Satya Nadella – Microsoft’s AGI Plan & Quantum Breakthrough
Estimated read time: 1:20
Summary
In a recent podcast, Microsoft CEO Satya Nadella discusses various technological advancements, including breakthroughs in AGI and quantum computing. Reflecting on past tech trends, he explores lessons learned and the future potential of AI and quantum computing. Nadella shares insights on the scalability of AI infrastructure, the integration of AI into business, and Microsoft's strategic investments. He also expresses excitement about Microsoft’s quantum breakthrough using Majorana zero modes, which could lead to the first utility-scale quantum computer. Throughout the conversation, he emphasizes the importance of aligning technology advancements with societal trust and economic growth.
Highlights
Satya Nadella discusses Microsoft's dual breakthroughs in AGI and quantum computing with Dwarkesh Patel. 🎉
Reflects on the evolution of tech trends and Microsoft's strategic decisions over the decades. 💡
Explains how scalability in AI infrastructure is a key to meeting growing global demands. 🏗️
Details on how Microsoft's Majorana zero modes are a potential game-changer in quantum computing. ⚛️
Importance of aligning tech developments with societal needs and legal frameworks. ⚖️
Key Takeaways
Microsoft's AGI advancements and quantum breakthroughs are set to transform the tech landscape. 🚀
Satya Nadella reflects on past tech trends, highlighting valuable lessons for future innovations. 📚
AI infrastructure's scalability is crucial for meeting future demands. 📈
Microsoft's Majorana zero modes could usher in the era of utility-scale quantum computing. 🖥️
Technological advancements must align with societal trust and economic growth for maximum impact. 🌍
Overview
Satya Nadella joined Dwarkesh Patel’s podcast to discuss significant breakthroughs at Microsoft, particularly in the realms of AGI and quantum computing. These advancements mark a pivotal moment in tech, promising to reshape digital landscapes by leveraging historical tech lessons and strategic foresight. Nadella’s insights into the scalability of AI infrastructure underscore its growing importance in an increasingly AI-reliant world.
A major highlight of the discussion was Microsoft's development of Majorana zero modes, hailed as a potential 'transistor moment' for quantum computing. This breakthrough could pave the way for utility-scale quantum computers, revolutionizing industries reliant on computational power. Nadella emphasized the importance of continuing down this groundbreaking path to deliver on its full potential.
Nadella also stressed that technology must advance hand-in-hand with societal trust and benefit. As new digital paradigms emerge, aligning them with legal norms and public trust is crucial. His conversation with Patel covered not only the technical aspects but also the economic and societal implications, promising an era where technology and humanity can grow simultaneously.
Chapters
00:00 - 00:30: Introduction The introduction chapter sets the foundation of the book by providing an overview of the main themes and objectives. It outlines the relevance and importance of the topics to be discussed, highlighting key questions and problems the book aims to address. Additionally, it offers a brief glimpse into the structure of the book, guiding the reader on what to expect in subsequent chapters. This chapter establishes the context and background necessary for understanding the detailed discussions that are to follow.
00:30 - 01:00: Microsoft's Recent Breakthroughs In this chapter, Satya Nadella discusses Microsoft's recent two major breakthroughs, which were significant enough to be featured in Nature on the same day. The breakthroughs include the development of the Majorana zero chip and advancements in world human action models.
01:00 - 02:00: Early Years in Tech and Lessons Learned The chapter titled 'Early Years in Tech and Lessons Learned' reflects on the technological developments and recurring patterns observed from the 1980s to the present. The speaker expresses excitement about noticing similar dynamics occurring again today, which takes them back to their initial years in the technology field. The initial part of the conversation also contains an appreciation for the host's podcast, noting the broad range of topics covered.
02:00 - 03:00: The Evolution of Client-Server Architecture The chapter discusses the historical context of client-server architecture evolution, focusing on debates from the 1990s. It highlights the technological debates between RISC (Reduced Instruction Set Computer) and CISC (Complex Instruction Set Computer), as well as the feasibility of building servers using x86 architecture. It also references the early days of Microsoft Windows NT and the development of a comprehensive technology stack from core silicon platforms to operating systems and application tiers.
03:00 - 04:00: Web's Impact on Technology and Business Models This chapter discusses the significant impact the web has had on technology and business models. It begins with a reference to ongoing litigations which have been part of the technological evolution. The transition to cloud computing and distributed systems is highlighted as a major turning point that changed traditional client-server models. The chapter suggests that the current changes in technology, likely driven by the web, are more comprehensive, affecting the full stack of technology and business models more thoroughly than any previous shifts witnessed by the narrator or others in the field. The chapter also encourages readers to consider various decision-making processes influenced by these technological advancements.
04:00 - 05:00: Long-term Trends and Secular Changes This chapter explores the differences between lasting technological trends and temporary ones, with a focus on the winners and losers of the 80s and 90s. It includes an analysis of Sun Microsystems' experience during the 90s dotcom bubble, discussing the concept of the data center build-out bubble, and its long-term impact demonstrated by the current prevalence of the Internet. The chapter aims to distinguish between secular (long-term, lasting) trends and ephemeral (short-lived) ones, setting the foundation for understanding what developments may stand the test of time.
05:00 - 06:00: The Cloud and AI: Market Predictions The chapter discusses the evolution of computing technologies, focusing on the transition from client-server architectures to modern cloud computing. It highlights the introduction of the graphical user interface and x86 architecture, which enabled the development of servers. Additionally, it recounts an experience from a developer conference in 1991 where Microsoft introduced the Win32 interface, signaling significant advancements in software development.
06:00 - 07:00: Hyperscalers and AI Workloads The chapter discusses the transition to x86 servers and the parallels with client-side computing. It highlights the secular trend towards centralizing computing power, predicting that what happened with client-side applications would occur on the server side as well. This led to the development of client-server applications, providing a clear model for application development.
07:00 - 08:00: Market Dynamics and Winner-Take-All Models The chapter discusses the impact of the web on market dynamics and the emergence of winner-take-all models. It mentions the launch of the Netscape or Mosaic browser in 1993 by Andreessen and his team, highlighting it as a significant game-changer. This development occurred just as Microsoft was gearing up for success in the client-server wave, signaling their impending victory in that sector as well.
08:00 - 09:00: Open Source and Model Diversity The chapter discusses the shift to the browser as a new application model. It describes the adaptations made to embrace this change, including integrating HTML into Word, developing a proprietary browser, and creating a web server to compete in the new landscape.
09:00 - 10:00: Consumer Markets and Network Effects The chapter discusses the often overlooked potential of business models on the web, particularly highlighting the rise of the search engine model as pioneered by Google. It emphasizes the importance of not just identifying technological trends, but also understanding where value creation will occur in the market. This is exemplified by Google's success in organizing the web through search, a path that others did not foresee as prominently.
10:00 - 11:00: Cloud Commoditization and Future of AI The chapter discusses the challenges and opportunities presented by shifts in business models and technology trends, specifically focusing on the commoditization of cloud services and the future of AI. It highlights the critical areas where value will be created in AI, emphasizing that there are two significant opportunities, particularly for successful hyperscalers.
11:00 - 12:00: Predicting Future Revenue Growth In the chapter titled 'Predicting Future Revenue Growth,' the discussion centers around the critical importance of computing power in determining the success of AI systems. It emphasizes that greater computing capabilities translate to a significant advantage. It highlights how technological advancements, particularly in GPUs and AI accelerators, are essential in scaling and enhancing AI workloads, such as those in ChatGPT. This inevitably impacts revenue growth prospects, as entities capable of handling extensive compute tasks position themselves as leaders in the AI field.
12:00 - 13:00: The Role of AI in Economic Growth The chapter discusses the increasing role of AI in bolstering economic growth. It highlights the expanding need for infrastructure that can support the voracious demand for computational power due to AI workloads. The conversation points out that AI systems require significant compute resources not only for their training phase but also during their testing phase. As AI agents become more prevalent, the economic systems must adapt to these increasing computational demands, thereby driving further growth in the underlying technological infrastructure.
13:00 - 14:00: The Balance Between Supply and Demand In this chapter titled 'The Balance Between Supply and Demand,' the discussion revolves around the impact of AI agents on compute usage. It highlights how AI isn't limited by a single user running a single program. Instead, one user can trigger numerous programs, massively increasing demand for computational resources. This shift is expected to significantly expand hyperscale businesses, such as Azure, and alter the competitive landscape, potentially leading to a winner-takes-all scenario.
14:00 - 15:00: Investment in AI and Return Expectations The chapter delves into the dynamics of investing in AI, emphasizing the importance of accurately identifying winner-take-all markets versus those that aren't. It highlights a personal anecdote regarding Azure and Amazon, illustrating early skepticism faced when entering markets dominated by major players like Amazon, with outsiders believing it was impossible to catch up. This reflects the broader theme of discerning market potential and investment returns in AI.
