Delivering 3 trusted platforms for the AI age

The following is adapted from my remarks at Microsoft Ignite this morning.?

With every platform shift, it is good to understand the context and underlying forces. And today it's scaling laws. With Moore’s law, we saw a doubling in performance every 18 months. With AI, we have started to see performance double roughly every six months.??

There is a lot of debate about whether we've hit the wall with scaling laws. At the end of the day, these aren’t?physical laws but empirical observations that hold true over long periods of time. So it's valuable to debate and have that motivation for innovation in model and system architectures as well as data regimes. In that context, if anything, we are seeing the emergence of a new scaling law with test-time compute, which OpenAI's o1 and Copilot's "Think Deeper" feature leverage to solve harder problems.?

Ultimately, these breakthroughs manifest in three capabilities:?

The first is a new universal interface that is multimodal. It supports speech, images, and videos both as input and output.?Second, we have new reasoning and planning capabilities. We essentially have a new "neural" algebra to approach complex problems. We can detect patterns and make predictions involving people, places, and things, as well as interactions between them.?And third, we now have the capability to support long-term memory, rich context, and teach these models how to use tools.?

When we put all of this together, we can build a rich agentic world defined by a tapestry of AI agents that can act on our behalf across work and life, across teams, business processes, and organizations.?

We are going to celebrate a lot of technology over the next multiple days, but it’s also worth reflecting on the words of philosopher John Haugeland. "The trouble with artificial intelligence,” he said, “is that computers don’t give a damn.”?

But we do!?

And that’s really what grounds us. Amidst all this rapid change, we remain focused on our mission to empower every person and every organization on the planet to achieve more.?

Today I want to focus on AI and business transformation, and how it drives growth and improves efficiency and operating leverage.

To that end, we are building three platforms: Copilot, Copilot devices, and the Copilot and AI stack.?

Security

Before I dive in, I want to talk about something we are prioritizing above all else: Security.?

At this conference, you will hear us talk a lot about our Secure Future Initiative , as we implement our principles of secure by design, secure by default, and secure operations. Most importantly, we are committed to continuous improvement, because our work is never done. After all, we are only as good as our ability to defend against the next novel attack.?

At the end of the day, we recognize that security is a team sport. It’s why we partner broadly with the security community, and why today we announced Zero Day Quest ., a new hacking event with $4 million in rewards focused on securing cloud and AI—the highest of any public hacking event in the industry. The quest starts today and will culminate in an in-person hacking event next year.?

Copilot

Now, let’s talk about Copilot. Copilot is the UI for AI.

It‘s rapidly becoming an organizing layer for work and how work gets done. Every employee will have a Copilot that knows them and their work, helping them unlock productivity, enhance creativity, and save time.?

Copilot Studio will allow you to create agents that automate business processes. And every IT department will have a control system to manage, secure, and measure impact.?

Over the past year, we have seen incredible momentum. What “lean” did for manufacturing, AI will do for knowledge work. It’s all about increasing value and reducing waste.?

And we’re just getting started. We are continuously innovating, shipping hundreds of updates since we made Copilot generally available a year ago. That includes real work on fundamentals like quality. Copilot responses are now more than two times faster on average, and response satisfaction has improved nearly three-fold.?

Going forward, we think about Copilot adoption in three ways: The first is adoption. The more employees use Copilot and Copilot agents, the faster they realize value. The second aspect is extensibility. Copilot is a platform you can extend with agents that scale what you can do. And the third is?measuring ROI.?

Let’s start with Copilot Pages. Pages is the first artifact of the AI age.

I use Pages myself to prep for customer meetings. Copilot can comb through everything about the customer from the web, LinkedIn, biz apps, and work artifacts like emails, documents, and presentations. I can put all of that into a Page and share it, including with the account team.?And then it becomes the first draft of meeting notes that are shared with everyone involved in real-time.?

Copilot is also deeply integrated across Microsoft 365. In Teams, Copilot can reason over past meeting chats and transcripts to get you quickly up to speed. With Teams screen understanding, Copilot can even answer questions about presentations and documents shared during meetings.?

In Word, Copilot creates a draft based on other Word documents, PowerPoint decks, PDFs, emails, and meetings. You never have to start with a blank page.?

In PowerPoint, you enter a prompt about what you want your presentation to be about, and Narrative Builder generates an outline. You can edit and refine its suggestions and create a first draft of the complete presentation.?

And in Outlook, prioritize my inbox is a total game changer. It helps you quickly get to the messages that matter, analyzing your inbox based on both the content of the email and the role of the sender.?

And I think what we are doing in Excel is fantastic. Take a look:?

