CEO of Amazon Web Services: Cloud, GenAI, two pizza rule and customer obsession.
In Good Company with Nicolai Tangen · 6 Dec 2023
Source: https://shows.acast.com/in-good-company-with-nicolai-tangen/episodes/ceo-of-amazon-web-services-cloud-genai-two-pizza-rule-and-cu
Transcript: https://podcasts.opennash.com/shows/in-good-company-with-nicolai-tangen/ceo-of-amazon-web-services-cloud-genai-two-pizza-rule-and-customer-obsession-e5ecc3/ (via OpenNash Podcasts, machine-generated)

Hi everyone, and welcome to our podcast in good company. I'm Nicolai Tangen, the CEO of the Norwegian company, Norges Bank Investment Management. In this podcast, I talk to the leaders of some of the largest companies we are invested in so that you can learn what we own and meet these impressive leaders. Today, I'm speaking to Adam Selipsky, CEO of Amazon Web Services, the largest cloud provider in the world. We own over 1% of Amazon. This translates to 80 billion Norwegian kroner, or over 8.1 billion US dollars.

This means that each Norwegian owns 15,000 kroner in this company. Most of the internet is run on cloud-based services, but how does the cloud work? How do they protect it, and how is AWS competing with Microsoft and Google? Tune in. Now, the cloud is the backbone of everything digital, and you control 1/3 of this market. Now, you have customers such as Netflix, Apple, NASA, CIA, Tesla, Spotify, and so on. So, just to kick off with the basics here, what is the cloud?

Uh, the cloud is essentially the ability for companies of all sizes, all types, to run all their infrastructure in a way that feels infinite and like it has no specific place. That's why the you have the metaphor of the cloud. So customers, uh, companies used to have data centers, uh, they would buy or rent the data centers, they'd buy servers, they'd hook up networking, uh, they'd worry about, uh, keeping those, uh, servers up and running. And when AWS, uh, launched in 2006 and pioneered the cloud, all of a sudden, there was this, uh, illusion of infinite capacity.

Like you could just tap into compute capacity, you could tap into storage capacity, and it didn't matter to you where it was or how it worked, you just had to concentrate on serving your customers, building your applications, instead of worrying about, uh uh, about IT infrastructure.

So how much of the total IT in the world is now in the cloud?

You know, it's so it's hard to know exactly, but, um, I'd estimate probably somewhere in the realm of 10% of IT has moved to the cloud, and if people think it's a lot higher than that, because there have been a lot of dollars that have moved to the cloud, uh, but they forget that, uh, there are a lot of IT dollars spent in the world. There's maybe several $1 1 trillion of IT spend per year. And so, uh, even though a lot of dollars have moved, we're probably still only in the, you know, 10 15% range, and, uh, that that means there's, uh, a a a huge amount of, uh, workloads that are still gonna be moving to the cloud for

years to come. So actually, we're very early in this cloud migration.

Stupid question, but how do you actually move the data to the cloud? I assume you don't transfer it through Wi-Fi.

Well, it's a it's not a stupid question at all, it's it's a very important question, and, uh, data moves a lot of different ways, um, it depends on how much data you have and where you are. Um, sometimes people just transfer data, uh, across the internet, uh, just from, uh, the servers that they ha- have right across the internet, but often when we have, you know, big organizations with a lot of data, um, a a it re requires, um, you know, higher throughput to get the data up into the cloud, and also you you really wanna be careful about, uh, about security and privacy.

So, uh, many of our customers set up, uh, uh, uh, dedicated connections, essentially a a direct connect, uh, over a private connection to AWS, and that can be, again, just like all other aspects of the cloud, that can be scaled up or scaled down to whatever size it needs to be, and we'll be securely transferring, uh, data over those, you know, private encrypted direct connections to us.

Now, you've been very important in the startup world because you enable companies to kick off with a business without making huge investments in IT structure. So what are some of the ventures that you have backed?

Uh, it's, uh, y- you're right, it's, uh, when AWS started, it was really with a vision of enabling, you know, a anybody, even a kid in a college dorm room, to have access to the same, you know, powerful, uh, compute infrastructure that even the largest, uh, companies, the largest enterprises in the world had. And so, that what that really meant was developers, and it really meant a lot of startups, and, uh, to this day, startups are still, you know, kind of deep, uh, uh, deep in the DNA of who we think we are.

So, in the early days, uh, there were, uh, companies like Netflix, uh, s uh, Airbnb, Pinterest, uh, of course, they're no longer considered startups, those are all enterprises now, but they were all startups. And then later on came companies like, uh, like Stripe and CrowdStrike, uh, again, big companies now who, uh, who used to be startups, and, uh, we we're still at it. So, uh, if you look at unicorns, the, uh, startups who are worth over $1 billion, over 80% of the world's unicorns, uh, run on AWS, and it's a great, great privilege to work with all of those innovators.

Wow, well done. Now, if we take these data centers, they are immense, right? Tell us about one of the bigger ones. Uh, I mean, how large are they?

Well, uh, th- the a AWS, um, is is built from an infrastructure perspective on this concept of regions, which is really, really important. So we have 32 independent infrastructure regions around the world, uh, many in the US and Canada, uh, UK, uh, many in the EU, um, Australia, India, China, Japan, Korea, uh, et cetera, et cetera. So 32, you know, full-fledged regions around the world. And each of those regions has, you know, multiple what we call availability zones.

