working across layers to turn apparent limits into solvable system constraints. Innovations in memory, networking and power show what this approach looks like in practice.

Today, memory is viewed as a supply problem or a component problem. In reality, it is a system problem.

In AI inference, memory is now setting the limits on system performance. It must hold larger models, preserve longer contexts and deliver data fast enough to keep the compute fed. And agents raise the bar even further. Generation, retrieval, tool use and persistent memory run together in loops that can last minutes or hours. The result is a much longer memory horizon, with far more information kept close to the compute and available across an expanding sequence of turns. Doing that efficiently at scale will define the next generation of AI infrastructure.

Our experience building the Azure Maia platform demonstrates that memory bottlenecks are not resolved by a single layer.

Model architecture, data science and compression can reduce the amount of KV cache, which stores the model’s working context during generation. Software can manage memory hierarchies more effectively, silicon can be optimized for data movement efficiency and compilers can place data closer to compute. No one change removes the constraint. Together, they increase the useful intelligence the system can deliver from the same memory resources.

That is useful yield: not simply adding bytes but getting more useful intelligence from every byte we already have.

As we zoom out to the cluster level, we see that intelligence does not come from one chip. It comes from thousands of chips operating as one system. Faster links matter, but the productivity of the system also depends on congestion management, failure recovery, workload placement, programming complexity and the boundaries between silicon, system and software. Together, they determine whether expensive compute is producing intelligence or sitting idle.

When architecting the platform for Maia, we did not begin with an existing networking design. We began with the outcome we wanted to deliver: efficient inference at fleet scale, designing across silicon, networking and system software. Instead of separate scale-up and scale-out fabrics, we built a two-tier scale-up network, integrated the NIC functionality directly into the chip and developed a custom transport layer.

The result is scalable, consistent performance across dense inference clusters, with a unified fabric that simplifies programming, improves workload flexibility and makes better use of available capacity. And with less network hardware needed to deliver this performance, we also lowered the cost of running the entire system.

Our objective is not merely to move data faster. It is to keep more compute productive and deliver more tokens from every watt and every dollar.

Moving from the cluster to the grid, AI has introduced new challenges around power availability, distribution and utilization. Racks have gone from tens of kilowatts to hundreds of kilowatts, and datacenter campuses can operate on the scale of gigawatts. Power used to be something the system simply plugged into. Now, it is something we design around, from the grid to the chip.

That is why the industry is rethinking power across the system. Solid-state transformers and 800-volt direct current power delivery can reduce distribution losses as power moves through infrastructure. Power and cooling are no longer downstream of the design, they are part of the product definition from the start. And increasingly, that co-design is needed all the way into the silicon.

Azure Cobalt 200, our Arm-based server CPU, shows what this looks like in practice. We designed Cobalt so that every core has its own voltage and frequency controls, paired with software-based, per-virtual-machine power capping. This finer-grained control enables targeted power adjustments while protecting the performance of critical workloads, allowing us to run more servers within the same power envelope.

As Cobalt demonstrates, hardware-software co-design enables us to more effectively turn every megawatt into customer value.

The pattern we see across memory, networking and power extends throughout the system: start with the useful output, then optimize the whole rather than any one layer. This first requires us to be precise about the output we are optimizing for and which design constraints are truly fixed. Are we optimizing for peak performance or sustained system throughput? Would the workload benefit from significantly more capacity with marginally less redundancy? What creates more value: a broader set of capabilities or significantly earlier customer deployment?

Not every constraint in today’s systems is a law of physics. Some are inherited from decisions made elsewhere in the system and can change only when we work across traditional boundaries. Evolution comes from the steady gains each company drives within its own domain, but transformation comes when we challenge those assumptions together and redesign the system as a whole.

The breakthroughs ahead will emerge from collaboration across the ecosystem, spanning hyperscalers and silicon providers, equipment makers and materials innovators, utilities and datacenter operators, model builders and software developers.

But tokens and intelligence are not the finish line. What we produce becomes the input for someone else’s work.

What matters next is how broadly that input translates into productivity across the economy and value in people’s lives, whether it helps a scientist accelerate discovery, a clinician identify a signal earlier, a student get help at the right time or a small business find a new path to growth.

That happens as AI becomes part of everyday work across industries and around the world. People build on it, new uses emerge, the tools improve and the value compounds.

For that cycle to spread, AI must be broadly accessible. And at scale, accessibility depends on efficiency. It is the only way to deploy enough intelligence, and at a cost that allows it to reach everyone.

When every person and every company can access intelligence, build on it and create value of their own, that is full yield.

We will continue to build capacity because the world will need it. But our defining measure of progress must be what comes out: not only chips or tokens, but useful intelligence translated into empowerment, opportunity and human achievement.

That is the yield imperative. And it is work our entire industry must take on together.

Rani Borkar leads the core organizations responsible for planning, architecting, developing and deploying hardware and infrastructure for Microsoft’s leading cloud computing platform — from silicon, to systems, to supply chain.

Learn more about Microsoft’s silicon to systems approach to Azure infrastructure.

The post The yield imperative: Turning AI infrastructure into useful intelligence appeared first on The Official Microsoft Blog.

Throughout this past fiscal year, customers across every industry and segment moved from AI experimentation to deploying AI for real-world business outcomes. They unlocked innovation and created new opportunities for growth. We saw the emergence of Frontier Firms as they moved beyond efficiency gains to focus on human ambition and embed AI at the core of how they operate. Successful customers are building an intelligence platform so their unique IQ — their knowledge, data, workflows, applications and expertise — can continuously compound, ensuring the value of AI accrues to the customer, not the model. They have a trust platform that is pervasive, with the ability to manage, govern, secure and measure AI across every business process.

Everything we are doing at Microsoft is empowering Frontier Transformation: Copilot enables AI in the flow of human ambition, Microsoft IQ amplifies and protects an organization’s IQ and Agent 365 is the trust platform that enables observability at every layer of the stack.

Businesses are not static, and neither are the AI systems that support them. As organizations evolve, AI systems must continuously learn and improve. Agentic workflows need to be built, observed and tuned against the outcomes organizations seek and the ROI they demand. Microsoft’s open, model-diverse and heterogenous platform powers that improvement loop. We also recently announced Microsoft Frontier Company, bringing our AI engineering approach to customers around the world to help them build these AI systems to accelerate measurable business outcomes.

Throughout the past year, we saw customers put these capabilities to work in powerful ways — embedding AI into core business processes, building agentic systems, strengthening security, accelerating innovation and creating new sources of value. The stories below highlight organizations leading Frontier Transformation, demonstrating how intelligence, trust and human ambition come together across industries.

To advance its journey to become a global AI-powered company, Atos Group deployed Microsoft 365 Copilot to 56,000 employees across 54 countries — from consultants to engineers to frontline workers — and was one of the first organizations globally to adopt Microsoft 365 E7: The Frontier Suite. Using Microsoft Foundry, Microsoft Copilot Studio and Agent 365, Atos is building, operating and governing a growing ecosystem of 19,000 AI agents through a unified operating model that brings together productivity, security, compliance and agent governance. As Atos embeds secure agentic AI across its workforce, the company is creating a repeatable model to continuously improve thousands of agents at scale while applying the same playbook to help customers accelerate adoption across highly regulated industries.

