CEOInterviews.AI
Start App
Andrew Jassy
President, Chief Executive Officer & Director, Amazon

‘We Have To Do It’: Amazon CEO Andy Jassy On Racing China In The AI Buildout

📅 Nov 17, 2026 Semafor 29 MIN 8737 VIEWS 60 SEGMENTS · 4 SPEAKERS
Moderator: Kate Rooney, Technology Correspondent, CNBC.

Questions asked in this interview

12
  1. 5:25What do you see in terms of how broad it is and what the potential is for this to expand and more people to actually start using it versus sort of this this bubble that we're in in Silicon Valley?
  2. 7:11Are there areas that you do worry about where you're building any sort of contingency plan within Amazon?
  3. 11:48Are you talking to other CEOs and are there sort of bad actors that have done this the wrong way and given the industry a bad name here?
  4. 13:12What is your take on what the Nvidia saga has meant for the industry and will mean for for the tech industry in the near term?
  5. 14:21Does it change the way that you think about where you can sell your chips and and the the potential for this business overseas, especially in China?
  6. 16:12Where do you stand on on all those arguments?
  7. 17:06Can you take us back to sort of that initial investment and what you saw in Anthropic and Dario Amodei at the time?
  8. 18:47But does anything about the relationship change now that they're a public company?
  9. 20:37What are your conversations like um with Sam and Dario on that side?
  10. 22:29Do you feel vindicated at all that you've sort of stuck with it?
  11. 25:59I mean having if you have these agents, this fleet of agents that's going out and doing price comparison, is it going to lower cost?
  12. 26:48What is the health of the consumer look like right now and what are you seeing in terms of spending habits?
Event Announcer 0:00 ↗
Please welcome Semaphor's Ben Smith.
Ben Smith 0:08 ↗
Appreciate the one person applauding. Um, um, it's great to see you all again this morning. Um, and and after just an amazing day of conversation. And um, you know, when when we when we started building this platform just basically about a year ago, it was based on this sort of intuition, this sense that that Silicon Valley and Glo and global capitals and and Washington, but but others as well, were on a kind of collision course. Um, last year basically felt like felt like the collision, which for those of us in the news business made it a incredibly kind of energizing and exciting moment to be having these conversations. But of course, the the story is so far from over and and and the thing that I have heard and have felt over the course of yesterday is just how inextricably linked at this moment political power is with the future of the tech industry and these huge these huge breakthroughs that we've been um, you know, that we've been covering. And so we feel really lucky to be here with you all at this particular moment.
Last yesterday we heard, you know, from we began the day hearing from Sam Alman talking about how Jensen Wong is his hero. Um, ended the day at a dinner where Calun al-Mubarak told us that winter is coming, but that he hopes it won't be like a negative 30°ree winter, just a a mild winter. Um, and had government officials from Singapore to Washington DC, you know, reacting to these very urgent crises that they're that they're managing right now. Um, so thank you all so much for for being here and we're we're excited for another day of conversations. Um, beginning this morning with CNBC's Kate Rooney in conversation with Andy Jasse and and and uh then end at 100 p.m. today. If you haven't seen yet on the program, Jeff Bezos will be here talking to our Liz Hoffman, hopefully also heckling Andy Jasse this morning. Um, so thank you all so much for joining us and uh look forward to the conversations through the day.
Event Announcer 2:18 ↗
Please welcome Andy Jasse of Amazon and Kate Rooney of CNBC.
Kate Rooney 2:29 ↗
All right. Uh, Andy, it's great to see you in person.
Andrew Jassy 2:33 ↗
Great to be here with you.
Kate Rooney 2:34 ↗
Welcome to the Bay Area.
Andrew Jassy 2:34 ↗
Yeah, thank you.
Kate Rooney 2:35 ↗
Uh, what a great event. Thank you so much for for having us. Um, nothing going on. Gosh, I wish we had more to talk about. There's it's been a very busy news week. Um, but I think AI is just such an interesting part of obviously Amazon's business. I've heard you on so many earnings calls talk about how consistently bullish you are, how big the opportunity is, but it's been interesting. We were just talking backstage about kind of the difference in in Wall Street's reactions. It can fluctuate more than than your message on AI. Give us sort of the big picture on what you're seeing, what signals you're seeing that actually justify all of the spending, 220 billion in capex that you're about to spend.
Andrew Jassy 3:13 ↗