E- each one has at least 3 availability zones per region, and that's a really, really important set of concepts, s because it provides incredible resilience for customers. So each region is built to operate independently, uh, if you have, uh, any kind of problem in one region, it's not going to, uh, affect you in another region. And even inside of a region, if you architect so that your, uh, your resources run, uh, inside of multiple availability zones, uh, inside of a region, they're also built to, uh, to to to be available separately, you know, not to fail, uh, for the same reasons.

And so, uh, our customers can build very, very highly resilient, uh, applications with very high, uh, uh high uptime, and, uh, you know, each e- each availability zone could be just 1 data center, it's usually a pretty large data center, or an availability zone could actually be, you know, a a a conglomeration of of a number of big data centers all put together. It all depends on how much customer demand we have in the region for, uh, for those workloads.

But we have a very large, you know, very scaled, uh, worldwide infrastructure.

So let's say, you and I, we jump in a car and we'll go and visit a big one. How big is it?

Uh, d d it you you're gonna see a whole series of data centers, they're gonna be hard to find, by the way, we don't advertise them and put our names on them, uh, you know, security is more important than, uh, than marketing, uh, but, you know, we w- we we can have data centers with, uh, you know, many, many, many thousands of servers in them.

So we open the door, you and I, we go into the center. What do we see? What's what's in there?

Well, to start with, by the way, we'll have trouble getting in, so, my my badge, my employee badge won't work at the data centers, because I don't need to go inside of them every day. I would have to get special permission, and, uh, you know, our very security-conscious, uh, customers, you know, want to know that there are, uh, extremely effective, uh, security controls, both physical as well as virtual controls. But assuming we get permission and we actually go inside, uh, we'll have to get past, you know, physical barriers to get inside, we'll have to get past, you know, significant security to go inside.

Uh, we wouldn't be allowed in to any, uh, sensitive area, but if we just pretend, uh, for a moment that we are, uh, we'd see, you know, many, many banks, uh, of uh, of servers, and, uh, really just a small number of human beings who are absolutely essential for keeping those servers running, for, uh, replacing any servers, uh, or disk drives, for example, with or or or power supplies with with issues, and, uh, i it it would look, uh, uh, hopefully, very clean and very orderly, hopefully, it would look very boring, because when everything's operating very smoothly, there's actually nothing to see, which is the best state of affairs.

Yeah, now I assume that when CIA trusts you with the data, it's probably pretty secure. How much does the data demand change during the year or during the week? I mean, something like Black Friday, when there's huge retail demand, what kind of fluctuations are you seeing overall?

O- overall, uh, AWS has, uh, actually very smooth demand patterns, and, uh, that's, uh, an effect of the scale at which we operate, uh, and that's one of the huge advantages we we offer our customers. So, uh, if you're a retailer, as you alluded to, um, in the the December holiday season or Black Friday leading up to it, you're gonna have a huge spike. If you're a tax company in the US, you're gonna have a huge spike in April when taxes are due.

If you're a news organization, you probably have a big spike, you know, first thing in the morning, uh, and maybe at lunchtime, uh, et cetera, et cetera. Um, and basically, with the scale at which AWS, uh, operates, we're we're over a 90 billion dollar a year revenue organization, again, with those, um, 32 full infrastructure regions around the world, um, we we essentially aggregate all of that demand. So when one company has high demand, another has low, and then they the they flip, and so that makes our infrastructure very, uh, very efficient and, uh, lowers our cost structure because we're making high utilization all of all of the, uh, infrastructure resources, and that, in turn, allows us to, uh, charge

really low prices for customers and to keep keep dropping prices to customers. We've lowered prices over 120 times, uh, in the past 17 years, uh, since we launched. But really, our ability to to manage all those peaks and valleys for customers, uh, throughout the course of their day, their week, their month, their year, um, i is one of the benefits that we provide to them.

And so, just for the listener, we have all of our operations in the cloud with you guys. We pay roughly 70,000, 000 Norwegian kroner per year for that service. Now, you were one of the pioneers in the in the cloud services, but now you have competitors such as Microsoft, Google, and they have been gaining ground and actually are growing slightly faster than you. So how are you competing with them? How do you win or lose business?

Well, I mean, we've had robust competition, uh, for a long time, a as we should, by the way. And, uh, I like to say, if you look to your left and you look to your right, and you don't see anybody next to you, you you may have dramatically overestimated the attractiveness of your market segment. Uh, and i it is a it is a fast-growing and attractive segment, and so, um, customers appropriately have, uh, have a lot of choices, which we think is is good for customers.

And by the way, I think at the end of the day is good for us. It keeps us sharper and makes sure that we we keep our focus on, uh, really what we need to do to to delight customers. So I I think that, uh, AWS, as you alluded to, uh, was the pioneer of the cloud. We launched in 2006, and, um, lots of folks scratched their heads and said, What does this have to do with selling books?

I can't I can't tell you how many times I got asked that question. And of course, the answer is it has nothing to do with selling books, but the technology that we built to sell books has everything to do with our exposing that same technology for, uh, many, many other organizations to use. But because it came from Amazon, that some people said, Oh, they're just a bookseller, and, you know, nobody will use this, and then startups began using it, and they said, Well, no enterprise will ever use it, and then enterprises started using it, and they said, Well, nobody will ever run mission-critical applications on it, and, uh, then we started doing things like, uh, powering Netflix and, uh, controlling,

uh, NASA's Mars Rover, you know, SUV-sized vehicles on the surface of Mars, and people started to get the idea that mission-critical workloads were indeed running in the cloud. And so I think we just proved ourself over a number of years, uh, we also moved very, very fast, uh, on product innovation, and it was probably 5 to 7 years, uh, before, uh, other companies really started to come out with even vaguely, uh, you know, competitive, uh, offerings.