Facing state-sponsored threats and complex global operations, ASM is strengthening cyber resilience with Microsoft Security Copilot, helping protect the intellectual property behind advanced semiconductor manufacturing. By bringing threat investigations into a unified AI-powered experience, ASM enables analysts to investigate incidents faster, apply consistent decision-making across global operations and accelerate the development of cybersecurity talent. The company reduced incident triage time by 68%, cut laptop compromise investigations from 25 minutes to eight and now saves 337 hours each week on investigations while redeploying 20% of its security operations staff to governance, risk and compliance initiatives.

Banco Popular Dominicano, the largest private-sector bank in the Dominican Republic, transformed operational risk management from periodic, sample-based reviews into continuous, AI-powered supervision. Using AURA — an ecosystem of specialized agents built on Microsoft Copilot Studio and Microsoft Power Platform — the bank monitors 100% of its operational risk universe in real time, up from roughly 40% coverage, and can automatically analyze changes, validate controls and surface issues as they occur. The shift has delivered seven times greater analytical capacity, reduced manual operating effort by 70%, achieved 98% methodological accuracy and enabled continuous processing of approximately 80,000 documents per week and more than 300 cases per day. Just as importantly, risk teams have moved from reacting to problems after the fact to anticipating and preventing deviations before they occur. These results demonstrate the power of AI democratization. By enabling business teams to build intelligent solutions themselves through low-code tools, Banco Popular transformed operational risk management while fostering a culture of innovation led by domain experts.

To help reduce the manual burden for employees while meeting the pharmaceutical industry’s strict data security requirements, Cactus Life Sciences modernized scientific workflows with Microsoft 365 Copilot and agents. The company has deployed more than 30 custom automation agents to streamline document review and structure data extraction and information retrieval across scientific writing and project management teams. Supported by a centralized knowledge repository and the Copilot Champions community, the company reports efficiency improvements of approximately 35% to 50% in structured data extraction. By automating labor-intensive tasks and maintaining human review and quality controls, the company is enabling scientific writers to focus on deeper analysis, synthesis and delivering exceptional science to clients.

Chow Tai Fook is redefining luxury retail with Microsoft 365 E5, Microsoft Purview, Microsoft Azure OpenAI Service, Microsoft Fabric and Microsoft Foundry. The company has deployed over 400 customized AI agents supporting more than 24,000 employees, with millions of AI interactions each month and core business-process efficiency gains exceeding 70%. Through its AI Fook super-agent ecosystem, frontline associates can instantly access product expertise, inventory insights and personalized recommendations, helping drive sales conversion improvements of up to 57% while delivering hyper-personalized omnichannel experiences at scale. With hundreds of AI agents operating across the business, Chow Tai Fook is creating a foundation where customer, product and operational intelligence can be applied across every interaction, helping personalize experiences and improve decision-making across its global retail network.

To accelerate AI adoption, EY moved AI from experimentation into enterprise-wide transformation. After deploying Microsoft 365 Copilot to 150,000 employees and realizing a 15% productivity gain, the firm is expanding the Microsoft 365 Frontier Suite across its global workforce of more than 400,000 people, embedding agentic AI capabilities across the enterprise. As Client Zero, EY is applying Microsoft technologies across its own operations, including Microsoft Power Platform, Microsoft Copilot Studio, Microsoft Azure, Microsoft Foundry and Microsoft Fabric. The results include 95% faster lead times, a more than 37% reduction in finance operating costs and up to a 90% reduction in manual workloads across key business processes.

To reimagine the grocery shopping experience, Grandiose Supermarkets created an AI-powered shopping companion — GrandChef — built on Microsoft Foundry and Azure OpenAI Service. By connecting meal inspiration, recipe discovery and product purchasing into a single experience grounded in live product and inventory data, GrandChef helps shoppers move from intent to purchase faster and with greater confidence. This has led to a 31% increase in conversion, a 20% lift in average basket value and shopping journeys that are 40% faster. Together, these gains are helping Grandiose Supermarkets create more personalized customer experiences while driving measurable business growth. By connecting customer intent, product data and purchasing decisions in a single experience, Grandiose is creating a feedback loop that helps continuously improve recommendations and shopping experiences.

To support audits across its global organization, Grupo Bimbo built two agents with Microsoft Copilot Studio and deployed them through Microsoft 365 Copilot and Microsoft Teams: the result was Audit Assist, an idea generated from a Microsoft-supported internal hackathon and Comatrix. By connecting auditors directly to approved guidance, procedures and templates in SharePoint, the company is helping teams work more efficiently across 39 countries while improving consistency and audit quality. The solution/AI agents reduced planning-phase audit time by 20% and accelerated risk and control matrix creation from days to seconds, enabling auditors to spend less time searching for information and more time on analysis and decision-making. By making approved audit knowledge instantly accessible across its global audit organization, Grupo Bimbo is creating a foundation where expertise can scale with the business — improving consistency, reducing rework and enabling auditors to focus on higher-value analysis and decision-making.

To create a unified foundation for data, automation and AI across its operations in 47 countries, Navien built a connected, intelligent operating model with Microsoft Fabric, Microsoft 365 Copilot, Microsoft Foundry and Microsoft Copilot Studio. By connecting fragmented data and processes across procurement, manufacturing, quality and customer service, the company is enabling more consistent, data-driven decision-making across its operations. Navien saved 28,000 hours annually through AI agents and automation, with 32% of employees using self-service analytics to make decisions without IT support. Through its migration to Azure, it also expects more than 1.4 million in total cost-of-ownership savings over five years. By connecting data, insights and workflows across the business, Navien is enabling faster, more consistent decision-making at scale across its global operations.

NHS England is accelerating AI adoption across the healthcare system with the largest implementation of its kind in the healthcare sector. Following a trial involving 30,000 workers across 90 NHS organizations — where users saved an average of 43 minutes of administrative time per day — NHS England is rolling out Microsoft 365 Copilot to over 500,000 clinicians and support staff. Organizations can use Microsoft Copilot Studio to build and deploy AI agents that streamline clinical, operational and administrative workflows. Through Agent 365, NHS England can govern and scale those agents across the healthcare system while enabling individual trusts to build solutions for local needs — creating a secure framework where agentic workflows can be deployed, managed and expanded consistently to improve service delivery, reduce costs and create more time for patient care.

Novo Nordisk is using AI to help researchers make faster, more quantitative decisions in pharmaceutical R&D. Working with Microsoft’s AI Acceleration Studio within the Forward Deployed Engineering team, the company built a governed reasoning agent on Microsoft Azure with its proprietary dataset, including more than 200,000 patient-years of harmonized clinical trial data, while maintaining rigor, oversight and compliance. The system has expanded the team’s capacity to evaluate potential opportunities from 5 to 10 strong ideas per quarter to more than 50, and the company expects it to reduce time to insight for exploratory analyses from weeks to minutes. By grounding AI in its datasets and governed workflows, Novo Nordisk is helping scientists turn decades of institutional expertise into a reusable intelligence layer that can help accelerate discovery across the organization.