15:00 - 16:00: Compute Infrastructure and Scalability The chapter discusses the nature of hyperscale infrastructure in the context of competition and market dynamics. It contrasts consumer markets, which can be winner-take-all, with corporate and enterprise markets that prefer multiple suppliers for their compute infrastructure. This preference is attributed to the informed nature of buyers like IT departments, who strategically choose to avoid dependence on a single supplier.
16:00 - 17:00: Global Expansion of AI Capabilities The chapter discusses the future of artificial intelligence with a focus on the dual presence of open-source and closed-source models. It draws parallels to the software industry, particularly mentioning how closed-source operating systems like Windows have inspired the growth of open-source complements. This balance between open and closed sources acts as a regulatory mechanism, ensuring that open-source alternatives are always available alongside closed-source ones. The chapter implies that such dynamics will shape the global expansion of AI capabilities.
17:00 - 18:00: Role of AI in Knowledge Work The chapter discusses the potential impacts of artificial intelligence on knowledge work, with a focus on mitigating the closed-source and winner-take-all market dynamics. The conversation suggests that if AI technology proves to be as powerful as anticipated, governmental intervention may occur to regulate its effects. It contemplates a non-monopolistic future for AI development that fosters a competitive and fair marketplace.
18:00 - 19:00: AI and Change Management in Organizations This chapter discusses the impact of AI, using ChatGPT as a key example, on change management within organizations. It highlights the concept of 'winner-take-all' effects due to network advantages, where successful AI applications like ChatGPT can achieve significant market dominance and consumer recognition early on. The mention of ChatGPT's consistent high ranking in the App Store exemplifies its rapid adoption and integration into everyday consumer use, showcasing the competitive advantage gained by early market leaders in AI technology.
19:00 - 20:00: New Workflows Enabled by AI The chapter titled 'New Workflows Enabled by AI' explores the evolving landscape of AI in various sectors. It discusses how AI can transform workflows by providing competitive advantages, particularly in consumer and enterprise markets. Different categories may see different winners, as AI models might become commoditized over time. The analogy is made with the cloud's evolution over the past decades, transitioning from merely being a chip in a box to a now expansive service platform.
20:00 - 21:00: The Future of AI and Managing AI Agents The chapter titled 'The Future of AI and Managing AI Agents' discusses the evolution and potential of artificial intelligence, particularly the concept of Artificial General Intelligence (AGI). It highlights how achieving economies of scale and adding value in cloud computing have led to high-profit margins. The transcript suggests that the development of AGI, which could assist in creating more advanced AI systems, including synthetic data and reinforcement learning (RL), provides a strategic advantage. This advancement could eventually lead to the emergence of automated AI researchers, further entrenching the competitive edge of those leading in this field. The speaker underscores the importance of staying ahead in the AI race, noting that at a large scale, nothing remains a commodity.
21:00 - 22:00: Microsoft's Quantum Breakthrough The chapter discusses the misconception that cloud computing is a mere commodity and emphasizes the significance of the expertise required to run a large-scale cloud service, or hyperscaler. Initially, many believed that cloud hosting businesses were not profitable, questioning the viability of hyperscalers like Azure. However, the chapter highlights how mastering the operation of a hyperscale system, such as managing the world's computing needs across more than 60 regions, proves the existence of a substantial business opportunity. Azure's success in effectively managing this scale demonstrates the critical nature of cloud computing knowledge and the business potential in hyperscale operations.
22:00 - 23:00: The Road to Utility-Scale Quantum Computing The chapter discusses the concept of investing in fields with large Total Addressable Markets (TAMs). It questions whether certain markets are 'winner-take-all' and suggests that it's advantageous to enter categories where various competitors can thrive. This reduces the risk associated with a market being dominated by a single player, making it more appealing for investors and innovators.
23:00 - 24:00: Comparisons with Other Quantum Efforts This chapter discusses the intersection of hyperscale computing with quantum efforts, emphasizing the need for models to operate on hyperscale compute. It highlights the continuous necessity of storage and traditional computing resources for maintaining states and running agent-based environments.
24:00 - 25:00: AI, Quantum, and the Future of Computing The chapter discusses the limitations of a single entity dominating with one AI model and highlights the advantages of hyperscalers, especially in terms of leveraging data centers and GPUs for both AI model training and inference.
25:00 - 26:00: Quantum Computing Timeline and Applications The chapter discusses what kind of hyperscaler Microsoft Azure represents in the context of quantum computing and cloud services. It explores whether Azure focuses on pre-training, O3-type inference, or hosting and deploying models. The discussion also touches on Azure's strategy concerning Moore's Law.
26:00 - 27:00: The Role of AI in Simulating Nature The chapter discusses how artificial intelligence has been increasingly used to simulate nature, highlighting the importance of refreshing the technology fleet regularly. It emphasizes the strategy of asset depreciation and the effective placement of fleet resources to handle various tasks efficiently. The chapter points out the occasional necessity for highly concentrated peak periods to address extensive training jobs.
27:00 - 28:00: Long-Term Research Decisions at Microsoft The chapter discusses the challenges Microsoft faces in managing long-term research due to the growing scale of their data processing needs. As data processing requirements grow, particularly in pre-training, Microsoft needs to expand its data center capabilities. Challenges emerge as data processing scales to a point where it surpasses the capacity of individual data centers, requiring coordination across multiple centers.
28:00 - 29:00: Microsoft's Research Strategy and Goals The chapter discusses Microsoft's research strategy and goals, particularly focusing on the importance of building a robust distributed computing infrastructure. This infrastructure is essential for supporting large training jobs, test-time computations, and potential reinforcement learning (RL) applications. The strategy emphasizes the need for increased training capabilities to create specialized models.
29:00 - 30:00: Cultural and Managerial Challenges The chapter titled 'Cultural and Managerial Challenges' discusses the implementation of distilled models tailored for various tasks, focusing on the need for a globally distributed inference fleet. It highlights the importance of minimizing latency by ensuring that data centers are strategically located around the world, rather than relying on a single, centralized data center. Aside from inference fleet distribution, the chapter also underscores the importance of proximity between storage, compute resources, and their associated tasks to maintain efficiency in a hyperscale environment.
30:00 - 31:00: Economic Implications of AI and AGI The chapter discusses the infrastructure requirements for AI and AGI technologies, emphasizing the importance of stateless AI accelerators, adequate storage for training data, and the need for systems that support multiplexing of multiple training jobs. It highlights the necessity of memory and execution environments for agents to perform programs effectively.
31:00 - 32:00: The AGI Debate and Human Labor The chapter discusses the financial growth potential of Artificial General Intelligence (AGI) and its impact on human labor. It highlights a projected increase in revenue from AI, estimating it could reach $130 billion if current trends persist. The chapter raises questions about the deployment of such intelligence on an industrial scale, pondering whether it will be through software like Office or hosted deployments by others. The discussion emphasizes the need for AGI to achieve the projected revenue goals and questions what the future utilization of this intelligence will look like.
32:00 - 33:00: Legal and Ethical Considerations in AI The chapter focuses on legal and ethical considerations in artificial intelligence, particularly in the context of rapid growth and commoditization of intelligence. The speaker acknowledges the excitement around artificial general intelligence (AGI) but emphasizes the importance of observing GDP growth as a primary indicator amidst such advancements. The discussion hints at the necessity of cautious progression, especially when considering the implications for corporations like Microsoft. The speaker acknowledges that while there's significant potential, it's equally essential to consider the constraints imposed by existing economic structures, especially in the developed world, which comprises approximately 2% of the global context.
33:00 - 34:00: AI Alignment and Deployment Challenges The chapter discusses growth challenges, drawing parallels to the Industrial Revolution. It touches on the necessity of achieving high growth rates, such as 10% or 7%, particularly in the developed world, to mirror the growth experienced during the Industrial Revolution.
34:00 - 35:00: The Future of SaaS and AI Integration This chapter discusses the integration of SaaS (Software as a Service) and AI (Artificial Intelligence) and its implications on different industries. It highlights that while the tech companies are prominent players in this integration, the real beneficiaries will be the broader industries that use these technologies. As SaaS and AI become more available and abundant, they are expected to drive significant productivity growth across various sectors.
35:00 - 36:00: Microsoft's Legacy and Future Vision The chapter titled "Microsoft's Legacy and Future Vision" discusses the criteria for measuring progress and success in the tech industry. It critiques the self-proclaimed achievements of companies claiming advancements in Artificial General Intelligence (AGI) as mere 'benchmark hacking'. The real measure of success, according to the text, is a substantial growth rate of the global economy, cited as 10%. Such growth would translate to an increase of $10 trillion in value annually, marking true advancement and industry wellbeing.
36:00 - 37:00: Reflection on Leadership and Company Growth This chapter discusses the significant role of hyperscalers in the growth of the world economy. It questions whether current financial investments, like $80 billion, might be too meager compared to the actual potential and demand, suggesting perhaps a goal as high as $800 billion. The chapter highlights that one of the primary limitations to tapping into this growth is having the adequate computational power necessary for deploying AI technologies effectively. Additionally, the classic supply-side approach assumes a model of building the technological infrastructure expecting subsequent demand.