And, uh, our founder, Jeff Bezos, has actually publicly called that 5 to 7 year head start the largest gift in business history. Now I have no idea if that's true or not, but it's just an interesting concept.

What does he mean by that?

Well, it means that we we had 5 to 7 years to build a large array of, you know, really powerful, really important services, uh, to gain traction, uh, to get a customer base, most importantly, to gain knowledge, to gain experience about what it took to operate really, really well in this new business.

But why were the other people so slow?

I think because we were so threatening to their business model, they didn't wanna believe it. If you look at the old guard of technology companies, particularly software companies, uh, in some cases, hardware companies, for 30 or 40 years, um, they, uh, they overcharged customers, uh, they intentionally locked them in so they made it hard to leave, and, uh, w- we really offered a completely new model of not having to commit, um, being able to, uh, elastically, you know, use as much or as little as you want, uh, to see your prices go down continuously, not going up continuously, uh, all while being highly available, highly secure, and and highly reliable.

Um, a and so it was it was really threatening to those high prices and high margins by by these other, uh, technology companies, and so it was easier to find excuses to dismiss us as not knowing what we were doing rather than to see the enormous customer value that was being created. Now, at some point, it it it became, uh, obvious and inevitable, uh, to see w- how customers were voting, and, of course, we had other customers get into, uh, uh, this this market segment.

And, you know, over time, you give it enough years and you're, you know, a competent technology company, there there are absolutely other, um, uh, other offerings out there, and, um, you know, we we try not to be naive and to understand what competitors are doing, but we still to this day, you know, focus the vast, vast majority of our time, our effort, our mental energy on, uh, being customer-obsessed and figuring out, you know, what customers need us to do as opposed to having a competitively driven strategy.

How does artificial intelligence change the landscape? Are all the language models now trained in the cloud?

I I I think AI is well, to start with, AI is enabled by the cloud. I mean, you know, re really, in the long run, if you look at the, uh, the amount of compute resources which are required both to train, uh, these models as well as to run inference or or or run these models in production, um, the, uh, the the capacity requirements and and therefore the, uh, the potential costs, uh, can can be really enormous.

And, uh, those will happen in the cloud because the economics and the, uh, the the capacity availability will be much, much higher in the cloud than elsewhere. So I really view, uh, generative AI and and AI in general as, uh, the next big, big thing that is happening in the cloud. Now it's not just one thing, uh, I mean, generative AI is going to, you know, fundamentally change, you know, many, many technologies will fundamentally impact probably every application that you and I interact with, both in our professional and our personal lives.

So it's fundamental, but it's not a it's not a different thing. It is the next huge thing that will happen, you know, in the cloud, and, uh, it's very much tied to your data strategy. So we have all of these, uh, uh, companies who have, uh, built data platforms on AWS and have vast amounts of data in the cloud, and that data is what helps them differentiate with generative AI. There are there are these models there, but we what you've gotta do is bring your data to the models, whether it's your customer service data, your investor data, uh, your, uh, uh, medical research data, you know, whatever it is, and and use that data to your advantage, you know, using

these new powerful, uh, large language models, and, uh, uh, helping com companies to really marry together their data and their generative AI strategies, we think, is what's gonna unlock, uh, the real power, the real value.

And is that why you now invest billions in Anthropic, which is competing with OpenAI?

Yeah, o o op um, uh, Anthropic, um, used AWS from the very beginning when they were founded, uh, a few years ago, and, uh, a as our relationship with them has deepened, um, I I think that, uh, we provided a a a really great opportunity for them to, um, to get access to massive amounts of compute capacity, which they need to train their Claude models. They're now on Claude 2, which is a a a really leading edge model, it's as, uh, as good as anything in the world, and at the same time, uh, they're really smart people, and we saw that there was an opportunity to collaborate with them and have them, uh, help us to optimize and improve, um,

uh, the the hardware, the, uh, the the the custom design chips that AWS, um, uh, designs, and, uh, we decided to deepen that partnership together, and as one element of it, we also did invest, uh, uh, i into Anthropic. So it's a tight partnership, it's a special partnership, Anthropic, um, now uses AWS as its, uh, primary, uh, cloud provider, and, uh, I think we'll be partnering together for a long time to come.

You got competitors in the US and China, but no European competitor. Why is Europe so slow here?

Um, I I think it's true that if you just look, you know, globally, um, uh, the the the largest, uh, public cloud providers, um, have been US-based companies, and as you said, um, e- especially inside of China, um, Chinese-based companies, um, but there are, uh, there are Eur- you know, smaller European-based, uh, competitors as well, and, um, I think, uh, you know, it's very dynamic and, uh, a quickly moving space, and, uh, I think there are opportunities for for companies from, um, uh, from all countries if they got a a a great value proposition to, uh, um, to, uh, gain market segment share.

So, you know, for now, this is the way it's played out, but, uh, uh, you know, we I I think, uh, we all need to stay very alert.

Well, you're very diplomatic, Adam, but when we look at many different industries, Europe is a bit behind in innovation, in technology, and so on. Now, wh- why do you think it is? Because when I look into some of your business principles, you have words like big, fasts, and so on, and these are not very European slogans.