As AI becomes part of investment decision-making, SimCorp is helping financial institutions bring AI into investment workflows without compromising governance, auditability or control. By unifying SimCorp One on Microsoft Azure and leveraging Microsoft Foundry, the company is helping portfolio managers, risk analysts and operations teams access insights faster, automate manual processes and spend more time on higher-value decisions. By standardizing how AI is deployed and governed across global investment operations, SimCorp is creating a foundation for trusted AI at scale, helping organizations embed AI into investment workflows while maintaining the controls required in highly regulated markets. One study found SimCorp One customers realized 134% ROI over three years, improved operational efficiency by 45 percent, saved 10 hours per person per week and accelerated time to market by 50 to 60 days.

Stellantis is accelerating AI-led strategy and digital transformation across its global business, co-developing more than 100 AI initiatives across sales, customer care, product development and operations. By applying AI-powered insights across the business, the automaker is streamlining product development and validation, advancing predictive maintenance and bringing new digital features and services to market faster for customers. Stellantis is also deploying an AI-driven global cyberdefense center to help protect vehicles, customers and operations worldwide while modernizing its infrastructure on Microsoft Azure with a targeted 60% reduction in its datacenter footprint by 2029, powering a more scalable and interconnected digital ecosystem for future digital and connected services and resilient operations.

Facing a growing volume of cyberthreats across a complex healthcare environment, St. Luke’s University Health Network is using Microsoft Security Copilot to help protect the systems clinicians and patients depend on every day. Across 15 campuses, 300 outpatient sites and more than 2.5 petabytes of data and patient records, the organization needed a unified view of threats across a complex environment. Security Copilot connects Microsoft Defender, Microsoft Sentinel, Microsoft Entra, Microsoft Purview and other security tools, helping analysts correlate threats faster, eliminate silos and respond with greater precision. St. Luke’s is saving nearly 200 hours each month in phishing alert triage and creating incident reports in minutes instead of hours, helping security teams focus more time on protecting patient care.

University of Kentucky unified more than 150 AI initiatives across classrooms, research labs, healthcare settings and administrative offices through its CATS AI governance framework. Standardizing on Microsoft’s AI portfolio — including Microsoft 365 Copilot, Microsoft Dragon Copilot, GitHub Copilot and Microsoft Azure — the university achieved campus-wide deployment, providing more than 70,000 students and employees with access to AI capabilities. Clinicians are using Dragon Copilot to reduce documentation burdens and spend more time with patients, and students are using GitHub Copilot to become active builders of digital solutions, such as the Socratic Tutor: an AI-powered learning platform aimed at helping medical students master complex curriculum.

Looking back on FY26, I am inspired not only by the continued pace of AI innovation, but by what our customers are achieving. Across every industry and segment, organizations are turning their unique IQ into strategic advantage with continuously improving agentic workflows. The companies leading this next phase are building, observing and tuning agentic workflows against business outcomes. Intelligence compounds, trust scales and AI can work in the flow of human ambition. As Frontier Firms redefine what is possible with AI, we remain focused on helping our customers amplify and protect their intelligence so they can transform how they operate, compete and grow.

Judson Althoff is the chief executive officer of the commercial business at Microsoft. He is responsible for the product strategy, sales, services, support, marketing, operations and revenue growth of the company’s commercial business, which operates in more than 120 regional and national subsidiaries globally.

The post Looking back on Microsoft’s FY26: From AI experimentation to Frontier Transformation appeared first on The Official Microsoft Blog.

Editor’s note: Updates with additional details on the model’s crash score.

The physics of cybersecurity are changing. Autonomous systems can now reason, adapt and operate continuously. At the same time, the cost of offense is falling, while the volume, velocity and complexity of what must be secured continues to grow. Attackers can generate exploits faster, scale campaigns further and operate with unprecedented efficiency. The approaches built for a world of human actors cannot keep pace with a world of AI, agents and machine-speed attacks.

Security needs a new cyber stack. A new cyber stack must continuously perceive risk across the entire digital estate, reason across vast amounts of context and take action at machine speed. It must learn and adapt as environments evolve, helping organizations stay ahead of threats. And because security is ultimately a human mission, it must amplify defenders with better insights and more powerful ways to act. The defining characteristic of the next generation of security systems will not be their ability to generate more alerts. It will be their ability to continuously perceive, reason and act.

That vision led us to build Project Perception. A new agentic security system designed for the realities of AI. It turns signals into real-time protections using AI to defend against AI.

Project Perception brings together signals, context, models and specialized agents into a continuously learning system of defense. It can reason, prioritize and act at machine speed while keeping humans firmly in control and empowering them with powerful new workflows.

Project Perception is based on a simple idea: effective defense requires continuous understanding of how an attacker sees the world, how a defender evaluates risk and how protections are improved over time. To accomplish this, Perception coordinates three classes of specialized agents. Red team agents identify potential paths to compromise before an attacker can exploit them. Blue team agents investigate, reason over context and determine what represents meaningful risk. Green team agents take corrective actions and strengthen defenses across the environment. Working together, these agents form a closed-loop system that continuously discovers, evaluates and improves an organization’s security posture.

A system like Project Perception is only as effective as the visibility it has, the actions it can take, the experience of the teams building it and the models it can use. Microsoft brings together all four.

We see across identities, endpoints, applications, data, clouds and AI systems, providing broad visibility across the digital estate. Equally important, we can help customers take action across those environments. Combined with decades of security research, threat intelligence and real-world operational experience defending organizations, these capabilities shape how Project Perception reasons, prioritizes and responds.

Security is a 24/7 mission. Organizations need protection that is highly effective, continuously available and affordable at scale. That requires more than access to the most capable model. It requires applying the right model to the right task. Project Perception adopts a multi-model architecture that combines frontier and specialized cyber models, optimizing for both quality and cost.

As part of this multi-model strategy, we are committed to bringing customers the best models for each security task, including innovating with our own specialized models. The first scenario is software vulnerability management, bringing MAI-Cyber-1-Flash inside MDASH, our software vulnerability multi-model team of agents. MDASH with MAI-Cyber-1-Flash delivers 96% (any-crash score*) on CyberGym, an industry benchmark. It also outperforms competitors like Mythos on the CyberGym leaderboard. And this same configuration delivers almost 50% of cost savings vs. the current MDASH configuration in market today. That’s the power of a well-tuned, multi-model system with access to uniquely rich historical training data. Next, Project Perception will take advantage of MAI-Cyber-1-Flash for many more security workflows, beyond the software vulnerability scenario.

We are bringing this vision to customers around the world through Project Perception, which enters public preview on August 3.

YouTube Video

Click here to load media A cyber stack built for agentic security

Delivering agentic security requires more than adding agents to existing workflows. It requires a new cyber stack, designed from the ground up.

The stack begins with signals and sensors that provide awareness across the digital estate. Security context transforms those signals into token-efficient understanding that agents can use. Models provide intelligence and reasoning. A harness coordinates models and agents across security workflows. Agents apply that intelligence across security workflows and actuators translate decisions into protection. Together, these layers create a continuous learning system that can understand risk, adapt to changing conditions and improve security outcomes over time.