37:00 - 38:40: Final Thoughts and Future Directions The chapter emphasizes the importance of aligning supply and demand in business strategies. It warns against getting caught up in self-hype without recognizing real customer value. Successful strategies require balancing risk-taking with practical application and customer focus to ensure both supply and demand sides are effectively managed.
Satya Nadella – Microsoft’s AGI Plan & Quantum Breakthrough Transcription
00:00 - 00:30
00:30 - 01:00 Satya, thank you so much
for coming on the podcast. In a second, we're going to get to the two
breakthroughs that Microsoft has just made, and congratulations, same day in Nature: the
Majorana zero chip, which we have in front of us right here, and also the world human action
models. But can we just continue the conversation
01:00 - 01:30 we were having a second ago? You're describing
the ways in which the things you were seeing in the 80s and 90s, you're seeing them happen again.
The thing that is exciting for me... Dwarkesh, first of all, it's fantastic to be
on your podcast. I'm a big listener, and I love the way that you do these interviews
and the broad topics that you explore. The thing that is exciting for me… It reminds me
a little bit of my, I'd say, first few years even
01:30 - 02:00 in the tech industry, starting in the 90s, where
there was real debate about whether it's going to be RISC or CISC, or, "Hey, are we really
going to be able to build servers using x86?" When I joined Microsoft, that was the
beginning of what was Windows NT. So, everything from the core silicon platform to the
operating system to the app tier- that full stack
02:00 - 02:30 approach- the entire thing is being litigated.
You could say cloud did a bunch of that, and obviously distributed computing and cloud
did change client-server. The web changed massively. But this does feel a little more
like maybe more full-stack than even the past that at least I've been involved in.
When you think about which decisions
02:30 - 03:00 ended up being the long-term winners in
the 80s and 90s, and which ones didn't, and especially when you think about- you were at Sun
Microsystems, they had an interesting experience with the 90s dotcom bubble. People talk about
this data center build-out as being a bubble, but at the same time, we have the Internet
today as a result of what was built out then. What are the lessons about what will stand
the test of time? What is an inherent secular trend? What is just ephemeral?
If I go back, the four big transformations
03:00 - 03:30 that I've been part of, the client and the
client-server. So that's the birth of the graphical user interface and the x86 architecture,
basically allowing us to build servers. It was very clear to me. I remember going to
what is PDC in '91, in fact I was at Sun at that time. In '91, I went to Moscone. That's when
Microsoft first described the Win32 interface and
03:30 - 04:00 it was pretty clear to me what was going to
happen, where the server was also going to be an x86 thing. When you have the scale advantages
accruing to something, that's the secular bet you have to place. What happened in the client was
going to happen on the server side, and then you were able to actually build client-server
applications. So, the app model became clear.
04:00 - 04:30 Then the web was the big thing for us,
which we had to deal with in starting, in fact as soon as I joined Microsoft, the
Netscape browser or the Mosaic browser came out what, I think, December or November of '93, right?
I think is when Andreessen and crew had that. So that was a big game-changer, in an interesting
way, just as we were getting going on what was the client-server wave, and it was clear that
we were going to win it as well. We had the
04:30 - 05:00 browser moment, and so we had to adjust.
And we did a pretty good job of adjusting to it because the browser was a new app model.
We were able to embrace it with everything we did, whether it was HTML in Word or building a new
thing called the browser ourselves and competing for it, and then building a web server on
our server stack and go after it. Except,
05:00 - 05:30 of course, we missed what turned out to
be the biggest business model on the web, because we all assumed the web is all about being
distributed, who would have thought that search would be the biggest winner in organizing the web?
And so that's where we obviously didn't see it, and Google saw it and executed super well.
So that's one lesson learned for me: you have to not only get the tech trend right, you also
have to get where the value is going to be created
05:30 - 06:00 with that trend. These business model shifts are
probably tougher than even the tech trend changes. Where is the value going to be created in AI?
That's a great one. So I think there are two places where I can say with some confidence.
One is the hyperscalers that do well,
06:00 - 06:30 because the fundamental thing is if you sort of
go back to even how Sam and others describe it, if intelligence is log of compute, whoever
can do lots of compute is a big winner. The other interesting thing is, if you
look at underneath even any AI workload, like take ChatGPT, it's not like everybody's
excited about what's happening on the GPU side, it's great. In fact, I think of my fleet even
as a ratio of the AI accelerator to storage,
06:30 - 07:00 to compute. And at scale, you've got to grow it.
Yeah. And so, that infrastructure need for the world
is just going to be exponentially growing. Right.
So in fact it's manna from heaven to have these AI workloads because
guess what? They're more hungry for more compute, not just for training, but we now know, for
test time. When you think of an AI agent, it
07:00 - 07:30 turns out the AI agent is going to exponentially
increase compute usage because you're not even bound by just one human invoking a program. It's
one human invoking programs that invoke lots more programs. That's going to create massive, massive
demand and scale for compute infrastructure. So our hyperscale business, Azure business, and
other hyperscalers, I think that’s a big thing. Then after that, it becomes a little fuzzy. You
could say, hey, there is a winner-take-all model-
07:30 - 08:00 I just don't see it. This, by the way, is the
other thing I’ve learned: being very good at understanding what are winner-take-all markets
and what are not winner-take-all markets is, in some sense, everything. I remember even in
the early days when I was getting into Azure, Amazon had a very significant lead and people
would come to me, and investors would come to me, and say, "Oh, it's game over. You'll never
make it. Amazon, it's winner-take-all."
08:00 - 08:30 Having competed against Oracle and IBM
in client-server, I knew that the buyers will not tolerate winner-take-all.
Structurally, hyperscale will never be a winner-take-all because buyers are smart.
Consumer markets sometimes can be winner-take-all, but anything where the buyer is a
corporation, an enterprise, an IT department, they will want multiple suppliers. And so
you got to be one of the multiple suppliers.
08:30 - 09:00 That, I think, is what will happen even on the
model side. There will be open-source. There will be a governor. Just like on Windows,
one of the big lessons learned for me was, if you have a closed-source operating
system, there will be a complement to it, which will be open source.
And so to some degree that's a real check on what happens. I think in models
there is one dimension of, maybe there will be a few closed source, but there will definitely be
an open source alternative, and the open-source
09:00 - 09:30 alternative will actually make sure that the
closed-source, winner-take-all is mitigated. That's my feeling on the model side. And by the
way, let's not discount if this thing is really as powerful as people make it out to be, the
state is not going to sit around and wait for private companies to go around… and all over the
world. So, I don't see it as a winner-take-all. Then above that, I think it's going to be
the same old stuff, which is in consumer,
09:30 - 10:00 in some categories, there may be
some winner-take-all network effect. After all, ChatGPT is a great example.
It's an at-scale consumer property that has already got real escape velocity. I go to the
App Store, and I see it's always there in the top five, and I say “wow, that's pretty unbelievable”.
So they were able to use that early advantage and
10:00 - 10:30 parlay that into an app advantage. In consumer,
that could happen. In the enterprise again, I think there will be, by category, different
winners. That's sort of at least how I analyze it. I have so many follow-up questions. We have to
get to quantum in just a second, but on the idea that maybe the models get commoditized: maybe
somebody could have made a similar argument a couple of decades ago about the cloud – that
fundamentally, it's just a chip and a box.
10:30 - 11:00 But in the end, of course, you and many others
figured out how to get amazing profit margins in the cloud. You figured out ways to get economies
of scale and add other value. Fundamentally, even forgetting the jargon, if you've got AGI
and it's helping you make better AIs – right now, it's synthetic data and RL; maybe in the future,
it's an automated AI researcher – that seems like a good way to entrench your advantage there. I'm
curious what you make of that, just the idea that it really matters to be ahead there.
At scale, nothing is commodity.
11:00 - 11:30 To your point about cloud, everybody would
say, "Oh, cloud's a commodity." Except, when you scale... That's why the know-how of
running a hyperscaler... You could say, "Oh, what the heck? I can just rack and stack servers."
Right. In fact, in the early days of hyperscale, most
people thought “there are all these hosters, and those are not great businesses. Will there be
anything? Is there a business even in hyperscale?” And it turns out there is a real business,
just because of the know-how of running, in the case of Azure, the world's computing of
60-plus regions with all the compute. It's
11:30 - 12:00 just a tough thing to duplicate.
So I was more making the point, is it one winner? Is it a winner-take-all or
not? Because that you've got to get right. I like to enter categories which are big TAMs,
where you don't have to have the risk of it all being winner-take-all. The best news to
be in is a big market that can accommodate
12:00 - 12:30 a couple of winners, and you're one of them.