Well, eh, we we we partner with a great number of European companies, whether it's systems integrators like, uh, Capgemini, um, whether it's software companies like, uh, uh, Sage, uh, in the UK, that I just, uh, spoke with last week, and I see a lot of innovation, uh, happening in Europe. Um, I I think, you know, a lot's been written and said about, uh, how venture capital works and how, uh, h- how much funding, you know, happens of companies, uh, i in the US, and I think a lot of the VCs are are looking more internationally these days, so I think there's there's a good chance that, uh, the innovation continues to to grow and flourish in a lot of different

countries around the world. Uh, you know, India's another place where people are really looking, uh, uh, l- looking at for innovation to come out of. Um, so, a again, I think it's very dynamic and, um, I think there have been great conditions for for technical innovation in the US for sure, uh, but I think you'll see it in other countries over time as well.

Now, these data centers are very energy-intensive. How do you plan for a more sustainable future?

Well, that is an incredibly important, it is is one of the most, uh, in ques- uh, important topics, um, that we could touch on. So, uh, uh, Amazon has taken a very public, uh, pledge to be, um, net zero carbon across all of Amazon retail, AWS, everything, uh, by 2040, uh, which is 10 years ahead of the Paris Accords and is a very, you know, big, bold, uh, public commitment. We did it publicly instead of privately because we wanted to, uh, both put a forcing function on ourselves, frankly, because it's difficult, uh, as well as to really catalyze and help inspire other organizations, other companies to join us in, uh, uh, helping to fight f- uh, fight climate change.

We have over 400 other organizations, including many large companies, who have signed on to the climate pledge, so we're we're really gratified by that. Now, we can't just wait till 2040, and we have a lot of interim, uh, goals along the way. One of which is to be 100% renewable energy powered by 2025. Now, obviously, that is just around the corner, and, uh, the good news is we are already over 90% renewable energy powered across all of Amazon, but a huge part of that, uh, of course, comes from, um, AWS, uh, data centers.

So we're, uh, already at 90% on our way to 100%, and one of the big ways we do that is by being, uh, the largest corporate purchaser of renewable energy on Earth over the past 3 years. And, uh, it's we've we've, uh, uh, done long-term funding for wind farms, for solar projects around the world, um, including, you know, the 1st offshore wind in Japan project and, uh, many others like it. I it's also really important that we innovate to just use less energy, period.

And, uh, uh, I mentioned custom design chips a little while ago. So, uh, the 1st, uh, a a AWS, we we certainly buy a lot of a lot of chips for our servers from a l- a lot of really great partners. But in addition, uh, for a number of years now, AWS has actually designed its own chips, uh, which customers can use in, you know, different different servers, virtual servers, and, uh, that general-purpose chip is called Graviton.

Our our Graviton-based, uh, um, uh, chip, um, actually consumes 60% less power than equivalent, uh, equivalent chips that our customers use. So, simply by moving into the cloud and then moving onto our Graviton-based capacity, uh, customers can, you know, uh, reduce their energy consumption by up to 60%.

If you look at the chips that you buy, how much do you now develop yourself, and how much are you still dependent on the likes of Nvidia?

Um, I I I think it's a big mix, um, you know, we we have in the kind of classical x86-based, uh, chip architecture, we're great partners with Intel, uh, with AMD, uh, in the GPU space, uh, we're great partners with NVIDIA and have a very tight relationship with NVIDIA. Um, so we we continue to, uh, to buy and use, you know, enormous quantities of chips from all of those, uh, all of those companies, and and I don't I don't foresee any change in that, uh, in the foreseeable future.

How could quantum computing change your business?

Um, I I think quantum computing is is is, uh, potentially, you know, very disruptive, very exciting, is also, I think, a lot of unknowns, and it's it's years away. So we're we're investing in quantum now, uh, both internally inside of Amazon as well as we've, uh, uh, funded a lot of research at, uh, uh, places like, uh, Harvard University, uh, Cal Poly, um, uh, Caltech, and, uh, we'll continue to do so. I I th- I think that, um, it it's a a a proposition which, you know, in a number of years from now, uh, could really be disruptive to, uh, to how computing is done.

Uh, but there are a lot it's very, very early, there are a a a lot of unsolved questions which which need to be solved befo- before it becomes a a viable technology at scale.

But when it is a viable technology, how is that world going to be different? What's going to be the main change?

Uh, the main change would be, uh, the ability to do, you know, vastly, uh, larger, more powerful, uh, computing jobs at, uh, much faster and at much lower cost. Uh, I mean, that is the that is the dream. Um, and, uh, I I think there's a good chance that the world gets there, but, uh, there's still s to be a lot of of scientific research, uh, as well as, uh, uh, applied, um, uh, applied computer science to to make that happen, and we're investing in it.

I know a lot of other companies, uh, are as well, and, uh, I think it's a great, uh, uh, uh, an area of great potential for the world, but, uh, at the same time, as we, uh, keep that eye towards the the future, uh, we're keeping the other eye on the technologies we need today, uh, to improve price performance for our customers.

And, in your opinion, how do we stack up against China, both on the AI side, but also on quantum computing?

Um, well, on quantum, I'd say it's it's it's too early to really answer that question, uh, I think it's it's it's hard to know, um, actually, who's p- who's proceeding at what pace. It's it's still very early. Uh, on AI, uh, I mean, again, to the best of my knowledge, I'll say, uh, I think that, uh, uh, the US and the Western world is is proceeding really well. I think that, um, you know, there are great great leaders, uh, in the AI space, and what I really like is that some of them are are big companies like Amazon, um, and some of them are, uh, very, very young startups, uh, like Anthropic, you know, or Cohere, or Stability AI, or

AI21, all of whom are are are great, uh, AWS startup partners. Hugging Face is another one, and, uh, i it's, uh, I I think that mix is really important for innovation, having the resources of big companies along with the fresh thinking of, uh, of of very young, uh, smaller companies, and so I think that, uh, uh, the the the US and the Western world is is very well positioned. I think we have a lot of questions around not only where the technology will go, but how you think about responsible AI and ethical AI and making sure that, um, these capabilities are used in the right way, and, uh, we're we're we're busy at work, um, as you can imagine, in

in in all of those areas, you know, technical, uh, out there in the market with customers as well as, uh, very intently with, um, regulators and policymakers.