While each layer provides important capabilities, the power of Project Perception comes from how they work together.

Effective reasoning requires more than raw signals. Agents need context.

Microsoft transforms its breadth of visibility, threat intelligence and security expertise into a security context that connects security data, knowledge and semantics across the digital estate. The result is a continuously updated representation of an organization’s assets, identities, relationships, risks and activities that gives agents a shared, near real-time, understanding of the environment they are helping to defend.

This shared understanding is foundational to how Project Perception operates. Rather than forcing agents to continuously gather, correlate and reconstruct context from raw signals, it provides them with immediate and token-efficient access to the information they need to reason over risk, prioritize actions and make decisions. By grounding every interaction in this rich security context, Project Perception improves the accuracy and consistency of reasoning while reducing the time, compute and cost required to operate at scale.

No single model will be optimal for every security task. Effective cyber defense requires applying the right model to the right problem at the right time.

For Project Perception, the right model is determined by the combination of quality, reliability, latency and cost. Rather than relying on a single model, Project Perception adopts a multi-model architecture that continuously selects the capabilities best suited to the task, optimizing for both effectiveness and economics. Because security is an always-on mission, sustainable economics are essential to operating protection at scale.

This approach is shaped by ongoing research, benchmarking and evaluation across frontier and specialized models. Our security researchers continuously assess models against real-world security workflows, enabling us to match each task with the model that delivers the best outcome. This allows customers to benefit from advances in AI without being tied to any single model.

Security teams do not need more information. They need better outcomes.

That is why actuators are a critical part of the cyber stack. Project Perception is deeply integrated across Microsoft Security products, enabling agents to connect insights to actions. Organizations can continuously reduce risk rather than simply identify it, helping defenders strengthen security while remaining in control.

Underpinning every layer of the cyber stack is a foundation of trust. Project Perception is built in alignment with Microsoft’s Responsible AI principles and inherits the security, compliance, governance and operational controls our customers already rely on. This ensures these capabilities are delivered with the same rigor, accountability and enterprise readiness that customers expect.

Security has always been a race between attackers and defenders. AI changes the speed, scale and economics of that race. Defenders need systems that can continuously perceive, reason and act alongside them.

 Project Perception is how we begin to build that future.

To learn more about Microsoft Security solutions, visit our website. Bookmark the Security blog to keep up with our expert coverage on security matters. Also, follow us on LinkedIn (Microsoft Security) and X (@MSFTSecurity) for the latest news and updates on cybersecurity.

Other resources:

• Enhancing AI Security Through Global AI Red Teaming 

• From research to reality: An interview with Microsoft VP of Security Research Taesoo Kim

Hayete Gallot leads Microsoft’s work to help organizations operate securely in an AI-driven world. Her scope includes identity, threat protection, compliance and data security at global scale.

Note: *Any-crash measures the ability of the agent to identify vulnerabilities that can crash the code under evaluation with an input that triggers any existing or 0-day vulnerability. Our 96% score is an any-crash score.

The post Rethinking security for the age of AI appeared first on The Official Microsoft Blog.

Today, we’re excited to share a long-term commitment to the Department of Energy’s (DOE) Genesis Mission, backed by a $60 million investment designed to accelerate AI for science and the breakthroughs it can deliver for the country. This deepened commitment includes Microsoft’s new Scientific Partnership Advancing Research & Knowledge coordination hub and program office, otherwise known as SPARK, that is focused on facilitating collaboration and scientific discovery for the Genesis Mission.

The Genesis Mission represents an ambitious vision, bringing together DOE’s 17 National Laboratories, world-class experimental facilities, decades of irreplaceable scientific data and next-generation computing into a single, unified platform capable of transforming how science gets done. The critical goal of the Genesis Mission is to double the productivity and impact of American research and innovation within a decade by embedding AI directly into the scientific process. It’s a bold step forward and exactly the kind of moonshot that defines American science.

At Microsoft, we believe that this goal is not only achievable, but also essential as a national security imperative and an economic engine for generations to come. With this commitment and the launch of SPARK, we’re ready to partner fully in this mission and build on our collaborations across government and with academia.

Today, Microsoft’s $60 million investment package will begin to accelerate AI for science in support of DOE and the National Laboratories. As sustained scientific impact requires more than infrastructure alone to enable lasting scientific outcomes, Microsoft’s investment package is structured in two parts:

• $40 million in Azure compute and AI credits, distributed over three years, to power the large-scale AI and scientific workloads at the heart of the mission. This gives researchers room to train, simulate and iterate at a scale that matches their ambition.
• $20 million in solution engineering enablement services. Dedicated engineering, architecture, deployment, adoption and acceleration support to turn cloud and AI capacity into operational scientific outcomes. This investment helps ensure that the Genesis Mission advances in practice and in principle.

Together, this investment reflects a simple conviction: While infrastructure opens the door, it is people, expertise and disciplined delivery that carry discoveries across the finish line. Microsoft is committing to both.

Great partnerships need more than good intentions and good technology — they need a clear way to work together. So, alongside this commitment, we are standing up a new program office and coordination hub: SPARK, or Scientific Partnership Advancing Research & Knowledge.

SPARK is the single front door and coordination hub for Genesis Mission collaboration with Microsoft. SPARK orchestrates the full breadth of what Microsoft brings to bear: program, technical, research, engineering, security, compliance, partner and field teams into one clear, consistent path from first idea to deployed science.

SPARK is built around five commitments:

• A dedicated Genesis Mission Program Management Office. This office handles intake, prioritization, a disciplined sprint-and-checkpoint cadence and a steady operational interface with DOE for alignment, reporting and sustained partnership.
• An AI for Science Center of Excellence. An integrated delivery team that moves use cases from concept to secure, compliant, scalable implementation, with enablement through office hours, hackathons and training grounded in responsible AI and reproducible research.
• Optimization of Azure credits. Ensures computing resources are directed toward the projects where they can accelerate science most.
• Management of technical services. Dedicated support to help labs adopt and operationalize AI-for-science capabilities, delivered by Microsoft teams and select partners.
• Joint research and development. A focused set of Genesis Mission-aligned challenge problems, co-developed by Microsoft researchers and DOE scientific leadership, with milestones, publications and IP governed by mutually agreed terms.

Underneath SPARK sits the full breadth of Microsoft’s secure, FedRAMP-authorized cloud and AI portfolio. This includes Microsoft Azure infrastructure, Microsoft Foundry and Zero Trust security through Microsoft Defender, Sentinel and Entra. Azure capabilities will help extend the DOE’s American Science Cloud capabilities by providing computing, AI, data and collaboration services that complement existing scientific infrastructure. Together, these capabilities will help accelerate discovery, enable secure collaboration across institutions and provide researchers with flexible access to advanced technologies and resources. This will enable DOE and National Laboratory scientists to move faster from hypothesis to discovery, without compromising the security, reproducibility and governance that mission-critical research demands.

Microsoft’s new commitments build on capabilities we recently announced through Microsoft Discovery and Microsoft Quantum. They are designed to accelerate scientific research and will support the Genesis Mission’s ambition to bring AI, advanced computing and emerging technologies more deeply into the scientific process.

