Actually, it's wonderful to be back here in Brazil and see all the exciting things happening. In fact, this morning, I already have had a chance to spend some time at our showcase meeting with entrepreneurs and startups, nonprofits, large companies and public sector institutions all already using this next generation platform, with AI and cloud and doing super innovative things. It's always an exciting time. I've been coming to Brazil over the last 30 years of my own Microsoft career. You know, across all the different eras, starting with the PC and client server to the web and the internet, cloud and mobile and now AI. And it's always exciting to see how the rate of diffusion of new technologies and new platforms goes up, non linearly. Right what's happening right now when it comes to AI, it's just there's no real difference between what's happening, perhaps in the West Coast, in the United States, or in Sao Paulo, and it's exciting to be able to sort of see that. So what I want to talk about is this next generation platform that we are living in, in the transition to and ground us in the innovation whenever you think about a new platform, you need to perhaps get a deeper understanding of what's the core driving before driving, in this case, with AI, happens to be scary laws, very Much like Moore's law. These are laws that you see in observation here. If you think about Moore's law, every 18 months you have doubling of what you would see with AI and deep learning in particular is doubling of performance perhaps every six months it started, in fact, all the way back in 2010 right around 2018 when the Transformers had started reflecting again. And you're now seeing the scaling clause really come to the conclusion in a very major way. And it's interesting also, when you think about scaling, it's not just compute scale. It's the combination of compute data and innovation, how you shape data, collect data, generate and synthesize data, as well as algorithms. So the combination of all of them. So for example, from the compute you're getting 100x improvement in performance. So that's the level of abundance that we have amongst that's the scary one. So we need some feedback how it manifests in increasing capability, right? Because it is this very enriched model that are capable of multimodal input and output. You can see, you can give it images in video input, and you can get gaps in images and video output, right? So it's fully multimodal, and you can have these long, full duplex conversations. So when you think about this, the computing the computing, human interfaith, going forward, is going to be completely changed because of these capabilities. The second major capability is the reasoning capability, because ultimately, you know, it's about, you know, if you sort of trace back 70 years of computing history. It's always been about, how can we digitize more artifacts of people, places and things and make sense? Can we get game inside? Can we do predictions? That has been a constant pursuit over the last seven years, and you know, where you've had great new technologies like databases relational algebra, and now we have a new neural algebra, and the neural reasoning leadership that helps us find backwards in digital neutral spaces and things, and helps us gain insight predict things. So you have new reasoning. And then, of course, we can put very good and then, of course, you can put credit purposes AI systems together by giving it memory, giving it context, giving it grounding. So when you bring all of these things together, what we are building, you can describe it as an agentic world, right? You have these AI agents, which you use, both in your personal life and at work, those AI agencies that are working on your behalf, that you trust when men are being working on behalf of your team, your organizational function, like that, your marketing and function
like it can be cross organizational as well. It can be business process overlaid on top of that, because there is going to be a very rich, aging world, or ejected world that we will create. So now, of course, to us in Microsoft, it's never about just technology for technology for us. Ultimately, our focus is to translate this powerful technology into real world impact by living up to our mission to empower every person and every organization on the planet to achieve that translation is everything, and that's where people like yourselves in this room really play a major, major role, because you have to do this in the Real World in one community, one one period of time. And to do so, we are focused on bringing three major platforms. The only three major platforms bring to form this agentic world. The first is copilot, plus agents. Copilot as the UI for AI, because even in this very rich agentic world, guess what, these agents are either working on our behalf. So that means we need to have an interface to it, and agents will need an interface to human activity. And so for us, copilot is the starting man, which is the UI to AI, but you can orchestrate all these agents, starting with a copilot. So copilot plus agents is the first platform we'll talk about. The second is, of course, your ability to build your own copilot, your own agent. So we have a full stack AI platform so that you can build all these rich agents and copilots yourself. And lastly, of course, there is going to be a complete regeneration of devices, starting with a PC, which are going to carry these copilots and agents, and they're going to enable what they describe as hybrid AI, so that you can take advantage of true distribution, computing and AI to be able to really get the