Do you think the tech companies are able to self-regulate AI?

Uh, I think tech companies it's very important that tech companies, you know, participate very vocally in the discussion. I don't think it'll only be tech companies self-regulating. I think that, uh, it will be, uh, partially up to governments and potentially multinational organizations, uh, to provide guidance and to provide regulatory frameworks, and I think countries should, um, have a voice in saying, Here's how we want AI to proceed, uh, inside of our our country, and I think that, um, tech companies have a lot of experience.

We're the ones with the technology and with the customers, so we have really important voices in that conversation, and we're helping very actively to shape those conversations. You know, I was at the White House this summer, uh, when President Biden announced, uh, the his, uh, voluntary commitments around AI, which about 7 companies signed up for. Um, later this week, I will be in the in the UK at the, uh, AI Safety Summit that the UK government, uh, is putting on, which I have a lot of, uh, world leaders at, and, uh, we're we're devoting the time and the resources to participating in those forums, participating in those discussions, and I think that, uh, industry needs to come together along with,

uh, academia as well as policymakers and regulators, uh, to figure out with this powerful set of new technologies, uh, how we can, uh, ensure, you know, safe, responsible, uh, uh, use with a lot of visibility into it, while at the same time, and this is essential, not q- squashing innovation. We've got to let companies keep on innovating, uh, particularly as some of the companies, ye- uh, some of the countries you mentioned, uh, are probably not going to worry so much, are gonna keep on innovating, uh, and so we have to make sure that, uh, our our countries and our companies keep on innovating unfettered as well.

Talking about innovation, you are also launching a satellite network to provide broadband akin to what Starlink is doing. So what is it? You launched a few satellites, you're looking at several thousands of them eventually. So so what is it? What is this going to do?

Well, uh, the, uh, it is very exciting. You're referring to Project Kuiper with a K, and, uh, it's a a whole satellite, uh, capability being launched actually by a sister division to AWS. It's it's being spearheaded by another part of Amazon, uh, although AWS partners, uh, very closely with them. So, uh, we will, over the next, uh, several years, have several thousand, uh, LEO or low Earth orbit satellites, uh, up around Earth. Um, just a few weeks ago, we launched our first 2 test satellites, which was a very exciting milestone.

Uh, those tests have gone really well, um, really as well or better than, uh, than what one could have hoped, and so, uh, those teams are gonna, uh, keep pressing forward, you know, really urgently and relentlessly and and as fast as possible, um, uh, getting the first production satellites up, uh, uh, next year and then moving into, uh, commercial availability for customers after that. Um, and the the the first and primary mission of Project Kuiper was to provide internet access to unserved and underserved, uh, people around the world, and there are billions of those people who are either unserved or underserved.

And, uh, I'm really excited for that mission. I think over the next several years, it's gonna be really exciting. We're obviously still early in that journey, but, uh, Amazon takes a very, you know, long-term view of most things, and, uh, we think it'll be an amazing set of offerings for customers for many years to come.

So that's a great place to pivot to your corporate culture because you are one of the longest term thinking companies I know. What kind of time frame do you operate within when you plan?

Yeah, it it's true. I do think that's one of the, um, most distinguishing things about Amazon is we we do tend to take very, uh, long-term views in our planning horizons. And, uh, it it's such a it's such a tool for us. It's it's so powerful because instead of looking at things on a, you know, a 90-day cycles, a quarterly basis, you know, we're thinking about, you know, what can we do 3, 5, 10 years from now to serve customers really well and to create great businesses for ourselves and to make to make decisions based on those criteria.

And a lot of companies, um, choose not to have that long-term viewpoint. I I think it's very limiting.

But what is it that makes people think short-term?

I I I think, um, uh, there's a lot of investor sentiment around, you know, quarterly, uh, er- earnings in the US, for example, and there's tremendous pressure, and I I I think that if you have a culture where, um, you know, you care a a lot about, um, uh, the the optics, short-term optics, uh, and you, uh, you just don't focus yourself as much on creating long, long-term value for customers, then you kinda end up in that trap.

Now, I'm not saying short-term results don't matter, you know, they do. And, uh, you know, having having targets and and hitting them, you know, matters, but, um, you know, really building building a business for the long term is what matters the most, and I think we've done a good job over time about, you know, trying to run a good business in the short term, uh, while really building, you know, enormous value for the long term.

Well, Sam Altman from OpenAI said that long-term thinking is a competitive advantage because so few people do it. What are some of the longest-term investments you've made at Amazon?

Well, we've talked about Kuiper and, uh, you know, that, uh, that probably started, call it 5 years ago, and, uh, i it's we're we're just at this exciting stage of starting to launch satellites, and it'll be another couple years before that full constellation is is up, so, uh, that is very capital-intensive for sure and and long-term proposition. Uh, I think we've been talking for the past, uh, 30 or 40 minutes about a a really good example, which is AWS.