To enable a unified research platform for Genesis Mission work, Microsoft will provide access to Microsoft Discovery, our integrated platform that unites AI models, simulation, data and experimental workflows driven by advanced cognition rooted in scientific method, within a single governed environment. Microsoft Discovery, now generally available, includes support for autonomous lab orchestration, integration of Microsoft Research’s AI models for science, continuous AI learning through the scientific loop and agentic memory, advances in Discovery Bookshelf for data curation and scalable indexing and deeper integration and multi-hop reasoning over enterprise science data estates, including data governance. Together, these capabilities are designed to help researchers move more quickly from data to insight, from simulation to experiment, and from promising idea to scientific breakthrough.

We also recently announced the preview of the Microsoft Discovery app, a local desktop experience that helps researchers, students and scientific teams begin working with Microsoft Discovery today. Our support for the Genesis Mission will include access to the Microsoft Discovery app to enable DOE and National Laboratory teams with early access to its capabilities.

Microsoft’s recent quantum progress further strengthens the foundation for Genesis Mission work. With Majorana-based quantum advances, including more reliable topological qubits and a roadmap toward scalable quantum computing, Microsoft is helping move quantum from long-range research toward practical scientific capability. For DOE and National Laboratory teams, that progress can open new ways to model complex materials, chemistry, energy systems and national security challenges that are difficult or impossible to solve with classical computing alone. Just as important, the same agentic AI for science approach behind Microsoft Discovery is helping accelerate quantum research itself, creating a reinforcing cycle in which AI speeds quantum breakthroughs and quantum computing expands what scientists can ultimately discover.

Working alongside and supporting federal researchers and universities, we’re already seeing real-world impact, from faster materials screening to advanced biosurveillance modeling. Four of the first projects taking shape under this partnership show what that looks like in practice:

Discovering critical energy storage materials and biosystems design (Pacific Northwest National Laboratory)
Using AI to dramatically accelerate the discovery of new energy storage materials through novel experimentation design and shrinking analysis of massive volumes of scientific data from years to weeks. In biosystems design, Microsoft Discovery is plugging directly into PNNL’s laboratory automation infrastructure to launch self-driving scientific workflows that autonomously design, run and fine-tune biological experiments in real time.

Strengthening biosecurity (Lawrence Livermore National Laboratory)
Pairing advanced AI with modern bioinformatics to identify emerging biological threats sooner and develop responses faster, strengthening the nation’s ability to stay ahead of the next outbreak and developing proactive defenses against evolving biosecurity risks. By applying advances in AI, scientific computing and biotechnology, Microsoft and the Lawrence Livermore National Laboratory are helping strengthen the nation’s ability to prevent, detect and respond to biological threats, advancing US biosecurity and biodefense priorities while supporting the long-term resilience and competitiveness of the US bioeconomy.

Autonomous labs for accelerated materials discovery (Johns Hopkins University Applied Physics Laboratory)
Supporting the development of AI-enabled autonomous (“self-driving”) laboratories through the integration of advanced foundation models and simulation tools — including MatterGen and MatterSim — to rapidly design, execute and optimize experiments for structural materials and superconductors. By combining autonomous experimentation with accelerated computational materials discovery, this approach dramatically shortens the path from scientific hypothesis to validated, manufacturable, mission-enabling materials, while exploring a vastly larger design space than is possible through traditional methods.

Accelerating nuclear energy permitting and autonomous energy operations (Idaho National Laboratory)
Applying AI to streamline the licensing, permitting and review of new nuclear projects. AI helps automate the development of complex engineering and safety analysis materials, a historically time-consuming and expensive process, to get this critical documentation into the hands of regulators and safety reviewers sooner. The lab also demonstrated that secure, distributed, hyperscale cloud can enable autonomous remote operations for nuclear power generators. Together, these efforts are helping bring safe and reliable power online faster and ensure scalable, efficient and reliable operations to meet the nation’s surging demand for energy.

We are entering an era where AI and quantum don’t just support the scientific process — they are essential to it. Through the Genesis Mission, Microsoft is committing to work as part of America’s research enterprise by providing hyperscale compute, advanced models, emerging quantum capabilities and dedicated expertise that run alongside the labs’ own world-leading systems. We are helping to ensure that the nation’s scientific strength is built on American technology. This commitment builds and expands upon the decades of collaboration through Microsoft Research and partnerships across the US government and our national labs. Now is the time to scale these breakthroughs across the entire R&D ecosystem.

The mission ahead is significant, and it should be. Whether it’s advancing the frontiers of energy, chemistry and materials science, hardening our national security or simply giving the country’s best scientists better tools to ask bigger questions, this is work worth doing and worth doing together. With a $60 million investment and SPARK, Microsoft is ready to help ensure America’s scientific leadership continues to accelerate.

Fueling scientific discovery is one of the most important things these technologies can do. AI and quantum are transforming how we tackle the toughest challenges facing the country, from sustainable materials and clean energy to biosecurity and drug discovery. The Genesis Mission is a pivotal step toward advancing that progress. We commend the DOE’s vision in building a unified, holistic approach to these critical national missions. For America to seize this opportunity and set the global standard, we must put these breakthrough technologies to work advancing the nation’s research and development.

We’re ready to deepen this partnership across government and academia. We move faster together. I’ve never been more optimistic about what’s possible when the nation’s scientific talent meets the best of modern AI and quantum. The Genesis Mission is that moment. Let’s get to work.

To get started, reach out to your Microsoft account team or email genesismission@microsoft.com.

Chris Barry leads Microsoft’s US Public Sector business. He partners with federal, state and local government and higher education agencies and organizations to accelerate digital transformation, bringing Microsoft’s secure cloud and AI capabilities to mission-critical work that serves communities across the nation.

The post Powering America’s Genesis Mission: Microsoft’s commitment to scientific discovery appeared first on The Official Microsoft Blog.

AI workloads are scaling faster than any single infrastructure approach can support — with more models, new agent-driven workloads and surging compute demand driving the need for greater specialization across the stack. To meet this need, Microsoft continues to evolve Azure’s infrastructure, including expanding its AI fleet with AMD’s most advanced AI and high-performance computing (HPC) solutions.

Our approach to AI infrastructure is designed to support the breadth of how AI systems are built and run. We closely work with industry innovators like AMD as well as our own purpose-built silicon and systems to provide customers with a comprehensive, open and heterogenous platform to achieve the best performance, cost and energy efficiency outcomes.

Building on our close collaboration with AMD, Microsoft is bringing AMD’s latest Helios AI platform and next-generation EPYC datacenter processors to Azure. These technologies will power three upcoming Azure offerings: HDv2 VMs for data processing, HXv2 VMs for electronic design automation (EDA) and ND MI455X v7 VMs for AI inference workloads.