That's what I meant by the hyperscale layer. In the model layer, one is models need ultimately
to run on some hyperscale compute. So that nexus, I feel, is going to be there forever. It's not
just the model; the model needs state, that means it needs storage, and it needs regular compute for
running these agents and the agent environments.
12:30 - 13:00 And so that's how I think about why the
limit of one person running away with one model and building it all may not happen.
On the hyperscaler side, and by the way, it's also interesting the advantage you as a hyperscaler
would have in the sense that, especially with inference time scaling and if that's involved
in training future models, you can amortize your data centers and GPUs, not only for the
training, but then use them again for inference.
13:00 - 13:30 I'm curious what kind of hyperscaler you
consider Microsoft and Azure to be. Is it on the pre-training side? Is it on providing
the O3-type inference? Or are you just, we’re going to host and deploy any single
model that's out there in the market, and we are sort of agnostic about that?
It’s a good point. The way we want to build out the fleet is [to], in some sense
ride Moore's law. I think this will be like
13:30 - 14:00 what we've done with everything else in the
past: every year keep refreshing the fleet, you depreciate it over whatever the lifetime value
of these things are, and then get very very good at the placement of the fleet such that you can
run different jobs at it with high utilization. Sometimes there are very big training jobs
that need to have highly concentrated peak
14:00 - 14:30 flops that are provisioned to it that also need
to cohere. That's great. We should have enough data center footprint to be able to give that.
But at the end of the day, these are all becoming so big, even in terms of if you take pre-training
scale, if it needs to keep going, even pre-training scale at some point has to cross data
center boundaries. It's all more or less there. So, great, once you start crossing pre-training
data center boundaries, is it that different than
14:30 - 15:00 anything else? The way I think about it is hey,
distributed computing will remain distributed, so go build out your fleet such that
it's ready for large training jobs, it's ready for test-time compute, it’s ready-
in fact, if this RL thing that might happens, you build one large model, and then after
that, there’s tons of RL going on. To me, it's kind of like more training flops, because
you want to create these highly specialized,
15:00 - 15:30 distilled models for different tasks.
So you want that fleet, and then the serving needs. At the end of the day,
speed of light is speed of light, so you can't have one data center in Texas and
say, "I'm going to serve the world from there." You've got to serve the world based on
having an inference fleet everywhere in the world. That's how I think of our
build-out of a true hyperscale fleet. Oh, and by the way, I want my storage and
compute also close to all of these things,
15:30 - 16:00 because it's not just AI accelerators that are
stateless. My training data itself needs storage, and then I want to be able to multiplex
multiple training jobs, I want to be able to then have memory, I want to be able to have these
environments in which these agents can go execute programs. That's kind of how I think about it.
You recently reported that your yearly revenue
16:00 - 16:30 from AI is $13 billion. But if you look at your
year-on-year growth on that, in like four years, it'll be 10x that. You'll have $130 billion in
revenue from AI, if the trend continues. If it does, what do you anticipate doing with all
that intelligence, this industrial scale use? Is it going to be through Office? Is it going
to be you deploying it for others to host? You've got to have the AGIs to have $130
billion in revenue? What does it look like?
16:30 - 17:00 The way I come at it, Dwarkesh, it's a
great question because at some level, if you're going to have this explosion, abundance,
whatever, commodity of intelligence available, the first thing we have to observe is GDP growth.
Before I get to what Microsoft's revenue will look like, there's only one governor in all
of this. This is where we get a little bit ahead of ourselves with all this AGI hype.
Remember the developed world, which is what? 2%
17:00 - 17:30 growth and if you adjust for inflation it’s zero?
So in 2025, as we sit here, I'm not an economist, at least I look at it and say we have a real
growth challenge. So, the first thing that we all have to do is, when we say this is
like the Industrial Revolution, let's have that Industrial Revolution type of growth.
That means to me, 10%, 7%, developed world,
17:30 - 18:00 inflation-adjusted, growing at 5%. That's the
real marker. It can't just be supply-side. In fact that’s the thing, a lot of people
are writing about it, and I'm glad they are, which is the big winners here are not going to be
tech companies. The winners are going to be the broader industry that uses this commodity that, by
the way, is abundant. Suddenly productivity goes
18:00 - 18:30 up and the economy is growing at a faster rate.
When that happens, we'll be fine as an industry. But that's to me the moment. Us
self-claiming some AGI milestone, that's just nonsensical benchmark hacking to me.
The real benchmark is: the world growing at 10%. Okay, so if the world grew at 10%, the world
economy is $100 trillion or something, if the world grew at 10%, that's like an extra $10
trillion in value produced every single year.
18:30 - 19:00 If that is the case, you as a hyperscaler...
It seems like $80 billion is a lot of money. Shouldn't you be doing like $800 billion?
If you really think in a couple of years, we could be really growing the world economy
at this rate, and the key bottleneck would be: do you have the compute necessary to
deploy these AIs to do all this work? That is correct. But by the way, the classic
supply side is, "Hey, let me build it and they’ll
19:00 - 19:30 come." That's an argument, and after all we've
done that, we've taken enough risk to go do it. But at some point, the supply and demand have
to map. That's why I'm tracking both sides of it. You can go off the rails completely when
you are hyping yourself with the supply-side, versus really understanding how to
translate that into real value to customers.
19:30 - 20:00 That's why I look at my inference revenue. That's
one of the reasons why even the disclosure on the inference revenue... It's interesting that not
many people are talking about their real revenue, but to me, that is important as a
governor for how you think about it. You're not going to say they have to symmetrically
meet at any given point in time, but you need to have existence proof that you are able to parlay
yesterday's, let’s call it capital, into today's demand, so that then you can again invest,
maybe exponentially even, knowing that you're
20:00 - 20:30 not going to be completely rate mismatched.
I wonder if there's a contradiction in these two different viewpoints, because one of the
things you've done wonderfully is make these early bets. You invested in OpenAI in 2019, even
before there was Copilot and any applications. If you look at the Industrial Revolution,
these 6%, 10% build-outs of railways and whatever things, many of those were not
like, "We've got revenue from the tickets, and now we're going to..."
There was a lot of money lost.
20:30 - 21:00 That's true. So, if you really think there's
some potential here to 10x or 5x the growth rate of the world, and then you're like,
"Well, what is the revenue from GPT-4?" If you really think that's the possibility
from the next level up, shouldn't you just, "Let's go crazy, let's do the hundreds of
billions of dollars of compute?" I mean, there's some chance, right?
Here’s the interesting thing, right?
21:00 - 21:30 That's why even that balanced approach to the
fleet, at least, is very important to me. It's not about building compute. It's about building
compute that can actually help me not only train the next big model but also serve the next big
model. Until you do those two things, you're not going to be able to really be in a position
to take advantage of even your investment. So, that's kind of where it's not
a race to just building a model, it's a race to creating a commodity that
is getting used in the world to drive...
21:30 - 22:00 You have to have a complete thought, not
just one thing that you’re thinking about. And by the way, one of the things is
that there will be overbuild. To your point about what happened in the dotcom era,
the memo has gone out that, hey, you know, you need more energy, and you need more compute.
Thank God for it. So, everybody's going to race. In fact, it's not just companies deploying,
countries are going to deploy capital,
22:00 - 22:30 and there will be clearly... I'm so excited to
be a leaser, because, by the way; I build a lot, I lease a lot. I am thrilled that I'm going
to be leasing a lot of capacity in '27, '28 because I look at the builds, and I'm
saying, "This is fantastic." The only thing that's going to happen with all the compute
builds is the prices are going to come down. Speaking of prices coming down, you recently
tweeted after the DeepSeek model came out about Jevons’ Paradox. I'm curious if you can flesh that
out. Jevons’ Paradox occurs when the demand for
22:30 - 23:00 something is highly elastic. Is intelligence
that bottlenecked on prices going down? Because when I think about, at least my use cases
as a consumer, intelligence is already so cheap. It's like two cents per million tokens. Do I
really need it to go down to 0.02 cents? I'm just really bottlenecked on it becoming smarter. If you
need to charge me 100x, do a 100x bigger training run. I'm happy for companies to take that.
But maybe you're seeing something different
23:00 - 23:30 on the enterprise side or something. What is the
key use case of intelligence that really requires it to get to 0.002 cents per million tokens?
I think the real thing is the utility of the tokens. Both need to happen: One is intelligence
needs to get better and cheaper. And anytime there's a breakthrough, like even what DeepSeek
did, with the efficient frontier of performance per token changes, the curve gets bent,
and the frontier moves. That just brings
23:30 - 24:00 more demand. That's what happened with cloud.
Here’s an interesting thing: We used to think “oh my God, we've sold all the servers in the
client-server era”. Except once we started putting servers in the cloud, suddenly people
started consuming more because they could buy it cheaper, and it was elastic, and they
could buy it as a meter versus a license,
24:00 - 24:30 and it completely expanded.
I remember going, let’s say, to a country like India and talking about
“here is SQL Server”. We sold a little, but man, the cloud in India is so much bigger
than anything that we were able to do in the server era. I think that's going to be true.