Well, you know, I I think this is the most transformative technology that we've experienced in our lifetime, which is saying a lot given what we've been through. And I I look back on the beginning of AWS and we felt like the business was growing really fast in the beginning. In the first three years, we had about a 57 million annual revenue run rate. And in the first three years of this generative AI wave, just in AI revenue, we it's over $2 billion. So about 440 times faster than the way we grew in AWS when we thought we were growing fast. So it is growing unbelievably quickly. Uh, you know, I look at um the demand that we have, you know, we always save a significant amount of demand for on-demand, you know, for for services like EC2 and Bedrock, our inference service, but the big chunks of capacity are largely reserved already in 27. Um, uh, much of it is reserved already in 28 and a good chunk in 29. So we we have very clear, very substantial demand signals that are unlike anything we've ever seen before.
Now the way that the um cash cycle works in a business like AWS is that you have to spend the cash um for data centers and for servers and networking gear well in advance of when you can monetize it. So when you're growing at this unprecedented rate like I mentioned, you end up having a lot of capbacks which, you know, creates some headwinds on your short-term free cash flow. But the economics of the business are such that once you get a few years out and you have all the data centers installed and you're monetizing with the servers and the networking gear that you have in place, you really like the free cash flow and the return of invested capital. And I I think we're going to be a much larger company. Um, uh, and AWS is going to be much larger business than we ever anticipated. We felt like when we were building AWS the last number of years, we we felt like AWS could be a few hundred billion dollar revenue run rate business. We now believe it can be over a trillion dollars in the next 5 to 10 years. Um, and so I think we're going to be much larger company with much more revenue and free cash flow with a very strong return on invested capital.
Kate Rooney 5:25 ↗
How broad is the demand? That's one question that we hear that a lot of the power users in AI, sort of the 1%, are making up most of the demand, that it's very top-heavy. A lot of these companies are also venture-backed which I think has caused some jitters. What do you see in terms of how broad it is and what the potential is for this to expand and more people to actually start using it versus sort of this this bubble that we're in in Silicon Valley?
Andrew Jassy 5:48 ↗
Yeah. Well, I, you know, I don't really think of it as a bubble. I mean, it's it's pretty uh the demand is pretty diverse. You know, I I first I'd just say that um today um uh we don't have, if you look at our backlog, which is about $580 billion, there's no one company that represents even a quarter of it. So it's it's pretty diverse. Um, you know, 80 over 80% of the Fortune 100 are using Bedrock or inference service. So it's grown really quickly. I still think if you were going to zoom out and look at what the uh what the demand has looked like early on, it's it's been barbelled in that you have um the AI labs who are consuming a lot of um compute and some very significant applications like ChatGPT or Claude Code and then you have a lot of enterprises who are getting value out of AI today in uh cost avoidance and productivity. So automating customer service or business process automation or fraud and then in the middle of that barbell are all the current enterprise production applications. Um, uh, increasingly are are using inference but still the vast majority of them are yet to use inference as pervasively as they will and that is, you know, that that is going to happen. To me it's not a question of if, it's just when and a lot of what we're building right now is focused on trying to help enterprises get there.
Kate Rooney 7:11 ↗
Do you get the bubble question a lot? And I wonder if there are pockets of AI that you worry about that are maybe overhyped. I mean, we've got this rate environment, interest rates are going up. Are there areas that you do worry about where you're building any sort of contingency plan within Amazon?
Andrew Jassy 7:28 ↗
Well, I would, you know, there are different questions about whether we worry about it for Amazon versus, you know, are there areas that have more risk? I I would say from an Amazon perspective uh we feel very convicted about what we're building and that it's going to return very well for the company and for our shareholders. And it's because in part because of all the demand signals we see, but also if you think about when you spend the capital, the the biggest amount of capital that you spend is for the servers and networking gear. And so whereas with the data centers you have to lay out the capital about a couple years ahead ahead of when you can monetize it, for the servers and the networking gear it's just a few months. So we have real visibility before we pull the trigger on that. If we don't see the demand we won't we won't do that. But so we feel very good about it from an Amazon perspective. You know, there are um I think it'll be interesting over time whether all the different players who are building data centers whether they have a broad enough business with enough services um that, you know, they'll build long-term sustainable businesses and we'll have to see. I mean, there is it is growing so fast I can understand why a lot of people are interested in it.