If you go back in time, um, you know, Amazon was was, you know, mainly a a retailer, uh, had a few retail categories, and then there was this crazy idea about, uh, you know, turning Amazon inside out, offering up the guts of Amazon, uh, for other companies to use, and, uh, a lot of people didn't understand it, other people, uh, thought it was crazy, and it was definitely a long-term proposition because it requires, uh, uh, not only building all of the software services that we that we offer customers today around compute and storage and database and machine learning and AI and, uh, call center operations and everything in between, uh, but also, uh, spending all of the the capital, you

know, required for the the data centers and the servers and all of that infrastructure that we've already discussed. And, um, you know, we knew it would it would take years, uh, for it to become, um, a a big business and and a profitable business, but at the same time, you know, we believed and still believe that i in time, in the fullness of time, it could be the biggest business that Amazon has.

It's probably the majority of the value of your company now, isn't it? If you look at the whole Amazon?

Har- hard to say. I I think that all of the pieces of Amazon work very well together. I I leave it to the, uh, to the investment community to try and sort out, you know, which pieces are, uh, are worth what, but, uh, we we see many, many companies now who want t to bid do business with all of the different areas of Amazon. So, you know, automotive companies who want to do business with AWS and integrate our our devices and entertainment systems into the vehicles and supply Amazon with, uh, uh, with electric, um, uh, delivery vehicles, and that's just the automotive industry.

So I I think all the pieces of AW- of Amazon work very, very well together.

But also Amazon retail business, which is where the whole thing started, was running for ages on negative and very thin margins, right?

Well, certainly in the early days, uh, uh, the eh, uh, Jeff became very, uh, I think, well known for the mantra of, uh, get big fast. That was, you know, really, a term here at Amazon, and that preceded my time here at the company, by the way, but, uh, I I I think that was that was a really big thing, and, uh, the company really believed that, you know, achieving that scale in retail was gonna be, uh, really, really important, and it would help price, it would help customer selection, and it would help convenience and delivery speeds if they could, uh, get big fast, and and they did, and again, that took, uh, that took a very long-term perspective instead

of just, you know, worrying about, uh, uh, the very, very short term. Uh, and as a result, uh, you know, the retail business ha- has been great. You know, it is a retail business, and obviously, you know, uh, retail businesses, um, you know, uh, uh, you have to be, uh, appropriately cost conscious. Those folks have always done a great job of really investing for the future while also, you know, being being really, really efficient. And, uh, it's it's been, um, uh, it's been great to partner with them and see the results that they've achieved.

A lot of companies say they are customer-focused. Now, you really are customer-focused. How can we see that?

So, it's a great question because you're right, everybody does say they're customer-focused, and, um, I w- I I think that a lot of people think that customer focus is somehow an emotional attribute. E- uh, and on this emotional scale, you'll go from, you know, maybe disliking your customers, we could all point to a few cust- a few companies who don't seem to like their customers very much, to maybe liking your customers, and then the highest on that scale would be, you know, loving your customers.

And I I think, uh, in my opinion, um, that is not what customer obsession, uh, is really about. Although I think loving your customers is a good thing. But at Amazon, and I I think the reason why, uh, we actually are so customer-obsessed, I I think that what it means is 2 things. 1, understanding customers at an incredibly deep level, a level that is hard to achieve, a level that takes effort continuously, and most companies simply, you know, don't go through all of the energy, don't expend the energy and go through the effort required to understand, you know, on each one of their services, each use case, case, what do customers really need from you?

What a what real problems do they have? And then, uh, as hard as that is, that actually turns out to be the the easier of the 2 things. Is the 2nd is to ensure that every important decision you make is taken from that customer perspective. I've seen a lot of companies, you know, outside of here who they actually understand a fair amount about the customers, but then when they decide to price something, it's just about profit maximization, or if they decide whether to build something, it's about, well, you know, can we afford to build it?

And they kinda drop the customer perspective outside the door, you know, for some weird reason, and, uh, I think you have to have the discipline to say, we we not only have all this knowledge about what customers really need, we were gonna bring it into the room and make sure that we're basing our decision around that, and you see that e- embody itself in a lot of different ways at Amazon. It's not just talk, it's not just a philosophy.

Uh, one example is, um, we we have this this fairly well-known mechanism called a, uh, press release / FAQ process, PRFAQ, what that means is that, uh, before we, uh, before we build anything of substance, a new service, um, a new AWS feature, we'll actually write a full press release along with a full FAQ. Doesn't matter whether we ever actually issue that press release or not, uh, and the idea being that you need to be able to describe in in in detail, uh, why what you're building matters, how it's gonna delight customers, or why would you bother to waste your engineers' time on it, and that we call that our working backwards process, and it's a process that, uh, I've

actually got a number of customers who have seen, loved, and and adopted, and we actually, um, go in and do workshops with a number of our customers to to teach that technique. It's it's very powerful, and a lot of our customers have have really, uh, uh, uh, grasped it.

It's interesting. I spoke to some of the people who were in charge of the selection of AWS as a partner, and as you can imagine, we are a government entity entirely. So these processes are really thorough and can take a long time.

As they should be.

But one of the really differentiating factors was that it was very, very clear to us that you really cared about us as a client, and that spirit has endured after we signed the contract, and you guys have really delivered, so well done.

Yeah, it it just seems so basic, uh, it's so important for AWS, you know, uh, I tell my teams all the time that, you know, your job is to go form long-term trusted relationships, and, uh, the business part will take care of itself if you do that. I I think so much springs from that, and, uh, people get really caught up in their own short-term, uh, concerns that it's easy to forget that, and, uh, we try here to make sure that we don't.