If you think about, if you want to really have, in the Global South, in a developing country,
if you had these tokens that were available for healthcare that were really cheap,
that would be the biggest change ever.
24:30 - 25:00 I think it's quite reasonable for somebody to
hear people like me in San Francisco and think
25:00 - 25:30 “they're kind of silly; they don't know what it's
actually like to deploy things in the real world”.
25:30 - 26:00 As somebody who works with these Fortune
500s and is working with them to deploy things for hundreds of millions, billions
of people, what's your sense on how fast deployment of these capabilities will be?
Even when you have working agents, even when you have things that can do remote work for you,
with all the compliance and with all the inherent
26:00 - 26:30 bottlenecks, is that going to be a big bottleneck,
or is that going to move past pretty fast? It is going to be a real challenge because
the real issue is change management or process change. Here's an interesting thing: one of
the analogies I use is, just imagine how a multinational corporation like us did forecasts
pre-PC, and email, and spreadsheets. Faxes went
26:30 - 27:00 around. Somebody then got those faxes and did an
interoffice memo that then went around, and people entered numbers, and then ultimately a forecast
came, maybe just in time for the next quarter. Then somebody said, "Hey, I'm just going to
take an Excel spreadsheet, put it in email, send it around. People will go edit it,
and I'll have a forecast." So, the entire forecasting business process changed because
the work artifact and the workflow changed.
27:00 - 27:30 That is what needs to happen with AI being
introduced into knowledge work. In fact, when we think about all these agents, the fundamental
thing is there's a new work and workflow. For example, even prepping for our podcast, I go
to my copilot and I say, "Hey, I'm going to talk to Dwarkesh about our quantum announcement and
this new model that we built for game generation.
27:30 - 28:00 Give me a summary of all the stuff that I should
read up on before going." It knew the two Nature papers, it took that. I even said, "Hey, go give
it to me in a podcast format." And so, it even did a nice job of two of us chatting about it.
So that became—and in fact, then I shared it with my team. I took it and put it
into Pages, which is our artifact,
28:00 - 28:30 and then shared. So the new workflow for me is
I think with AI and work with my colleagues. That's a fundamental change management of everyone
who's doing knowledge work, suddenly figuring out these new patterns of "How am I going to get
my knowledge work done in new ways?" That is going to take time. It's going to be something
like in sales, and in finance, and supply chain. For an incumbent, I think that this is going
to be one of those things where—you know,
28:30 - 29:00 let's take one of the analogies I like to use
is what manufacturers did with Lean. I love that because, in some sense, if you look at it,
Lean became a methodology of how one could take an end-to-end process in manufacturing and become
more efficient. It's that continuous improvement, which is reduce waste and increase value.
That's what's going to come to knowledge. This is like Lean for knowledge work, in
particular. And that's going to be the hard work of management teams and individuals who are doing
knowledge work, and that's going to take its time.
29:00 - 29:30 Can I ask you just briefly about that
analogy? One of the things Lean did is physically transform what a factory floor
looks like. It revealed bottlenecks that people didn't realize until you're really paying
attention to the processes and workflows. You mentioned briefly what your own
workflow—how your own workflow has changed as a result of AIs. I'm curious if we
can add more color to what will it be like to run a big company when you have these AI agents
that are getting smarter and smarter over time? It's interesting you ask that. I was thinking,
for example, today if I look at it, we are very
29:30 - 30:00 email heavy. I get in in the morning, and I’m
like, man my inbox is full, and I'm responding, and so I can't wait for some of these
Copilot agents to automatically populate my drafts so that I can start reviewing and sending.
But I already have in Copilot at least ten agents, which I query them different things for
different tasks. I feel like there's a new
30:00 - 30:30 inbox that's going to get created, which is
my millions of agents that I'm working with will have to invoke some exceptions to me,
notifications to me, ask for instructions. So at least what I'm thinking is that there's
a new scaffolding, which is the agent manager. It's not just a chat interface. I need
a smarter thing than a chat interface to
30:30 - 31:00 manage all the agents and their dialogue.
That's why I think of this Copilot, as the UI for AI, is a big, big deal. Each of us
is going to have it. So basically, think of it as: there is knowledge work, and there's a knowledge
worker. The knowledge work may be done by many, many agents, but you still have a knowledge worker
who is dealing with all the knowledge workers. And that, I think, is the
interface that one has to build.
31:00 - 31:30 You're one of the few people in the world who
can say that you have access to 200,000… you have this swarm of intelligence around you in
the form of Microsoft the company and all its employees. And you have to manage that, and
you have to interface with that, how to make best use of that. Hopefully, more of the world
will get to have that experience in the future. I'd be curious about how your inbox,
if that means everybody's inbox, will look like yours in the morning.
Okay, before we get to that,
31:30 - 32:00 I want to keep asking you more about AI,
but I really want to ask you about the big breakthrough in quantum that Microsoft Research
has announced. So can you explain what's going on? This has been another 30-year journey for
us. It's unbelievable. I'm the third CEO of Microsoft who's been excited about quantum.
The fundamental breakthrough here, or the vision that we've always had is, you need a physics
breakthrough in order to build a utility-scale
32:00 - 32:30 quantum computer that works. We took the path
of saying, the one way for having a less noisy or more reliable qubit is to bet on a physical
property that by definition is more reliable and
32:30 - 33:00 that's what led us to the Majorana zero modes,
which was theorized in the 1930s. The question was, can we actually physically fabricate
these things? Can we actually build them? So the big breakthrough effectively, and I know
you talked to Chetan, was that we now finally have existence proof and a physics breakthrough
of Majorana zero modes in a new phase of matter
33:00 - 33:30 effectively. This is why we like the analogy
of thinking of this as the transistor moment of quantum computing, where we effectively have a
new phase, which is the topological phase, which means we can even now reliably hide the quantum
information, measure it, and we can fabricate it.
33:30 - 34:00 And so now that we have it, we feel like with that
core foundational fabrication technique out of the way, we can start building a Majorana chip.
That Majorana One which I think is going to basically be the first chip that will be capable
of a million qubits, physical. And then on that, thousands of logical qubits, error-corrected.
And then it's game on. You suddenly have the
34:00 - 34:30 ability to build a real utility-scale quantum
computer, and that to me is now so much more feasible. Without something like this, you
will still be able to achieve milestones, but you'll never be able to build a utility-scale
computer. That's why we're excited about it. Amazing. And by the way, I
believe this is it right here. That is it.
Yes. I forget now, are we calling it Majorana? Yes, that's right. Majorana One. I'm
glad we named it after that.
34:30 - 35:00 To think that we are able to build something
like a million-qubit quantum computer in a thing of this size is just unbelievable. That's the
crux of it: unless and until we could do that, you can't dream of building a
utility-scale quantum computer. And you're saying the eventual million qubits
will go on a chip this size? Okay, amazing.
35:00 - 35:30 Other companies have announced 100 physical
qubits, Google's, IBM's, others. When you say you've announced one, but you're saying
that yours is way more scalable in the limit. Yeah. The one thing we’ve also done is we've
taken an approach where we've separated our software and our hardware. We're building
out our software stack, and we now have, with the neutral atom folks, the ion trap folks,
we're also working with others who even have
35:30 - 36:00 pretty good approaches with photonics and what
have you, that means there'll be different types of quantum computers. In fact, we have what, I
think that the last thing that we announced was 24 logical qubits. So we have also got some fantastic
breakthroughs on error correction and that's what is allowing us, even on neutral atom and ion
trap quantum computers, to build these 20 plus,
36:00 - 36:30 and I think that'll keep going even throughout
the year; you'll see us improve that yardstick. But we also then said, "Let's go to the first
principles and build our own quantum computer that is betting on the topological qubit."
And that's what this breakthrough is about. Amazing. The million topological
qubits, thousands of logical qubits, what is the estimated timeline to scale up to
that level? What does the Moore's law here,
36:30 - 37:00 if you've got the first transistor, look like?
We've obviously been working on this for 30 years. I'm glad we now have the physics
breakthrough and the fabrication breakthrough. I wish we had a quantum computer because
by the way, the first thing the quantum computer will allow us to do is build
quantum computers, because it's going to be so much easier to simulate atom-by-atom
construction of these new quantum gates. But in any case, the next real thing is,
now that we have the fabrication technique,
37:00 - 37:30 let us go build that first fault-tolerant quantum
computer. And that will be the logical thing. So, I would say now I can say, "Oh, maybe
'27, '28, '29, we will be able to actually build this." Now that we have this one gate, can I
put the thing into an integrated circuit and then actually put these integrated circuits into a real
computer? That is where the next logical step is.
37:30 - 38:00 And what do you see as, in '27, '28, you've got
it working? Is it a thing you access through the API? Is it something you're using internally
for your own research in materials and chemistry? It’s a great question. One thing that I've
been excited about is, even in today's world… we had this quantum program, and we added
some APIs to it. The breakthrough we had maybe two years ago was to think of this HPC
stack, and AI stack, and quantum together.