benefit. So these are the three platforms. So let's dive starting with copilot processes. We build with copilot a full design system, a design system that starts with a chat interface, right that can get you access to information, whether it's inside your work or in the web. And then you also created a new artifact for pages. And then we also incorporated copilot into all the places you work today, right in Word, Excel and PowerPoint, Outlook and team. In fact, one of the most exciting things that we just did last week was launch of copilot in Excel with Python support, right? So whenever you get a spreadsheet, now it's not just attention. You can think of it as a spreadsheet that comes with a data analyst. Then you can query more facts and more insights on the data that is there in the spreadsheet. It literally creates a new workbook, does its flashback analysis for you, and you'll be instructed, right? So think about the sense of empowerment. I always say when Excel first came into my life is when number sense really changed for me, and I think it's one such moment now, data science is going to be similarly democratized, like what Excel did. Now, there is this other element called copilot studio, which allows you to then orchestrate agents into these workflows with copilot. In fact, copilot plus agents orchestrated by copilot studio is perhaps one of the most powerful things that I think we are going to all start exercises, just like anyone who created an Excel spreadsheet and wrote a formula or did some macro. Think of that level of agency is what we will have in this ejected world to be able to create our own ranges. In fact, one of the things that we do is we take the rich availability of all the data, right? One of the most important databases in any organization is the core communications database, and the database which has all the documents inside of the organization. That's where the internal asset knowledge of an organization is, right? So it's all inside something called the Microsoft graph. You can take that graph, you can then take even this other artist called Dataverse, which has all of the business process data and all the connectors to every line of business application and bring the richness of all that data with the power profile extruder to these agents to be able to automate the next sort of workflows and work artifacts, right? So, in fact, a good example of this, let's say you want to build a field service agent that somebody in the front line wants to use copilot as the user interface with natural language, but it's grounded for field service activity. All I do is go into copilot studio, give it some instructions, give it some knowledge and data. The data is sometimes SharePoint document, in this case, happens to be doing dynamic back end build service. I can wire it all up, give it some actions, and then make it integrate with copilot. And so it's as simple as that. So this is about that low code, no code developer of agents that any one of us in the workforce can be doing to bring the power of data within the agent to a co pilot, to change the expertise now that is really available at the fingertips to a full service technician who is on or visit your customer, for example. So that's your system you built, and in fact, you only receive the productivity gains, right? So it's no longer about waiting for the actual business impact. We are seeing it in marketing, in sales, in finance, in customer service, in every function across the enterprise. The productivity gains are coming fast and furious with this design system that we just talked about. And so now we already have tremendous amount of customers who are using this right here in Brazil and across all industries. So with finance in every industry possible, people are using this image, but this morning, I had a chance to meet a math teacher and his student, who used copilot studio to build an application to be able to not just in operation, but also essentially create a personal change of Whatever student in this room, right? That was a fantastic example of how people inside an organization can change how the organization works through the power of these digital technologies, because that's what it's not about just, you know, a couple of people who have access to this technology. It's about everyone inside the organization having access to technology change how your organization works to drive that external outcome in private that's exciting to see. Now, last week, we were very thrilled to announce what I think is the next phase of copilot. In this next phase of copilot, one of the core things that we've done is bring all the web's knowledge, all of the work's knowledge, and then also created for the fullest time and AI, first work artifact, right? If you think about it, every time there has been a platform shift, it has come with a new artifact, an artifact like, for example, when Windows was first created, we had word except PowerPoint and platform. And then when the web happened, we were able to collaborate in real time in all of those documents and spreadsheets. And now, with the pages, we have the ability to take where they are and collaborate with other units inside of our workforce. So that's what pages enable. So this web plus work plus pages, I think, is going to redefine, very much like back in the 90s, how the Office Suite redefine Work, Work artifacts and workflow. We take this web plus work, business chat with web plus work and pages is going to redefine work and work artifacts and workflows. Going forward, to show you this in action, I want to invite up on stage my colleague, Kelly gone up I all.