You have some really cool leadership principles which are publicly known. What is the most important one for you?

This is like choosing between your children, it's it's just not it's not fair. We have, um, we have 16 leadership principles that they're, uh, not just in some manual, they're alive and well. Uh, I sometimes call them the operating system of, uh, of Amazon or or the central nervous system. Uh, customer obsession is vital for sure. Uh, I I think the, uh, uh, another one which is absolutely core is invent and simplify. And, uh, that's really about innovating a and also, uh, embodies the principle that, um, innovating doesn't mean just making things more complicated.

It means, uh, uh, making them more powerful, but also more simple. And, uh, uh, you know, we w- we've always tried to, you know, move very fast as a company, and in fact, one of the reasons why AWS was founded was because Amazon wanted to move faster, and they wanted to create the shared layer of infrastructure services so all the other teams at Amazon could worry about their customers and not about infrastructure. And, uh, so we've I think we've built in many, many mechanisms to to move fast.

We we hire people who like to build, who are innovative. We organize for speed. Uh, a lot of people think speed is preordained. We think speed is a choice, and so we organize ourselves, both culturally as well as, you know, just our teams themselves, uh, to be able to innovate and to be able to do it really, really quickly, and so I I think that invent and simplify, uh, is incredibly important for customers, and it's actually been a a huge advantage for Amazon, uh, in the market segments we're in as well.

You have something called a 2-pizza team. What is that?

So a 2-pizza team. So, uh, the concept of that is, you know, wherever possible, you should have a team no larger than can be fed with 2 pizzas. And, uh, we've had a lot of teams at Amazon over the years who actually are 2-pizza teams, and even when, you know, with our scale, uh, you obviously have teams that, uh, that grow larger than that, and, uh, we very often try and refactor or slice those, uh, those teams up to create even smaller, more nimble teams.

Um, but even where we don't actually do that, it's it's the concept that's alive and well, and the concept is, you know, stay small, stay nimble, and very, very importantly, you know, control your destiny. Uh, you know, don't don't take dependencies wherever where you don't have to on many other teams, you know, figure out how to, uh, be able to to to own your own progress, uh, overcome your own obstacles, and make rapid progress for customers.

What's special about the meeting culture in Amazon?

Well, we do do something which is, um, uh, unusual, uh, which is we, uh, we pretty much never use, uh, PowerPoint to run an internal meeting. Um, so for most internal meetings, uh, somebody writes, uh, a a narrative, uh, a a document, and, uh, it could be shorter, it could be up to 6 pages, no longer than 6 pages, and when we come into a meeting, the first thing we do is say hello, and then we all sit down and read.

So we might spend the first 15 to 30 minutes of a meeting simply reading. And when everybody's finished reading the document, there's no presentation. Um, somebody just says, Okay, what questions do we have? And, uh, it's it's it's an incredibly powerful mechanism. It's to me, it is one of, like, the most powerful ways to, uh, uh, uh, to move the business forward because it creates real understanding. With with, uh, with PowerPoint bullets, it's it's so easy to be on the surface.

Um, it's so easy also to not get through the material because, uh, you just interrupt the presenter. With the narrative, you are forced. It is very obvious if you've if you have not thought through a problem problem deeply with a 4 page or a 6 page, uh, uh, narrative. Um, so you you you have to have depth of thought. You have to have been able to lay out, uh, all of the issues, uh, to have a good conversation.

And in addition, it's very egalitarian. By the end of reading the document, everybody has the same information. You know, nobody has special knowledge, and then, uh, anybody in the room, regardless of who they are, um, can ask questions, and, uh, it's all from the same fact base. And so, it's been, uh, uh, one of the most to me personally, one of the most amazing mechanisms that I've seen at Amazon, uh, to really help, uh, create deep understanding, to make decisions, and to move our businesses forward.

Moving on to leadership, what is the most challenging part of being the CEO of AWS?

Uh, I would say, well, I sometimes joke that I, uh, uh, I I have, uh, I fear that I'll wake up in the morning and and somebody will tell me that I work at a big company. Now, I I guess I I do happen to work at a big company, but, uh, I never want it to feel that way. I never want our customers to feel that way. And, um, you know, there's this con- concept of of, uh, insurgents versus incumbents, and, you know, in insurgents don't mind disrupting, insurgents, you know, just wanna create new things and wanna create value, and, uh, incumbents, um, wanna preserve value and worry about what they already have instead of worrying about what they

can build for customers. And I think, um, you know, it's very easy as you grow larger and you have revenue streams and you have businesses to slip into that incumbent mentality, and I think one of the most important jobs I have as a leader is to try and remind everybody at AWS that it is still day one. You know, it is very, very early in the cloud journey, um, that, uh, there's way more work to do than we have already done, and that, uh, our customers, you know, demand and expect us to be innovating on their behalf.

It's our job to do that and figure out how to create a good business for ourselves along the way, not to figure out, you know, how to create a good business and bring customers along with us. And I think you gotta have a lot of leadership, a lot of cultural mechanisms to continuously remind, you know, everybody across what is now a pretty scaled and pretty sizable team about, you know, h- how we believe we wanna operate and how we want our world to work.

You spent a few years outside Amazon as the CEO of Tableau. When you came back, how has your leadership changed? And also, how does your leadership change with age and experience?