YouTube Video

Click here to load media Built for AI data systems — Azure HDv2

CPU infrastructure is essential to the performance and efficiency of modern AI systems. AI accelerators depend on high-density, power-efficient CPU compute to process data, coordinate workloads and keep pipelines running at scale. Without this, training jobs don’t have enough data to learn from, and agents don’t have enough capacity to perform tasks on behalf of customers. Azure HDv2 virtual machines are one of our latest offerings designed from the ground up to eliminate these bottlenecks and empower massive agentic workload adoption.

Co-designed with AMD, HDv2 VMs expand Azure’s portfolio of purpose-built solutions for the most demanding CPU workloads from AI customers, including data preparation, search, reinforcement learning and agent coordination at scale. Featuring nearly 500 physical 6th Gen AMD EPYC CPU cores, 4 terabytes of RAM, 32 terabytes of local NVMe storage and 400 Gb Azure Boost networking, HDv2 VMs are built for the workload needs of our most demanding AI customers.

The AI era has created tremendous need and opportunity for firms developing the silicon products that power this infrastructure. For this reason, Azure HX virtual machines, launched in partnership with AMD in 2023 and featuring AMD’s unique 3D V-cache technology, have seen significant adoption among silicon design firms working to bring more capable and efficient AI silicon to market. Today, we are announcing the next step in our workload optimized journey for these customers, HXv2.

HXv2 virtual machines build on and extend the strengths of HX. They both continue the differentiation Azure offers for RTL simulation workloads by again employing 3D V-cache technology, while offering significant improvements to single threaded performance and memory. HXv2 VMs will feature 176 AMD 6th Gen EPYC CPU cores with a clock frequency of more than 5 GHz, 50% more addressable cache per core and VM sizes with nearly 2 or 4 terabytes of RAM, helping customers optimize their workloads to memory needs.

Azure HXv2 is also designed to support a broader range of technical computing workloads including scientific simulation, engineering analysis and other distributed memory applications. The significantly increased per VM and per core performance, and the inclusion of 800 Gb InfiniBand, enable large-scale MPI-based simulations and make HXv2 an ideal fit for a wide variety of HPC customers.

AMD, a leading HX-series customer, highlights this impact directly:

“Engineering teams are pushing the limits of simulation, chip design and scientific computing. At AMD, we experience those demands firsthand as we design future AMD EPYC CPUs and AMD Instinct GPUs. Azure HX is an important platform for scaling complex EDA workloads, and we’re excited about Azure HXv2, which is designed to deliver even greater performance and scalability. We look forward to continuing our collaboration with Microsoft as we help advance infrastructure for the world’s most demanding engineering and scientific workloads.”

— Mark Papermaster, Executive Vice President and CTO, AMD

The HXv2 also leverages Microsoft’s long-standing collaboration to optimize Synopsys AI-powered EDA solutions on Azure:

“As AI compute continues to push the limits of semiconductor design, our collaboration with Microsoft on the Azure HX-series demonstrates a shared vision for enabling customers to deliver next-generation AI systems with precision and scale in accelerated design cycles. These systems have enabled Synopsys customers to reliably and efficiently leverage cloud-based compute, extending EDA workloads beyond traditional infrastructure constraints so they can meet ambitious development schedules while maximizing design quality and delivering dramatic performance gains.”

— Shankar Krishnamoorthy, Chief Product Development Officer, Synopsys

ND MI455X v7 is designed for the reasoning, search and agentic workloads behind modern AI services. 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.

Together, these new capabilities expand Azure capabilities while giving customers more flexibility to choose the right compute for each unique AI workflow: from inference, to data systems, to chip design. Customer choice is a core design principle built directly into Microsoft Azure, and we’re excited to bring AMD’s most advanced innovations at production scale.

To learn more about Azure’s high-performance computing and AI infrastructure capabilities, visit Azure.com.

Scott Guthrie is responsible for a set of hyperscale cloud computing solutions and services including Azure, Microsoft’s cloud computing platform, generative AI solutions, data platforms and information and cybersecurity. These platforms and services help organizations across the globe solve urgent challenges — and transform for the future.

The post Microsoft expands Azure AI and HPC infrastructure with AMD appeared first on The Official Microsoft Blog.

Amy Coleman, EVP and Chief People Officer, shared the following communication with employees today.

When I stepped into this role, I promised to communicate more openly with you and share the “why” behind our decisions.

Today we are eliminating around 4,800 roles, about 2.1% of our global workforce, as we focus our people, investments, and energy on the priorities that will keep Microsoft positioned to deliver for customers in a fast-changing industry. The people whose jobs are impacted today are our colleagues and friends. They have made meaningful contributions to Microsoft, and we are deeply grateful for everything they have done.

Decisions like these are never easy, and you have my commitment that we are always looking for ways to reduce the need for job eliminations. Whenever possible, our priority is to place people into new roles aligned to the company’s highest priorities and greatest areas of opportunity. Over the past year, we have redeployed more than 4,000 employees into new roles, including another 500 this month. We will also transition four of our gaming studios to operate under new management, with the goal of preserving both their intellectual property and ongoing projects. In addition, more than 30% of eligible employees chose to participate in our recent voluntary retirement program, and we will continue exploring similar approaches in the future. While this doesn’t change the difficulty of today’s news, we will continue to do everything we can to create opportunities for our people, reduce the need for job eliminations where possible, and responsibly support those affected with care and respect.

The “why” is this: Our business is changing because the world around it is changing. The way technology is built, deployed, and used is transforming faster than at any point in my time here. Our customers’ needs are shifting, the business models that serve them are shifting, and that means the work itself – what we do, where we focus, and how we’re organized – has to transform too. Companies don’t get to choose whether their industry changes; they only get to choose whether they change with it. That means we will need to adjust resources and roles and shift how we operate so we can have the greatest impact for our customers.

I also want to be direct that the roles eliminated today are not being replaced by AI. At the same time, what is true is that AI is changing how work gets done. Some of the tasks we do every day can now be automated, and that means we all need to keep learning, keep building new skills, and keep adapting as the work evolves. Our customers are navigating this same shift, and they’re counting on us to help them through it. We can’t do that well unless we’re doing it ourselves. This comes down to two commitments: making the decisions needed to drive differentiated customer value, and supporting the people affected by them.

First, we will make the hard changes required to build differentiated products and services that deliver differentiated customer value. We are aligning our investment, people, and energy to our business priorities. Today’s changes mostly fall within our Commercial and XBOX organizations. In our Microsoft Commercial Business, they build on last week’s Frontier Company announcement, reshaping how we work and embedding our engineering experts alongside customers so we can help them accelerate their technology deployments. In XBOX, we are restructuring to position the business for long-term success. Engineering teams across the company will also evolve their structure and priorities to meet customer needs and innovate for the future.

Second, we will do this thoughtfully.  As mentioned above, we are working on alternative solutions to job eliminations, and beyond this, we will continue to invest in equipping employees with new skills, including in AI. For those who are impacted, we provide financial support and resources to help them take their next step.

I know many of you want to help those who are leaving but aren’t sure how. Reach out and check in on your colleagues. Use your network to bring people together, share what makes them exceptional, and help create connections to opportunities that might not happen otherwise.

We are still early on this journey, and there will be more changes ahead; other parts of our business will need to make similar changes. Each time, you can hold us to the two commitments.