Thank you, Satya, I've been using co pilot in my day to day work for over a year, and it's remarkable how it's transformed how I tackle workflow. Personally, I use it for things like summarizing information, jumpstarting the project, or finding that one mode file, even how I prepare for customer meeting here, can give me up to speed in phones co pilot can
give me up to speed in just I simply asked. I asked propeller to recap my meetings that I had, that I prepared for this call, and I asked it to help me build an agenda within moments, finds my continuous meeting and builds an agenda with action item based off outstanding questions from that call. Now this is just a fraction of what propelled can do today. I want to show you the more powerful, collaborative experiences. Let's be in action. So let's say if you were in sales at an EV charging company, and those are your business opportunities in the city. There's a new business opportunity we've never been to before, to start using some basic information about LGBT and because procurements drowned in web data, it's performing music, programming and pull syllable information is a great start to answer your question. Now you want to turn this simple response into a page. Here. You can keep looking at your phone with colleagues or with so many you can even share a page just like we share the workshop. Now this research section is looking a little light, so tagging your colleagues to help build an outcome once you do evidence, immediately starts with review. Now, while your colleagues on those latest interviews, you continue to do research build out the business case, what we just saw was kind of bringing in information from the web, putting it on a canvas and enabling you to collaborate with your coworkers. But what happens when you put in your work data, when I'm talking about your emails, chats, documents, meetings and more, that's where pages really comes to life. Now the first thing we're going to do is pivot from Web's tour. Can you take started on a plan for your proposal, but you want to be sufficient and do some existing work. So who else propelled to record the biggest plan and start on this one? Propelled this will deliver rules work by veterans and owners and formats at all. Now you're ready to collaborate with your quality plan. Here you can see coworkers already updating the workout schedule and finding comments. Now let's get to work on the proposal outline. As a first step, you need to understand what your customer is looking for. Your coworker had an interdisciplinary meeting with the customer last week. So right here in pages, they have referenced the recording of that meeting and insert a bulleted list requirements from that call. In just seconds, people is able to simplify the discussion and give your team a TTA. Now that you understand what it's going to take to win the business, you're ready to start building that proposal last year, you similar deal, and you want to reuse that great work. So you ask Copilot to reference the proposal as a starting point and update it to the new customer requirements. With one last click, co pilot pulls it all together for you. This is truly a new pattern. Source bringers get our content from web and work in an economic specific habit combined with human to human AI collaboration. Back
to you. Thank you, Shelly, but what really wonderful to see this, in fact, when I think about doing my daily habits now, how much it has changed with pages, right? I mean, chat has already changed, essentially, how we work on this. We go to it all the time, even when I'm thinking about sort of any topic and I want something summarized across all my documents, all my web, your channel, but now you start creating these new artifacts and features because you want to collaborate in real time with people and AI together. Think with AI, collaborate with people is the new workflow at work. Now let's move to this next big platform, which is the full stack AI development platform to build your own co pilot with your own agents. You know to us, it starts, of course, with the massive investment in the core infrastructure with Azure. We now have 60 plus regions around the world. We have across data centers. We're building these intelligent factories essentially everywhere. Right here in Brazil, we have two regions already, and it's exciting to see the growth of these regions and how this innovation we're just talking about. And today, I am really, really excited to announce that we are doubling down. We are going to invest 14 point 7 billion AI in Brazil. In fact, my honorable
vice president of Brazil is here and is going to be speaking after me to really talk about how all of this ultimately, is about economic growth and social benefits in Brazil because of all of these investments that are being made, ultimately to be able to translate that to that very large material to have the vice president deal with us now, in terms of the core negative investment community, it's about bringing the best performance of AI to every organization, every developer and every user of AI. It starts with silicon innovation. Silicon innovation, whether it's coming from our partners in India or our own silicon investors in mind, so we're going to do the best civic but on top of that, all we are doing,