Kate Rooney 8:44 ↗
When we talk about data centers, midterms were a big moment for the data center backlash. What do you think the AI industry so far is getting wrong and why was there this public rebuke of data centers? I wonder just your thoughts about how this could be done better and what if anything a slowdown, potential moratoriums in certain states will mean for your business.
Andrew Jassy 9:06 ↗
Well, you know, look, everything that we do today is driven by data centers. You know, banking, trading stocks, shopping, entertainment, ride-hailing, ordering food. I mean, everything we do in our lives um runs through data centers and that's before AI which we just talked about as the most transformative technology of our time. And so uh, you know, I think that collectively in the country and really around the world we had not invested um in upgrading the grid in a meaningful way in the way we were gonna we were going to need to. So at some point you were going to have to make a bunch of these investments. It just turned out AI happened happened quickly that pushed a lot of these um to happen sooner than than it otherwise would. I I think that um we're in a moment of time right now where um it's happening really quickly and and sometimes when things are happening quickly, you worry about it.
And but I I if you think about the type of buildout that's required for AI and and the pace it's growing that we're talking about, we haven't experienced a buildout like this of infrastructure in the United States since probably the 1950s when we were building out the highways and the bridges and we have to do it. I I I cannot fathom uh any scenario where the United States stops building data centers and allows another country like China to establish all the norms. Like I I think that would be a historic mistake by the United States that we would we would all regret for a very long time. I do think at the same time uh there are ways that we can all do it better with, you know, uh and and partner more closely with our communities. And I I think if you look at the data centers being built and they're being built quickly, there are differences between some of the companies that are doing it more responsibly and partnering with communities and those that aren't. I just think collectively as a group, we got to remember that we we are building it in neighborhoods. And, you know, there are some there's a lot of good things um in those neighborhoods, too. If you look at, for instance, in St. Joseph County in Indiana where we have a big presence. You know, if you look at the next few years, we'll contribute over $3 billion of taxes, whereas the previous land owners contributed about a million dollars of taxes. All those taxes go towards the community. They go towards schools and utilities and things that matter in those neighborhoods. And so, you know, uh um partnering with those communities, building programs like we've built with Built Together where we've contributed another billion dollars to things that matter in those neighborhoods and in those communities are important. And I think we can all partner better than we have in the past. But I I also think it's I just cannot imagine a scenario where we allow ourselves to um to to chase rather than to lead here.
Kate Rooney 11:48 ↗
Are you talking to other CEOs and are there sort of bad actors that have done this the wrong way and given the industry a bad name here?
Andrew Jassy 11:57 ↗
There's just a lot of there is a lot of building and some of it are um cloud providers and some of it are um uh, you know, real estate folks and some of it are just developers and, you know, we're not privy to all the folks that are doing it. We're just trying. We've set a set of standards that we think are really important um to both partner well with communities and to have the infrastructure that people need and we're going to work hard to abide by.
Kate Rooney 12:20 ↗
I also wonder about the position that hyperscalers are in given that you have so many existing data centers. I've talked to other CEOs about this who actually say if there is a slowdown or some sort of moratorium, you guys might actually be advantaged here because you've already got these existing buildings.
Andrew Jassy 12:38 ↗
I've heard that theory too and uh it's I'm not saying it's wrong. We we do have um we have a lot of infrastructure. We put a lot of capacity in in 2026. We're putting, you know, a substantially larger amount of that capacity in in 2027 and 2028, much of which is in progress, but it's not what we're rooting for. It it's incredibly important for the US to build out this uh this data center infrastructure so that we can allow the type of innovation and growth that that wants to happen here. So we we are rooting for the ability to build out that infrastructure and we're going to do everything we can to help.