Now, let’s talk about how to extend Copilot. Today, I was very excited to announce Copilot Actions. With Actions, you can use Copilot to reduce the amount of repetitive, everyday tasks that you do. It is like Outlook rules for the AI era. Actions are a simple but powerful way for you to scale the work you do, and they work across the entire Microsoft 365 system.?

We also introduced new agents you can use with your team. You can think of them as teammates?scoped to specific roles and permissions. You can add a Facilitator Agent to your Teams meeting to help you stay focused, moderate discussions, and follow up on action items; our Project Manager Agent in Planner helps automate key steps in the project management workflow; and our Employee Self-Service Agent?provides useful information and answers to policy questions at your company.?

We are also announcing SharePoint agents. Every SharePoint site now has a built-in agent. These agents provide instant access to real-time information and insights from your knowledge base in the flow of your work.?

We are also giving you the ability to easily create your own agents with Copilot Studio. Sometimes we mystify AI and agentic creation. But it’s as simple as creating a Word doc, PowerPoint slide, or an Excel spreadsheet. ?

You can also make these agents autonomous using Copilot Studio. And they can always raise an exception in Copilot for input. After all, at the end of the day, even autonomous agents need a UI to interface with us. ?

And of course, when we talk about extensibility, that includes changing how you interact with your bespoke business applications, too. We are excited to share that many of our partners have built their own agents and connectors in Microsoft 365 Copilot. This includes Adobe, LinkedIn, SAP, ServiceNow, Workday, and even new AI-first agents built by companies like Cohere.?

Here it is in action:

Now, let’s talk about measurement. After you adopt Copilot and extend it with agents, you need to measure the ROI. And to do that, today we introduced Copilot Analytics .?

Our goal is to show how Copilot usage is directly translating to business outcomes across Sales, Marketing, Finance, and more.??

Copilot devices

This AI era will be defined not just by what’s happening in the cloud, but also what’s happening at the edge. ?

Over the past year, we have introduced an entirely new class of Windows PCs designed to unleash the power of distributed AI across cloud and edge. We call these Copilot+ PCs .?

We are working across our entire ecosystem—including AMD, Intel, and Qualcomm, along with every major OEM—to bring these to life, with silicon capable of an incredible 40-plus TOPs. When it comes to fundamentals, they are the fastest PCs on the market and include all-day battery life. ?

We are also delivering entirely new endpoints. Three years ago, we introduced the Cloud PC category with Windows 365, which securely streams your personal Windows desktop from the cloud to any device, whether it runs iOS, Android, or is a mixed reality headset like the Meta Quest.?

We have seen great momentum and adoption as we bring Cloud PCs to remote and temporary workers, IT, developers, frontline employees, and for disaster recovery. In fact, it is my go-to dev desktop, equipped with full access to GitHub, VSCode, Azure SDK, CLIs—all set up to use on any of my devices.?

Today, I am excited to announce the next step with Windows 365 Link :?

It’s a simple, secure, purpose-built device for Windows 365. It is admin-less and password-less. And security configurations are enabled by default and cannot be turned off.?

Form factors like Windows 365 Link give you yet another choice for secure Windows endpoints.?We recognize how mission-critical Windows is, and we are committed to security and resiliency as first-class priorities.?And today we announced?the Windows Resiliency Initiative , as we double down on our commitment to make Windows secure and reliable for customers and their mission-critical workloads. As part of this work, we are making changes to OS access, introducing new features, and, in partnership with our ecosystem, establishing new guidelines for Safe Deployment Practices.?

Copilot & AI stack?

We are also exposing every layer of our tech stack so you can build your own copilots and agents.

Infrastructure?

At the infrastructure layer, we are building Azure as the world’s computer. Over the past year, we have made datacenter investments in 15 countries across six continents. We now have 60-plus datacenter regions—more than any other provider.?

And we are not stopping there. Today, we went further, announcing Azure Local .?

Enabled by Azure Arc, Azure Local extends Azure services across hybrid, multi-cloud, and edge locations, with a central control plane. It brings Azure services to customers' distributed locations, whether in retail, hospitality, or industrial manufacturing, so they can run mission-critical workloads across cloud and edge.?

When it comes to silicon, our Cobalt 100 VMs became generally available last month. Customers and partners like Databricks, Siemens, and Snowflake are seeing up to 50 percent improvements in price-performance with these VMs. Our own media processing capabilities in Teams now run on Cobalt 100 as well.?

Our approach to silicon includes a deep commitment to security, and today we announced our first new in-house security chip, Azure Integrated HSM .?

It is a dedicated hardware security module that hardens key management, meaning encryption and key signing can remain within the bounds of the device without compromising performance or security. Starting next year, it will be part of every new server deployed on Azure, enhancing security for both confidential and general-purpose VMs and containers.?