from science systems work right when people AI this workload, whether it's agents or whether it's training, is unlike anything else, right? These are data parallel, synchronous workloads where you optimize everything from the network to the storage to the accelerator and memory access into reviews of fantastic systems work that is extremely tons of things. Work that is extremely tough to fit for delivering this AI tokens for you both when you run free trading runs, already doing eight runs, and so that's, I think, one of the biggest use cases of innovation. The second of course, is right next to your AI stack for training or inventory. You have your data stack because ultimately, your AI is basically taking all the data and using it to train and or using it to ground your image. And so we have to obtain all the data to have nexus with your new AI compute. Now that means you want to bring everything from Oracle database or one end to snowflake or other end to support everything in Azure. But one of the most exciting areas of innovation for us is building a new AI stack for operational stores, whether it's cosmos or SQL or Postgres or new vector DBS and Azure Search. What we're building in AI first analytic stack with fabric so the rich data layer, coupled with the innovation the AI infrastructure are the core building blocks for anyone trying to build AI applications now we then have the best tool chain, right? So you have your AI infrastructure, you have your data. So the next thing as a consideration for anybody building an AI application is about having a fantastic tool chain. It starts with the app services, right? Where you need an app platform. In fact, AI doesn't stand alone, right? You need application services that you always used in the cloud to be right there with you when you're thinking about AI. So we have App Services, we have logic apps, we have Azure functions, we have AKs, which is the container service. So that's the first grid. And next to this is a new AI platform that's getting created, which is all about Azure. Ai. You want to be able to use new evaluation services across all models. You want to be able to ground these models using hybrid search. So all of those capabilities come by default in Azure AI. And of course, you have the model catalog and models as a service. And talking about models, Azure has the broadest set of models available. Of course, they are all the unique open AI models, right? This whole one model that's just come up is pretty exciting with its reasoning capability. But we also have all open source models, whether it's from llama or misfrom or even other closed source models from this electro here. So you have a full five selection of models that you can then use to build your applications. And then, outside of history, is the tool chip. Right by the clock from 1975 to 2024 is focused on one thing is building the best tool chip. And for us, the combination of VS code and GitHub. And now get out with GitHub co pilot has been absolutely changed. So we are very, very excited to bring the best tool chain to everybody building these AI applications right now, right here in Brazil, we are close to 5 million GitHub developers, and it's fantastic. It's the second largest, or soon to be second largest community of developers, and it's fantastic to see the progress you can get out right if I get on copilot now has this new feature that we are enabling to even o1 where you can code optimize. So for example, if you have some kind of an encoding algorithm, you can use Omar to go in and say, optimize that algorithm, right? So it's no longer just continuation, but it's even optimization. So it's very, very good to see the progress these products are making, and the five plus million developers here using all these products, and we have many customers already here, partners already using these products, I had this morning a chance to meet with folks from Alban Einstein hospital. They've been working with us for many years, but it's great to see them now take advantage of even all the AI work on top of all the data work they did, because one of the foundational contributions they're making is to make sure that Brazilian health data is really represented well so that global health, you know, and all the drug regimes and drug discoveries grounded in the data that represents all of the Brazilian health records. I also had a chance to meet with professor who built a complete new agent of theirs, which is powering operations, everything from customer service to their internal employee operations. In fact, I met the governor of the central bank here who's just unwilling to visionary when it comes to thinking about the financial stack, everything from the payments to finance or programmable finance to, in fact, what they're doing with cdbc and digital currencies. It's just exciting to see the vision that he has to ensure that Brazil has the most advanced system when it comes to even reusing the transaction costs for every everything that happens in the economy. And then I had a chance to meet three women entrepreneurs who started this company radar faith, which is exciting to see three women start a company in one of the most important sectors, which is about health and wellness and the progress they're making. Let's roll the video and give
you a feel for it. Brazil,
personal, personal,
happy This
Coffee
could
education,
services, what does this?
It's really wonderful to see that article.