Uh, I I think that, uh, over time, partially as you say, from age and experience, and and partially from just being able to step away and, uh, operating in a different environment and get getting a new look, uh, on, uh, uh, on the world. Uh, I I think one thing I've really tried to focus on over the past few years is, um, uh, leading with empathy. And, um, some people confuse, by the way, empathy with sympathy.

You know, v- very different. To me, empathy means understanding understanding. So, an, uh, understanding where people come from, understanding why they're saying what they're saying, instead of just focusing on the words coming out of somebody's mouth, trying to go back a step and understand, you know, are they saying this because they're worried about a program? Are they saying this, you know, what what's going on, or or what's going on with the person, the human being? Where where are they coming from?

And, uh, in order to do that, uh, I've tried to really, really focus, as obvious as it may sound, I've tried really, really hard to focus on listening more over the past few years.

Can empathy be learned? Is it an acquired skill?

Absolutely. I think it starts with your listening skills. So, uh, I've figured out over the past few years that, um, I don't learn that much when I talk. And that I learn a lot more when I listen. So, I I think they're they're very tactical things, very tangible things, I should say, like, uh, like listening.

And would your wife agree to this?

On some days, on my better day, and maybe maybe not on all days, but a good question. Um, so I I think absolutely it can be learned. I like to think that, um, you know, I've gotten better, uh, not perfect, but better and better at this, uh, over the years, and I think, um, if you have empathy and and really start and really listen and understand where people are coming from, um, that allows you to be, you know, really open and honest, and, uh, in return, and, you know, I th- my personal leadership style, I I I try to really be, uh, you know, open and honest and transparent and let people understand what I'm thinking and why I'm thinking

it. And, um, I I think that that really resonates with with most people, if if they feel that you've brought them along the discussion, you that you've brought them along the decision process. Uh, they've had a voice, they've been heard. There's there's been thoughtful discussion. I've had many examples where I I've had to make a decision, which, you know, a lot of people disagree with, and I've found that in the vast majority of cases, th- they don't like the fact that we've decided, you know, something that that they wouldn't have decided, but they're okay with it.

They're okay with it because of the process, they're okay because they feel respected, because they feel listened to, and they trust, uh, that their leaders are, um, even if not always, you know, in complete agreement with them, that they have the, uh, best intentions and that they're, uh, they're going through a thoughtful, appropriate process.

How do you deal with stress?

Stress? What stress? No, we we're in a very, uh, I I'm very privileged, we're in a incredible, uh, uh, business, I have an incredible position, I'm I'm very, very lucky, but, um, it is fast-paced, uh, of of course, there's there's stress. Um, I think in a couple ways. One one of which is at work, having a really, really good team and and knowing that I can, uh, I can talk to the people on my team. Um, people used to tell me, Oh, the CEO job is very lonely.

I I've never found it lonely. I think that's a choice. And I don't find it lonely because, uh, my senior leaders, you know, know what's going on. I don't keep things inside my own head. I share with them, um, they they help me, uh, they make my ideas better, they bring new ideas, which are, um, you know, usually better than mine, and, uh, we really operate as a team, and I think that reduces my personal stress at work.

And then, I think it's really important to set boundaries and to have time for, uh, whatever it is that's important to you. Um, for me, you know, my family's really important. Spending time with the family is important. Um, there's also some s uh, sporting things I love to do. I love to water ski in the summer. I love to play tennis year-round, and, um, you gotta carve out time for those things to to stay whole and to stay healthy.

Do you read?

I don't read nearly as much as I would like to, uh, I I I do read, but, uh, uh, unfortunately, at this particular stage, it's a a l- little more episodic, but, uh, I do love to read, and I've been, uh, been a lifelong reader since I was a little kid.

We have tens of thousands of young people listening to this. What is your advice to them?

Uh, I think advice to young people, um, well, I wish there was probably a lot of advice I should have been given when I was young, so, uh, what to choose. I would say, uh, maybe 2 things, um, 1 is, um, try and remember that you have a long life, hopefully, and a long career, and and I think people early in their career are are measuring themselves against their their peers that they went to university with or that they started a job with, and, you know, whether you do something 12 months earlier or later than I do, you know, it really doesn't matter.

Uh, what matters is the experience you gain, so just gain experience, build skills, you know, do things which you can see, uh, could be valuable to you, you know, down the line. And, uh, I promise that, um, you know, if you gain those skills, it really won't matter that, you know, 20 years from now, um, that I had some promotion 6 months before you did. Um, so just take that long-term view about skill-building, and the second thing I'd say, and I really wish that, um, you know, I had learned this earlier in my career, is this concept of being, um, vocally self-critical.

And, uh, most people are able to, you know, very quickly, uh, uh, critique somebody's shirt, uh, whether they like the coffee they're drinking that day, but when, uh, when when the camera turns on themselves, uh, they lose that objectivity because it's it's it's painful, and it makes you insecure, and I think if you can understand what you're really good at, what you're not so good at, uh, what you really wanna work on, if you have that clarity of understanding, uh, it's the old adage, you can't manage what you can't measure, and if you can't sort of understand yourself, how can you really hope to, uh, to know what to work on?

And I think if you can just be objective, uh, with yourself about what the things you wanna work on are, no matter how good you actually are at other things, um, that is an incredibly powerful skill.

Well, it's been really great talking to you. We've been investing in Amazon for more than 20 years, and we are just so grateful for all the value that you have created. And also, thank you for running our operations up in the cloud.

Well, it's a it's a privilege. Uh, we love working with you. We love working with all of our customers. Uh, we have got an incredibly passionate team, and, uh, we're just getting started, so we'll keep going.

Fantastic. Thank you so much.