38:00 - 38:30 In fact, if you think about it, AI is like an
emulator of the simulator. Quantum is like a simulator of nature. What is quantum going
to do? By the way, quantum is not going to replace classical. Quantum is great at what
quantum can do, and classical will also... Quantum is going to be fantastic for anything
that is not data-heavy but is exploration-heavy
38:30 - 39:00 in terms of the state space. It should be
data-light but exponential states that you want to explore. Simulation is a great one:
chemical physics, what have you, biology. One of the things that we've started doing
is really using AI as the emulation engine. But you can then train. So the way I think of it
is, if you have AI plus quantum, maybe you'll use
39:00 - 39:30 quantum to generate synthetic data that then gets
used by AI to train better models that know how to model something like chemistry or physics or what
have you. These two things will get used together. So even today, that's effectively what
we're doing with the combination of HPC and AI. I hope to replace some of
the HPC pieces with quantum computers.
39:30 - 40:00 Can you tell me a little bit about how
you make these research decisions which,
40:00 - 40:30 in 20 years time, 30 years time, will
actually pay dividends, especially at
40:30 - 41:00 a company of Microsoft's scale? Obviously, you're
in great touch with the technical details in this project. Is it feasible for you to do that
with all the things Microsoft Research does? How do you know the current bet you're making
will pay out in 20 years? Does it just have to emerge organically through the org, or
how are you keeping track of all this? The thing that I feel was fantastic is when Bill,
when he started MSR back in '95 I guess. I think
41:00 - 41:30 in the long history of these curiosity-driven
research organizations, to just do a research org that is about fundamental research and MSR,
over the years, has built up that institutional strength so when I think about capital allocation
or budgets, we first put the chips in and say,
41:30 - 42:00 "Here is MSR's budget." We gotta go at it each
year knowing that most of these bets are not going to pay off in any finite time frame. Maybe
the sixth CEO of Microsoft will benefit from it. And in tech that is I think a given.
The real thing that I think about is, when the time has come for something like
quantum or a new model or what have you, can
42:00 - 42:30 you capitalize? So as an incumbent, if you look at
the history of tech, it's not that people didn't invest. It's that you need to have a culture that
knows how to take an innovation and scale it. That's the hard part, quite frankly, for CEOs and
management teams. Which is kind of fascinating.
42:30 - 43:00 It's as much about good judgment as it is about
good culture. Sometimes we've gotten it right; sometimes we've gotten it wrong; I can tell you
the thousand projects from MSR that we should have probably led with, but we didn't. And I always
ask myself why. It's because we were not able to get enough conviction and that complete thought
of how to not only take the innovation but make
43:00 - 43:30 it into a useful product with a business
model that we can then go to market with. That's the job of CEOs and management teams:
not to just be excited about any one thing, but to be able to actually execute on a complete
thing. And that's easier said than done. When you mentioned the possibility of three
subsequent CEOs of Microsoft, if each of them increases the market cap by an order of magnitude,
by the time you've got the next breakthrough, you'll be like the world economy or something.
Or remember, the world is going to be growing
43:30 - 44:00 at 10%, so we'll be fine.
Let's dig into the other big breakthrough you've just made. It's amazing that
you have both of them coming out the same day, in your gaming world models. I'd love if
you can tell me a little bit about that. We're going to call it Muse. It's going to be the
model of this world action, or human action model. This is very cool. One of the things is
that obviously, Dall-E and Sora have been unbelievable in what they've been able
to do in terms of generative models. One
44:00 - 44:30 thing that we wanted to go after was
using gameplay data. Can you actually generate games that are both consistent
and then have the ability to generate the diversity of what that game represents,
and then are persistent to user mods? That's what this is. They were able
to work with one of our game studios, and this is the other publication in Nature.
The cool thing is what I'm excited about is
44:30 - 45:00 bringing--we're going to have a catalog of games
soon that we will start using these models, or we're going to train these models to
generate, and then start playing them. In fact, when Phil Spencer first showed it
to me, he had an Xbox controller and this model basically took the input and generated the
output based on the input. And it was consistent with the game. That to me is a massive moment
of “wow”. It's kind of like the first time we
45:00 - 45:30 saw ChatGPT complete sentences, or Dall-E
draw, or Sora. This is one such moment. I got a chance to see some of the videos
in the real-time demo this morning with your lead researcher Katja on this. Only
once I talked to her did it really hit me how incredible this is, in the sense that
we've used AI in the past to model agents,
45:30 - 46:00 and just using that same technique to model
the world around the agent gives consistent real-time – we'll superimpose videos of what this
looks like atop this podcast so people can get a chance to see it for themselves. I guess it'll
be out by then, so they can also watch it there. This in itself is incredible. You, through your
span as CEO, have invested tens of hundreds of billions of dollars in building up
Microsoft Gaming and acquiring IP. In retrospect, if you can just merge all of
this data into one big model that can give
46:00 - 46:30 you this experience of visiting and going
through multiple worlds at the same time, and if this is the direction gaming is headed,
it seems like a pretty good investment to have made. Did you have any premonition about this?
I wouldn't say that we invested in gaming to build models. We invested, quite frankly, because-
here's an interesting thing about our history: We built our first game before we built
Windows. Flight Simulator was a Microsoft
46:30 - 47:00 product long before we even built Windows.
So, gaming has got a long history at the company, and we want to be in gaming for gaming's
sake. I always start by saying I hate to be in businesses where they're means to some
other end. They have to be ends unto themselves. And then, yes, we're not a conglomerate.
We are a company where we have to bring all these assets together and be better owners
by adding value. For example, cloud gaming is
47:00 - 47:30 a natural thing for us to invest in because that
will just expand the TAM and expand the ability for people to play games everywhere.
The same thing with AI and gaming: we definitely think that it can be helpful in
maybe changing- it's kind of like the CGI moment, even for gaming long-term. And it's great.
As the biggest, world's largest publisher, this will be helpful. But at the same time,
we've got to produce great quality games. I mean, you can't be a gaming publisher without, sort
of, first and foremost being focused on that.
47:30 - 48:00 But the fact that this data asset is going to
be interesting, not just in a gaming context, but it's going to be a general action model
and a world model, it's fantastic. I mean like, you know, I think about gaming data as perhaps,
you know, what YouTube is perhaps to Google, gaming data is to Microsoft. And so
therefore I'm excited about that. Yeah, and that's what I meant, just in the sense
of like, you can have one unified experience across many different kinds of games. How does
this fit into the other, separate from AI,
48:00 - 48:30 the other things that Microsoft has worked on
in the past, like mixed reality? Maybe giving smaller game studios a chance to build these AAA
action games? Just like five, ten years from now, what kinds of ways could you imagine?
I've thought about these three things as the cornerstones of, in an interesting way,
even five, six, seven years ago is when I said the three big bets that we want to place [are]
AI, quantum, and mixed reality. And I still believe in them, because in some sense,
what are the big problems to be solved?
48:30 - 49:00 Presence. That's the dream of mixed
reality. Can you create real presence? Like you and I doing a podcast like this.
I think it’s still proving to be the harder one of those challenges, quite honestly. I thought
it was going to be more solvable. It's tougher, perhaps, just because of the social
side of it: wearing things and so on.
49:00 - 49:30 We're excited about, in fact, what we're
going to do with Anduril and Palmer, now, with even how they'll take forward the
IVAS program, because that's a fantastic use case. And so we'll continue on that front.
But also, the 2D surfaces. It turns out things like Teams, right, thanks to the pandemic,
we've really gotten the ability to create essentially presence through even 2D. And that
I think will continue. That's one secular piece. Quantum we talked about, and AI is the other one.
So these are the three things that I look at and
49:30 - 50:00 say, how do you bring these things together?
Ultimately, not as tech for tech's sake, but solving some of the fundamental things that
we, as humans, want in our life, and more, we want them in our economy, driving our productivity.
And so if we can somehow get that right, then I think we will have really made progress.
When you write your next book, you've got to have some explanation of why those three pieces
all came together around the same time, right? Like, there's no intrinsic reason you
would think quantum and AI should happen in
50:00 - 50:30 2028 and 2025 and so forth.
That's right. At some level, I look at it and say: the simple model I have
is, hey is there a systems breakthrough? To me, the systems breakthrough is the quantum thing.
Is there a business logic breakthrough? That's AI to me, which is: can the logic tier be
fundamentally reasoned differently? Instead of imperatively writing code, can you have
a learning system? That's the AI one.
50:30 - 51:00 And then the UI side of it is presence.