It's really wonderful to see that these gaming missions and social all brought together with AI create new products that I think that's the type of innovation that I think is going to have real impact. So that brings us to the third cluster, these copilot plus agent devices, starting with the copilot plus PC, it's really exciting for us to take what we have done in the cloud of making a complete stack that is for AI to do exactly the same thing when it comes to edge devices, right starting right into silicon innovation without partners, Whether it's Qualcomm or AMD and retail or others who all want to bring some of the best silicon innovation in perhaps 30 plus years to the PC. It's really exciting to see when you when you use any AI application now on copilot plus PC, it's really using the 45 plus costs that is available in the NPU, and that means your CPU and your GPU are left free for everything else. So you can be laid on it with your GPU Max, and you can have your AI assistant on the NPU system architecture for the next generation of devices. And of course, on top of this runtime called hyper AI, where you can now take the power of the cloud and combine it and compose it with the power of the client or the edge. That is interesting. These are not about two separate things. It's about two things coming together in support of your application, because you want the best of both. So that's what our development platform really enables. And we're excited to see where the innovation already happened, whether it's from Microsoft or the third party that I'm able to build. To give you a feel for some of the Innovation in copilot postpies as well, or Video you
it's wonderful to have you know better back to life, better performance and of course, co pilot, co create, and and live captions all built into these new devices. And it sets the stage, I think, for what is going to be next generation of device innovation and application innovation, and top of these ranking mobile devices. So that's the three platforms that we're building. But I want to close out by talking about, perhaps, what is the most important considerations across all three of these platforms, which is trust, and in particular, in today's age trustworthy. AI, right to me, trust is all about having consistency over time, and it starts with commitments that we make. It starts with impact the security What started with our security initiative in November of last year. In fact, just this week, we reported the progress we made. This is something that we plan to do every quarter. We take this as some of the most important imperative Microsoft inside the company, the security of all else. And so that's what we intend to do, see us report the progress on it. The next is privacy, especially the age of AI, it's very important for us to stand behind nothing. Your data is your data. We don't train our model of your data, and you can be assured that there is no leakage of information. If you bring a copilot in and use it inside your organization, you're able to use that reasoning engine on your data, but that reasoning engine is not taking your data out. So these communities are on privacy matter, and we're making them very strong. And then, of course, it's about responsive to AI. Ultimately, when you have all these rich models in every application, you're going to use multiple models, you want to have guardrails around them. You want to ensure that there's no bias. You want to have a fantastic email system to be able to evaluate for safety before. And so that's we're doing, which is just concept principles for safety, but we want to really ensure that we are taking AI safety as an end to end system. That's our commitment. And through those we are launching new capabilities just this week. For example, we launched a new confidential computing we have in Azure to help with that privacy community. We are launching new APIs for correctness, which is around AI safety. So in Azure, AI now we can use AI to correct AI, right? That's the correctness roundedness. API, how we continually innovate on watermarking in many, many other areas, so that you have the ability to use AI in a trustworthy way. And ultimately, this is about building towards that mission that I started with, which is empowering every person and every organization on the planet to achieve more, and that achieving more is grounded in creating that economic opportunity, expanding economic opportunity, earning trust, protecting fundamental rights, and doing it all while in A sustainable way. That's the four pillars that cut across our mission. And of course, for us, it starts by really empowering every person, every organization, one country at a time. So that means right here in Brazil, and today, I'm so pleased to announce that we are making a commitment to training and up 5 million people in Brazil. Because we think that the most important passage is,
you know, ultimately, when we have both these intelligence factories and the skills right here in Brazil, I think what you all will be able to do in education, in healthcare, in industry, in public sector, is going to drive the next level of social and economic growth in Brazil. And I can't wait to see all the impact you're going to have, and you already see some of the early signs of it. So I leave you with a video montage of all the impact you're having right here in Brazil. Thank you all very, very much. In the
new era, one where access to information is access to expertise, from the farm to the lab, from the boardroom to the classroom, this new generation of AI is for everyone, everywhere. Now, anyone can save time with a personal assistant. What
used to take months. I've taken me
business of the belt. I'm saving about 50% of time, and that's time that I can use to do other innovative things.
Anyone can access a personal tutor to learn new skills. I think the
scenario has a potential to completely reimagine the way everything completely reimagine the way every single student learns in the world.
This is a new way to analyze with a personal coach,
I can teach back to you not supposed to be AI is unlocking
AI is unlocking creativity for us, all descriptions of social details