During my time at Microsoft, I’ve seen this company reinvent itself again and again. What makes that possible has always been our people – their resilience, creativity, and willingness to keep learning.

Thank you for everything you bring to Microsoft.

Amy

Read more: Resetting XBOX.

The post The latest in our company transformation appeared first on The Official Microsoft Blog.

The pace of AI adoption is moving incredibly fast. Customers have moved well beyond experimentation and understand the importance of adopting AI to transform their business. They are now concentrating on delivering measurable business outcomes and demonstrating a return on their AI investments, while ensuring their intelligence is amplified and their IP is protected.

Today we are introducing Microsoft Frontier Company, a new operating business focused on delivering Frontier Transformation through AI for our customers around the world. It will provide a unique combination of skills inclusive of deep industry knowledge, change management and continuous improvement experience, and enterprise-grade AI engineering expertise. This goes beyond what has been labeled as Forward Deployed Engineering (FDE) and will be the largest, most capable, outcome-driven engineering organization in the industry. We are making a $2.5B investment in Microsoft Frontier Company, embedding 6,000 industry and engineering experts at customers to co-design, co-innovate, deploy and continuously improve AI systems at scale based on measurable business outcomes.

I recently wrote more about my conviction that Intelligence + Trust are the two most important components of any AI solution and how our customers can use different levers to manage cost.

Companies need to establish an intelligence platform so their unique IQ — their proprietary data, expertise, workflows and decision-making processes — compounds over time from within, using their choice of models to build AI solutions and workflows.

They need a trusted platform that allows them to observe, govern, manage and secure AI solutions across every layer of the technology stack, using FinOps to assess their ROI.

Enterprise AI engineering expertise with deep industry knowledge is required to build a system that acts as a continuous loop of improvement between the two platforms to fine tune agentic business processes, ensuring that a customer’s intelligence compounds over time and delivers real business outcomes.

This is what Microsoft Frontier Company was built to do: focus on end-to-end Frontier Transformation, enabling customers to amplify their IQ with AI while refining their differentiated value in the markets that they serve.

Early results demonstrate meaningful impact: Our engineers and industry experts partnered with LSEG (London Stock Exchange Group) to embed AI into LSEG Workspace, helping finance professionals ask complex questions and get quick answers across structured and unstructured financial content. The solution is underpinned by a foundation that is iteratively refined through client feedback and real-time user testing that accelerates each cycle and steadily improves model quality and scope. From LSEG to Land O’Lakes to Unilever to Novo Nordisk, our differentiated approach is already delivering measurable outcomes on our customers’ Frontier Transformation journeys.

To achieve scale, we will work closely with our partner ecosystem to extend this unique value to our customers across all markets and segments globally. We have robust FDE partnerships with our Global SI partners, including Accenture, Capgemini, EY, KPMG, PwC and others.

Central to this approach is a principle that is non-negotiable: a customer’s IQ is protected. Their data, their IP, their competitive advantage — none of it is used to train models in ways that commoditize what differentiates them in their industry. Satya put it clearly recently: there is no societal permission for an AI future that eats the intelligence of the companies it’s deployed inside. We built Microsoft Frontier Company to make sure that does not happen.

We protect that intelligence with a model-diverse, open, heterogeneous AI platform. Customers shouldn’t be locked into a single model any more than they should be locked into a single technology vendor. Microsoft’s platform gives organizations the flexibility to run the right model for each scenario — whether it comes from OpenAI, Anthropic, Microsoft AI, open source or a specialized model tuned for a specific industry — without ceding control to any one of them.

To lead this new organization, I have asked Rodrigo Kede Lima to be the President of Microsoft Frontier Company. Rodrigo brings 30 years of industry experience, and for the past six at Microsoft has led enterprise-wide transformations as a sales leader in the Americas and Asia. He has been at the forefront of helping customers and partners translate technology shifts into business outcomes, and understanding how platform innovation, engineering and partner ecosystem collaboration come together to drive growth.

I am excited about all the things that Microsoft Frontier Company will do for our customers to realize the gains of Frontier Transformation. At the end of the day, it comes down to Intelligence + Trust and empowering our customers to achieve meaningful outcomes and a return on their investments.

Learn more at www.microsoft.com/en-us/frontier-company.

Judson Althoff is the chief executive officer of Microsoft Commercial Business. He is responsible for the product strategy, sales, services, support, marketing, operations and revenue growth of the company’s commercial business, which operates in more than 120 regional and national subsidiaries globally.

The post Microsoft Frontier Company: AI engineering that amplifies and protects your intelligence appeared first on The Official Microsoft Blog.

As demand for cloud and AI services continues to grow, datacenters are becoming more essential than ever. Communities also want to better understand how this infrastructure affects local resources, particularly water. At Microsoft, water stewardship has been a priority since our first datacenter builds in the early 2000s and remains a core part of our strategy today and into the future. It underpins both our Community-First AI Infrastructure initiative and our company-wide commitment to become water positive by 2030, meaning we will replenish more water than we withdraw. We are pairing our progress with continued transparency so people can understand not only how much water we use, but also how we are working to reduce that use over time.

Through continuous innovation and advancements in cooling technologies, we have improved our water use effectiveness (WUE), measured in liters per kilowatt-hour (L/kWh), by nearly 90% since our first generation of datacenters in the early 2000s. Our average WUE has decreased from 2.3 L/kWh to 0.27 L/kWh in 2025, reflecting decades of innovation and our ongoing commitment to reducing the water intensity of our datacenters while meeting the growing demand for cloud and AI services.

Across our entire owned fleet of datacenters, we are committed as a company to a 40 percent improvement in datacenter water-use intensity by 2030. As of 2025, we have reduced our water-use intensity by 25 percent, putting us well over halfway toward our goal. This strong progress reflects the impact of our continued investments in water-efficient cooling technologies, operational improvements and responsible water management practices.*

In FY25, Microsoft reached an important milestone toward its 2030 water positive commitment, replenishing more water than it withdrew across its global operations for the year. We have made this progress by decoupling datacenter growth from water use through resilient, responsible water stewardship practices and the deployment of increasingly efficient cooling technologies — demonstrating that digital growth and sustainable water management can advance together. We are committed to building on this progress and working toward sustaining water positive performance over time as we continue advancing toward our 2030 goal.

Beginning with some of our earliest datacenter designs, we prioritized water efficiency while minimizing impacts on energy use through the deployment of high-efficiency economizing chillers operating at elevated water temperatures. As early as 2008, we adopted direct air cooling with evaporative assist as the primary cooling approach across our datacenter fleet. This design uses significantly less electricity and up to 90% less water than traditional water-based cooling systems by relying on water only when outside temperatures exceed 85°F (29.4°C). In parts of Northern Europe, no water is required for cooling throughout the year, while in other cooler regions like Dublin and Amsterdam, water is used less than 5% of the time. In more temperate climates like Virginia, water is typically required only 10% of the year, while in the hottest climates like Phoenix, water use may increase to as much as 40% of the year. As a result, approximately 90% of our 2025 owned fleet operates using highly efficient, low- to zero-water cooling systems.