Going back to AI for a second, in your 2017 book… 2019 you invest in OpenAI, very early,
2017 is even earlier, you say in your book, "One might also say that we're birthing a new species,
one whose intelligence may have no upper limits." Now, super-early, of course, to be talking
about this in 2017. We've been talking in a granular fashion about agents, Office
Copilot, capex, and so forth. But if you
51:00 - 51:30 zoom out and consider this statement you've
made, and you think about you as a hyperscaler, as the person doing research in these models as
well, providing training, inference, and research for building a new species, how do you think
about this in the grand scheme of things? Do you think we're headed towards
superhuman intelligence in your time as CEO? I think even Mustafa uses that term. In fact he’s
used that term more recently, this “new species”.
51:30 - 52:00 The way I come at it is, you definitely need
trust. Before we claim it is something as big as a species, the fundamental thing that we've
got to get right is that there is real trust, whether it's personal or societal level trust,
that's baked in. That's the hard problem. I think the one biggest rate limiter to the
power here will be how does our legal… call
52:00 - 52:30 it infrastructure, we’re talking about all
the compute infrastructure, well how does the legal infrastructure evolve to
deal with this? This entire world is constructed with things like humans owning
property, having rights, and being liable. That’s the fundamental thing that one has to first
say, okay what does that mean for anything that
52:30 - 53:00 now humans are using as tools? And if humans are
going to delegate more authority to these things, then how does that structure evolve? Until that
really gets resolved, I don't think just talking about the tech capability is going to happen.
As in, we won't be able to deploy these kinds of intelligences until we figure out how to…?
Absolutely. Because at the end of the day, there is no way. Today, you cannot
deploy these intelligences unless and
53:00 - 53:30 until there's someone indemnifying it as a human.
To your point, I think that's one of the reasons why I think about even the most powerful AI
is essentially working with some delegated authority from some human. You can say, oh, that's
all alignment and this, that, and the other. That's why I think you have to really get these
alignments to work and be verifiable in some way, but I just don't think that you can deploy
intelligences that are out of control. For example, this AI takeoff problem may be a real
problem, but before it is a real problem, the real
53:30 - 54:00 problem will be in the courts. No society is going
to allow for some human to say, "AI did that." Yes. Well, there's a lot of societies in the
world, and I wonder if any one of them might not have a legal system that might be more
amenable. And if you can't have a takeoff, then you might worry. It doesn't
have to happen in America, right?
54:00 - 54:30 We think that no society cares about it, right?
There can be rogue actors, I'm not saying there won't be rogue actors; there are cyber criminals
and rogue states; they're going to be there. But to think that human society at large
doesn't care about it is also not going to be true. I think we all will care. We know
how to deal with rogue states and rogue actors today. The world doesn't sit around and say
“we’ll tolerate that”. That's why I'm glad
54:30 - 55:00 that we have a world order in which anyone who is
a rogue actor in a rogue state has consequences. Right. But if you have this picture where you
can have 10% economic growth, I think it really depends on getting something like AGI working,
because tens of trillions of dollars of value, that sounds closer to the total of human wages,
around $60 trillion of the economy. Getting that magnitude, you kind of have to automate labor
or supplement labor in a very significant way.
55:00 - 55:30 If that is possible, and once we figure
out the legal ramifications for it, it seems quite plausible, even within your
tenure that we figure that out. Are you thinking about superhuman intelligence? Like,
the biggest thing you do in your career is this? You bring up another point. I know David Autor
and others have talked a lot about this which is, 60% of labor- I think the other question
that needs to happen, let’s at least talk about our democratic societies. I think that
in order to have a stable social structure,
55:30 - 56:00 and democracies function, you can’t just have a
return on capital and no return on labor. We can talk about it, but that 60% has to be revalued.
In my own simple way, maybe you can call it naive, we'll start valuing different types of
human labor. What is today considered
56:00 - 56:30 high-value human labor may be a commodity.
There may be new things that we will value. Including that person who comes to me and
helps me with my physical therapy or whatever, whatever is going to be the case that we value,
but ultimately, if we don't have return on labor, and there's meaning in work and dignity in work
and all of that, that's another rate limiter to any of these things being deployed.
On the alignment side, two years ago,
56:30 - 57:00 you guys released Sydney Bing. Just to be clear, I
think given the level of capabilities at the time, it was a charming, endearing, kind
of funny example of misalignment. But that was because, at the time, it was like
chatbots. They can go think for 30 seconds and give you some funny or inappropriate response.
But if you think about that kind of system--that, I think to a New York Times reporter, tried
to get him to leave his wife or something--if
57:00 - 57:30 you think about that going forward, and you
have these agents that are for hours, weeks, months going forward, just like autonomous swarms
of AGIs, who could be in similar ways misaligned and screwing stuff up, maybe coordinating with
each other, what's your plan going forward so that when you get the big one, you get it right?
That is correct. That's one of the reasons why
57:30 - 58:00 when we usually allocate compute, let's allocate
compute for what is that alignment challenge? And then more importantly, what is the runtime
environment in which you are really going to be able to monitor these things? The
observability around it? We do deal with a lot of these things today in the classical
side of things as well, like cyber. We don't just write software and then just let it go.
You have software and then you monitor it. You monitor it for cyber attacks, you monitor
it for fault injections, and what have you.
58:00 - 58:30 Therefore, I think we will have to build
enough software engineering around the deployment side of these, and then inside the
model itself, what's the alignment? These are all, some of them are real science problems.
Some of them are real engineering problems, and then we will have to tackle it.
That also means taking our own liability in all of this. So that's why I'm
more interested in deploying these things in
58:30 - 59:00 where you can actually govern what the scope of
these things is, and the scale of these things is. You just can't unleash something out there in
the world that creates harm, because the social permission for that is not going to be there.
When you get the agents that can really just do weeks worth of tasks for you, what
is the minimum assurance you want before you can let it run a random Fortune 500?
I think when I use something like Deep Research,
59:00 - 59:30 even, the minimum assurance I think
we want is before we especially have physical embodiment of anything, that I
think is kind of one of those thresholds, when you cross. That might be one place.
Then the other one is, for example, the permissions of the runtime environment in
which this is operating. You may want guarantees that it's sandboxed, it is
not going out of that sandbox.
59:30 - 60:00 I mean, we already have web search and
we already have it out of the sandbox. But even what it does with web search and
what it writes -- for example to your point, if it's just going to write a bunch of
code in order to do some computation, where is that code deployed? And is that
code ephemeral for just creating that output, versus just going and springing
that code out into the world?
60:00 - 60:30 Those are things that you could, in
the action space, actually go control. And separate from the safety issues, as you think
about your own product suite, and you think about, if you do have AIs this powerful, at some
point, it's not just like Copilot- an example you mentioned about how you were prepping
for this podcast- it's more similar to how you actually delegate work to your colleagues.
What does it look like, given your current suite, to add that in? I mean, there's one question about
whether LLMs get commodified by other things.
60:30 - 61:00 I wonder if these databases or canvases or
Excel sheets or whatever -- if the LLM is your main gate point into accessing all these things,
is it possible that the LLMs commodify Office? It's an interesting one. The way I think
about the first phase, at least, would be: Can the LLM help me do my knowledge work using
all of these tools or canvases more effectively?
61:00 - 61:30 One of the best demos that I've seen is a doctor
getting ready for a tumor board workflow. She's going into a tumor board meeting, and the first
thing she uses Copilot for is to create an agenda for the meeting because the LLM helps reason about
all the cases, which are in some SharePoint site. It says, "Hey, these cases -- obviously, a tumor
board meeting is a high-stakes meeting where you
61:30 - 62:00 want to be mindful of the differences in cases
so that you can then allocate the right time." Even that reasoning task of creating an agenda
that knows how to split time- super. So, I use the LLM to do that. Then I go into the meeting,
I'm in a Teams call with all my colleagues. I'm focused on the actual case versus taking notes,
because you now have this AI copilot doing a full transcription of all of this. It's not just a
transcript, but a database entry of what is in
62:00 - 62:30 the meeting that is recallable for all time.
Then she comes out of the meeting, having discussed the case and not been distracted
by note-taking. She's a teaching doctor; she wants to go and prep for her class. And so she
goes into Copilot and says, "Take my tumor board meeting and create a PowerPoint slide deck out of
it so that I can talk to my students about it."
62:30 - 63:00 So that’s the type. The UI and the scaffolding
that I have are canvases that are now getting populated using LLMs. And the workflow itself is
being reshaped; knowledge work is getting done. Here's an interesting thing: If someone came to me
in the late '80s and said, "You're going to have a million documents on your desk," I would say,
"What the heck is that?" I would have literally
63:00 - 63:30 thought there was going to be a million physical
copies of things on my desk. Except, we do have a million spreadsheets and a million documents.
I don’t, you do. They're all there. And so, that's what's
going to happen with even agents. There will be a UI layer. To me, Office is not
just about the office of today; it's the UI layer for knowledge work. It'll evolve as the
workflows evolve. That's what we want to build.