Kate Rooney 13:12 ↗
We'll talk about geopolitics in a second. I want to get your take on Nvidia and some of the fallout with Jensen Wong this week. I know you don't know the details exactly but he was is such a central figure in the AI space. Nvidia is so central as a financier to a lot of the startups. You use Nvidia chips. What is your take on what the Nvidia saga has meant for the industry and will mean for for the tech industry in the near term?
Andrew Jassy 13:37 ↗
You know, I I uh, you know, I I have a lot of respect for Jensen. I have a lot of respect for Nvidia. They're a really important partner of ours and have been for a long time and will be for as long as I can foresee in the future. I I think the AI um the growth in AI as we were just talking about is staggering and it's, you know, it's more than one company. It's it's a number of companies that are having unprecedented success and growth right now and I I, you know, I they'll have to work out how uh how they move forward as a company. But we'll be partners with them either way and I, you know, we're excited about what we're building.
Kate Rooney 14:15 ↗
Is there any impact on Amazon?
Andrew Jassy 14:18 ↗
I I don't think it has an impact on Amazon.
Kate Rooney 14:21 ↗
When you think about your own semiconductor business, you have this relationship with Nvidia. I sort of think of of it as a a frenemy relationship. You might push back on that, but you are very reliant on their chips, but you're also building this competitive custom silicon business. Export controls are now back in the spotlight. Does it change the way that you think about where you can sell your chips and and the the potential for this business overseas, especially in China?
Andrew Jassy 14:48 ↗
Well, in the in the technology space, you you have to be used to coopetition. That's just it's how it works, especially if you're if you're working across uh lots of different uh um products like we are. And you're right, we we have we've had a very long and um fruitful partnership with Nvidia and uh and I expect that to continue for as long as I can foresee and at the same time we're also building our own custom silicon in Trainium which is, you know, doing incredibly well. We, you know, the last time we talked about it which was in Q2, Trainium had, you know, I'll say our chips business which is inclusive of both Trainium which is our AI accelerator and Graviton which is our CPU chip um, you know, was over a $25 billion annual revenue run rate business. So we we have, you know, we have a very large offering on our own and the majority of what we actually uh use for Bedrock which is our inference service which continues to grow real quickly that, you know, the majority of that are Trainium chips and yet we also have Nvidia chips. We use some of them ourselves, we we have customers who use Nvidia and so I I expect that will continue to be the case for a long time and and uh, you know, I would say on the export controls, you know, we will abide by whatever the rules are. And and that's, you know, in any country in which we operate that's what we have to do and that's what you hear too.
Kate Rooney 16:12 ↗
It seems like there's been two different camps. There is sort of the side of China shouldn't have access to the best US chips. And then there's the other camp which is we want US technology to be everywhere. We want that sort of to be the currency of the world. Where do you stand on on all those arguments?
Andrew Jassy 16:29 ↗
It's a it's a hard one. Uh and I uh, you know, national security matters a lot. I understand both sides of that argument and uh, you know, and at different times we may choose different paths there but again we we have long abided by whatever the laws are of the land and and we'll do the same here.

28 more exchanges in this transcript

Sign in free to read the rest of this interview. No card required.

Sign in to read the full transcript

Cite this transcript

APA, MLA, BibTeX
APA

Jassy, A. (2026, November 17). ‘We Have To Do It’: Amazon CEO Andy Jassy On Racing China In The AI Buildout [Interview transcript]. Semafor. CEOInterviews.AI. https://ceointerviews.ai/interview/3323525/

MLA

Andrew Jassy. "‘We Have To Do It’: Amazon CEO Andy Jassy On Racing China In The AI Buildout." Semafor, 17 Nov. 2026. Transcript, CEOInterviews.AI, https://ceointerviews.ai/interview/3323525/.

BibTeX
@misc{jassy2026_3323525,
  author       = {Andrew Jassy},
  title        = {‘We Have To Do It’: Amazon CEO Andy Jassy On Racing China In The AI Buildout},
  howpublished = {Interview transcript, Semafor. CEOInterviews.AI},
  year         = {2026},
  month        = {nov},
  url          = {https://ceointerviews.ai/interview/3323525/},
  note         = {Speaker-attributed transcript with timestamps}
}