YouTube Video

Click here to load media

While the majority of our existing datacenters are already highly water efficient, we did not stop there. In 2024, Microsoft introduced a new datacenter design optimized for AI workloads that consumes zero water for cooling during operations, further reinforcing our commitment to water stewardship by design. This chip-level cooling solution delivers precise zonal temperature control without water evaporation by recirculating water through a closed-loop, direct-to-chip cooling system. As our datacenter fleet continues to expand, the addition of these zero-water designs will further reduce Microsoft’s water use intensity over time.

Datacenter cooling methods, explained:

• Cooling towers: Traditional systems that remove heat by evaporating water year-round.
• Hybrid fluid coolers: Evaporates water for cooling during hot summer conditions and switches to dry mode when ambient temperatures cool down.
• Direct air: Uses outside air for cooling, with little to no water use. Water is used only when outside air is above 85°F.
• Air cooled chillers: Uses mechanical refrigeration and outside air to remove heat from closed coolant loops with zero water evaporation.
• Liquid-cooled AI DCs: Uses closed-loop, direct-to-chip cooling to provide precise chip-level temperature control, removing heat efficiently with zero water evaporation.

Design innovation is only part of the story. We are also improving the efficiency of existing facilities that use water through a continuous focus on optimizing temperature and humidity setpoints, enabling more precise environmental control and eliminating overcooling. In addition, we regularly audit water use and compare actuals against design expectations using real-time weather data and operational analytics. This helps ensure our datacenters are performing as intended and enables us to quickly identify and address any unexpected water use. These efforts, combined with ongoing hardware and operational improvements, are all aimed at using as little water as possible.

Specifically in our Phoenix, Arizona, datacenters, implementation of these advancements led to a 23% year-over-year improvement in WUE in FY25 alone. We are now deploying these advancements across our direct-evaporatively cooled datacenters globally.

Operational improvements like these are one reason Microsoft has been able to report significant long-term reductions in water intensity across datacenter generations. They also point to the next phase of our work: expanding the use of recycled and alternative water sources wherever possible.

In addition to driving efficiency, we also prioritize using recycled, reused or non-potable water wherever possible in our operations. We have expanded the use of these non-potable water sources in some of our most water-intensive regions, helping reduce demand on freshwater supplies. For example, in Quincy, Washington, Singapore and San Antonio, Texas, three of our key locations for advancing water stewardship, we leverage 74%, 99% and 79% recycled, reused or non-potable water sources, respectively.

Rainwater harvesting systems are now operational at select datacenters in the Netherlands, Sweden and Ireland, with additional installations planned in Canada, the United Kingdom, Finland, Italy, South Africa and Austria. To illustrate the potential impact, Microsoft’s new datacenters in Quebec are expected to collect up to 1.5 million liters of rainwater annually, depending on local precipitation levels. This water can be used to further offset the already low water withdrawal at these sites. Expanding the use of alternative water sources in this way helps reduce pressure on municipal water supplies while supporting efficient datacenter operations.

As needed, we implement on-site water treatment systems that enable facilities to recycle water multiple times for cooling operations. These systems produce purified water suitable for reuse within cooling systems, reducing overall dependence on utility water supplies. Together, these efforts demonstrate how engineering innovation and operational excellence can work in concert to meaningfully reduce water use at the facility level.

We work closely with local utilities to reduce strain on community resources, and plan ahead for the sourcing and infrastructure needs associated with our operations. Beyond our operations, Microsoft’s Datacenter Community Pledge commits us to protecting local watersheds, engaging stakeholders and investing in projects that strengthen regional water resilience — helping ensure datacenter growth supports both environmental sustainability and long-term community well-being. Where system improvements are required, Microsoft funds those upgrades in full so communities do not have to shoulder the cost of supporting our operations.

Beyond our own footprint, we invest directly in community water infrastructure by modernizing water systems, expanding access, increasing reliability and helping utilities maintain stable rates and pressure. These investments create shared value for both Microsoft and the local communities we work closely with by strengthening critical infrastructure and supporting long-term water resilience. For example, near our datacenter in Leesburg, Virginia, Microsoft is funding more than $25 million in water and sewer improvements to help ensure that the cost of serving our facilities does not fall on local ratepayers. Since 2020, we have invested more than $500 million in more than 75 water and wastewater infrastructure projects that deliver meaningful community co-benefits.

Finally, we pair all of this work with our broader water positive commitment, our goal to replenish more water than we withdraw by 2030, while advancing our Community-First AI Infrastructure approach, which prioritizes delivering measurable benefits to the communities where we operate. This approach extends across our entire datacenter footprint, including leased facilities. In FY25, we replenished more water than we withdrew globally, marking a significant milestone in our water stewardship journey.**

We prioritize and pursue projects designed to deliver meaningful water contribution to each local community. For example, in the greater Phoenix area and nearby Nevada communities, we partner with FIDO Tech and local utilities to deploy AI-enabled leak-detection that identifies and repairs hidden breaks in aging water systems. By preventing water loss before it occurs, these efforts help keep more water in circulation, improving reliability for residents and effectively increasing the amount of usable water available across the system.

Across the Midwest, we work with The Nature Conservancy to restore historic oxbow wetlands — crescent-shaped water bodies that naturally recharge groundwater, reduce flood risk and enhance habitat for native species. These wetlands act as natural reservoirs, capturing and slowly returning water to local aquifers over time. The goal is to create more stable water availability for agriculture, healthier ecosystems and increased resilience for nearby communities throughout the year.

As water challenges become increasingly complex around the world, Microsoft remains deeply committed to protecting water as a vital natural resource. We continue to advance datacenter innovations that reduce water use intensity while supporting the growing performance demands of cloud and AI services. Through zero-water cooling designs optimized for AI workloads, water reuse initiatives both on and off our campuses and community-focused stewardship programs, we are working toward a future where digital growth and responsible water management go hand in hand.

We are also exploring zonal cooling architectures that more precisely align cooling approaches with the needs of different hardware types, improving efficiency while supporting a diverse mix of AI and traditional workloads.

Datacenters are essential infrastructure for the digital economy, and we believe they should be built and operated in ways that benefit the communities they serve. Over the last decade, we have demonstrated that technological advancement and environmental stewardship can progress together, and we remain committed to continuing that journey as we build the datacenters of the future.

Top image caption: Aerial view of Microsoft datacenter campus in Wisconsin.

Judy Priest leads technology strategy, innovation and research for Microsoft’s global cloud and AI infrastructure, driving advances in datacenter architecture, sustainability, power, cooling, energy and emerging technologies that enable reliable, scalable services.

Steve Solomon is a professional engineer leading the engineering strategy behind the company’s global cloud and AI infrastructure. He specializes in datacenter design, sustainability, power and cooling innovation, helping advance reliable, efficient and community-focused infrastructure at hyperscale.

*Footnote 1: Measured as water withdrawals per megawatt (MW); based on a 2022 baseline.

**Footnote 2: To understand how Microsoft calculates and tracks water replenishment data, please refer to our FY 2024 Environmental Data Fact Sheet.

The post Inside Microsoft’s two-decade push to cut water intensity while scaling for growth appeared first on The Official Microsoft Blog.