63:30 - 64:00 I do think the SaaS applications that
exist today, these CRUD applications, are going to fundamentally be changed because
the business logic will go more into this agentic tier. In fact, one of the other cool things
today in my Copilot experience is when I say, "Hey, I'm getting ready for a meeting
with a customer," I just go and say, "Give me all the notes for it that I should
know." It pulls from my CRM database, it pulls from my Microsoft Graph, creates a composite,
essentially artifact, and then it applies even
64:00 - 64:30 logic on it. That, to me, is going to transform
the SaaS applications as we know of it today. SaaS as an industry might be worth hundreds
of billions to trillions of dollars a year, depending on how you count. If really
that can just get collapsed by AI, is the next step up in your next decade 10X-ing
the market cap of Microsoft again? Because you're talking about trillions of dollars...
It would also create a lot of value in
64:30 - 65:00 the SaaS. One thing we don't pay as much
attention to perhaps is the amount of IT backlog there is in the world.
These code gen things, plus the fact that I can interrogate all of your SaaS
applications using agents and get more utility will be the greatest explosion of apps, they'll
be called agents, so that for every vertical,
65:00 - 65:30 in every industry, in every category, we're
suddenly going to have the ability to be serviced. So there's going to be a lot of value. You can't
stay still. You can't just say the old thing of, "Oh, I schematized some narrow business
process, and I have a UI in the browser, and that's my thing." That's ain’t going to
be the case. You have to go up-stack and say, "What's the task that I have to participate in?"
You will want to be able to take your SaaS
65:30 - 66:00 application and make it a fantastic agent
that participates in a multi-agent world. As long as you can do that, then I
think you can even increase the value. Can I ask you some questions
about your time at Microsoft? Yeah.
Is being a company man underrated? So you've spent most of your
career at Microsoft, and you could say that one of the reasons you've been able to add so much
value is you've seen the culture, the history, and the technology. You have all this context by
rising up through the ranks. Should more companies
66:00 - 66:30 be run by people who have this level of context?
That's a great question. I've not thought about it that way.
Through my 34 years now of Microsoft, each year I felt more excited about being at
Microsoft versus thinking that, oh, I'm a company person or what have you. I take that seriously,
even for anybody joining Microsoft. It's not like
66:30 - 67:00 they're joining Microsoft as long as they feel
that they can use this as a platform for their both economic return, but also a sense of purpose
and a sense of mission that they can accomplish by using us as a platform. That's the contract.
So I think yes, companies have to create a culture that allows people to come in and
become company people like me. Microsoft
67:00 - 67:30 got it more right than wrong, at least in
my case, and I hope that remains the case. The sixth CEO that you’re talking about, who’ll
get to use the research you’re starting now, what are you doing to retain the
future Satya Nadellas so that they're in a position to become future leaders?
It's fascinating. This is our 50th year, and I think a lot about it. The way to think about
it is, longevity is not a goal; relevance is.
67:30 - 68:00 The thing that I have to do and all 200,000 of
us have to do every day is: Are we doing things that are useful and relevant for the world as we
see it evolving, not just today, but tomorrow? We live in an industry where there's no franchise
value, so that’s the other hard part. If you take the R&D budget that we will spend this year,
it’s all speculation on what's going to happen
68:00 - 68:30 five years from now. You have to basically go
in with that attitude, saying, "We are doing things that we think are going to be relevant."
So that's what you have to focus on. Then know that there's a batting average, and you're not
going to get- you have to have a high tolerance for failure. You have to take enough
shots on goal to be able to say, "Okay, we will make it to the other side as a company."
That's what makes it tricky in this industry.
68:30 - 69:00 Speaking of- you just mentioned that
you're two months away from your 50th anniversary of Microsoft’s founding. If you
look at the top 10 companies by market cap, or top 5, basically, everybody else but
Microsoft is younger than Microsoft. It's an interesting observation about why
the most successful companies often are quite young. The average Fortune
500 company will last 10 to 15 years.
69:00 - 69:30 What has Microsoft done to remain relevant for
this many years? How do you keep refounding? I love that, Reed Hoffman uses that term,
"refounding." That's the mindset. People talk about founder mode, but for us mere
mortal CEOs, it's more like refounder mode. To be able to see things again in a fresh way
is the key. To your question: can we culturally
69:30 - 70:00 create an environment where refounding becomes
a habit thing? Every day we come in and say, "We feel we have a stake in this place to be
able to change the core assumptions of what we do and how we relate to the world around us. Do
we give ourselves permission?” I think many times,
70:00 - 70:30 companies feel over-constrained by either
business model or whatever. You just have to unconstrain yourself.
If you did leave Microsoft, what company would you start?
Company I would start? Man. That’s where the company man and me sort
of says, “I'll never leave Microsoft.” If I were thinking of doing something, I think
picking a domain that has... When I look at the
70:30 - 71:00 dream of tech, we've always said technology is
about the biggest, greatest democratizing force. I feel like finally, we have that ability.
If you say those tokens per dollar per watt is what we can generate, I would love
to find some domain in which that can be applied, where it is so underserved.
That's where healthcare, education...
71:00 - 71:30 Public sector would be another place. If you take
those domains, which are the underserved places, where my life as a citizen of this country or
a member of this society or anywhere, would I be better off if somehow all this abundance
translated into better healthcare, better education, and better public sector institutions
serving me as a citizen? That would be a place.
71:30 - 72:00 One thing I'm not sure about, hearing
your answers on different questions, is whether you think AGI is a thing. Will there
be a thing which automates all cognitive labor, like anything anybody can do on a computer?
This is where I have a problem with the definitions of how people talk about it. Cognitive
labor is not a static thing. There is cognitive labor today. If I have an inbox that is managing
all my agents, is that new cognitive labor?
72:00 - 72:30 Today's cognitive labor may be automated.
What about the new cognitive labor that gets created? Both of those things have
to be thought of, which is the shifting… That's why I make this distinction, at least
in my head: Don't conflate knowledge worker with knowledge work. The knowledge work
of today could probably be automated.
72:30 - 73:00 Who said my life's goal is to triage my
email? Let an AI agent triage my email. But after having triaged my email, give me a
higher-level cognitive labor task of, "Hey, these are the three drafts I really want you
to review." That's a different abstraction. But will AI ever get to the second thing?
It may, but as soon as it gets to that second thing, there will be a third thing.
Why are we thinking that somehow, when we
73:00 - 73:30 have dealt with tools that have changed what
cognitive labor is in history, why are we worried that all cognitive labor will go away?
I'm sure you've heard these examples before, but the idea that horses can still be good for certain
things, there are certain terrains you can't take a car on. But the idea that you're going to see
horses around the street, they’re going to employ millions of horses, it’s just not happening.
And then the idea is, could a similar thing happen with humans?
But in one very narrow dimension?
73:30 - 74:00 It's only 200 years of history of humans where
we have valued some narrow sort of things called "cognitive labor" as we understand it.
Let's take something like chemistry. If this thing, quantum plus AI really helped us
do a lot of novel material science and so on, that's fantastic to have novel material science
being done by it. Does that take away from all the other things that humans can do?
Why can't we exist in a world where there
74:00 - 74:30 are powerful cognitive machines, knowing that
our cognitive agency has not been taken away? I'll ask this question, not about you, but in a
different scenario, so maybe you can answer it without embarrassment. Suppose on the Microsoft
board, could you ever see adding an AI to the board? Could it ever have the judgment, context,
and holistic understanding to be a useful advisor?
74:30 - 75:00 It's a great example. One of the things we added
was a facilitator agent in Teams. The goal there, it's in the early stages, is can that facilitator
agent use long-term memory, not just on the context of the meeting, but with the context
of projects I'm working on, and the team, and what have you, be a great facilitator?
I would love it even in a board meeting, where it's easy to get distracted. After all,
board members come once a quarter, and they're
75:00 - 75:30 trying to digest what is happening with a complex
company like Microsoft. A facilitator agent that actually helped human beings all stay on topic and
focus on the issues that matter, that's fantastic. That's kind of literally having, to your point
about even going back to your previous question, having something that has infinite memory
that can even help us. You know, after all,
75:30 - 76:00 what is that Herbert Simon thing? We are all
bounded rationality. So if the bounded rationality of humans can actually be dealt with because there
is a cognitive amplifier outside, that's great. Speaking of materials and chemistry stuff,
I think you said recently that you want the next 250 years of progress in those fields
to happen in the next 25 years. Now, when I think about what's going to be possible in the
next 250 years, I'm thinking like space travel,
76:00 - 76:30 and space elevators, and immortality, and
curing all diseases. Next 25 years, you think? One of the reasons why I brought that up was, I
love that thing of, the industrial revolution was the 250 years. We have to take this entire change
from a carbon-based system to something different. That means you have to fundamentally reinvent
all of what has happened with chemistry over the last 250 years. That's where I hope we have
this quantum computer, this quantum computer
76:30 - 77:00 helps us get to new materials, and then we can
fabricate those new materials that help us with all of the challenges we have on this planet.
And then I'm all for interplanetary travel. Amazing. Satya, thank you so much for your time.
Thank you so much. It's wonderful. Thanks. Great, thank you.