Our goal is to remove any bottlenecks that stand in the way—whether it is performance, latency, or resource constraints. That is why today we are expanding Azure Boost with our first in-house DPU.?

DPUs are processors specifically architected to accelerate data-centric workloads, absorbing multiple components of a traditional server into a single piece of silicon. It runs cloud storage workloads at three times less power and four times the performance.?

When it comes to AI, we are building new intelligence factories for the world. Tokens per watt plus dollars?is the new currency for performance. It is all about maximizing value in the most efficient way.?

The pace of innovation across the industry is phenomenal, and we are working with our partners to bring you more choice across cost and performance. ?

Today, we announced the preview of NVIDIA Blackwell AI infrastructure on Azure. With 72 Blackwell GPUs in a single NVLink Domain and InfiniBand in the backend, these racks are optimized for the most cutting-edge and high-scale AI workloads.?

We are also working closely with AMD. We were the first cloud to offer VMs powered by AMD’s MI300X GPU, and we are using this infrastructure to power Azure OpenAI Service, too. ?

And today, we are introducing Azure HBv5 , co-engineered with AMD. It is up to eight times faster than any other cloud VM, setting a new standard for high-performance computing. It will be generally available next year.?

Data?

Now, let’s move to data. There is no AI without data. For you to build your AI apps, you need to rendezvous your data with your AI compute effectively. We are building out the full data estate to do just that.?

At the core of our platform is Microsoft Fabric. It brings together all your data, as well as your analytics workloads, into one unified experience. With OneLake in Fabric, it is easy to unify your data no matter where it lives—in Azure, on premises, or even in AWS or GCP—to create the data layer for your AI workloads.?

We are very excited to announce that we are bringing our flagship operational database, SQL Server, natively to Fabric, with Microsoft Fabric Databases .?

Fabric simplified every aspect of an organization’s analytics needs. We want to do the same for operational databases. With Microsoft Fabric, customers now have an enterprise data platform that serves the full range of use cases—from batch data to real-time, to massive transactional performance—all in a unified product.?

Azure AI Foundry?

Now, let’s talk about building your AI applications. Every new generation of apps has brought a changing set of needs. Just like web, mobile, and cloud led to new app platforms, AI is transforming how we design, customize, and manage apps today.?

That is why we are building a first-class app server for the AI age. We call it Azure AI Foundry .?

With Foundry, we are unifying all our models, tooling, safety, and monitoring solutions into a single experience integrated with the world’s most popular developer tools, available as a standalone SDK and as a portal.?

It all starts with our approach to models. We know models and model choice sit at the core of every use case. We want to help you choose the right model for the right job. In fact, we now have over 1,800 models in our catalog. ?

OpenAI continues to set the pace for model innovation, and all of their latest models are available on Azure. We also support open-source models from Meta and Mistral, as well as providers like Cohere. New categories of models are also emerging. In fact, we have added over 20 industry models designed for more specialized use cases from partners like Bayer, Paige, Rockwell, Siemens, Sight Machine, and others.?

With all this model diversity, it has never been more important to have the right tools to choose models for the job at hand. With Foundry, we are adding model experimentation capabilities, meaning for the first time you will be able to experiment with several of these models to compare outcomes and choose the best model that works for you.?

We’re also excited to introduce a new service designed to simplify the creation of AI-powered agents with Foundry. Our new Agent Service helps developers build, deploy, and scale AI apps that automate business processes.?

While anyone can go to Copilot Studio to build agents with just a few clicks, developers who need more control can go directly to the Agent Service and build agents code-first, grounded with data wherever it lives—be it public data from the web, enterprise data in SharePoint, or leveraging Fabric’s OneLake to unify your data estate across your apps and cloud providers. These agents can take action across any system using our 1,200-plus Logic Apps connectors.?

We also know that the multi-agent space is evolving quickly, so we are ensuring our early multi-agent frameworks can all be used within Foundry Agent Service.?

AI apps have specific operational considerations, too, and organizations also need a centralized way to manage AI costs, performance, safety, and security. That is why I was excited to share that we are also bringing new management capabilities to Foundry. For example, we are introducing AI Reports, which developers can use to document and share their application's use case and evaluation results.?

Safety is the most important feature of any AI application. We continue to ensure we have the best tools to build safe and secure AI apps, including Prompt Shields that detect and block the manipulation of outputs that could harm your business's content.?

Here's AI Foundry in action:

Close

I’ll end where I started. It all comes back to our mission: to empower every person and every organization on the planet to achieve more, one community, one country at a time.?

As we enter this new age of AI, it is up to us to empower human achievement. That is why, over the past year, we have helped train and certify over 23 million people in AI and digital skills. And we are committed to working to help millions more. Because learning these skills can change people’s lives. In fact, it already has.??

?