From Wired, this is the big interview. I'm Katie Drummond. Amazon Web Services, better known as AWS, is back in the spotlight. I'm not talking about the most recent outage that nearly shut down 30% of the internet. Instead, it's because of the recent announcement from CEO Matt Garman that's focused on the very AI-centric future of one of the world's largest cloud platforms. Garman, a one-time AWS intern, is now guiding the company through perhaps its most transformative moment since its founding, introducing new AI systems, new partnerships, and new questions about who and what will dominate the next decade of cloud and artificial intelligence. In this conversation, we talk about the strategy, the pressure, and the big decisions Garman has to make, shaping the infrastructure that powers, well, almost everything.

All right, I'm ready.

Okay, let's do it.

Matt Garman, welcome to The Big Interview.

Thank you. Thanks for having me.

So, we always start these conversations with some very quick questions, like a warm-up. Are you ready?

Uh, sure. Go ahead.

Okay, he's he's ready. It's too late now. Okay.

If AWS had a mascot, what would it be?

We have a big S3 bucket sometimes that goes around, so we'll call it that.

Wait, sorry. What is an S3 bucket?

Well, an S3 bucket is like a thing that you store your S3 objects in, but we actually have a large foam uh big bucket that walks around and is actually looks like a a a paint bucket.

So, you do have a mascot.

Well, S3 has a bucket has has a mascot. It's probably the closest we have, and I like it, so.

Perfect. What's the most expensive mistake you've ever made?

Uh, personally or professionally?

Either.

Ooh. That's a good question. I think probably personally, the most expensive mistakes I ever made was uh playing basketball too long and I tore my Achilles. So, that cost me about 9 months of being able to walk. So, you know, that was I probably should have known that into my 30s, I was well past basketball playing age, but I lost a little bit of time there.

That sounds personally expensive. Psychologically, that sounds very expensive.

Exactly.

If if you could rename the cloud today, what would you call it?

Uh, what is it called today?

The cloud.

Oh. Okay. I actually think the cloud is a pretty good name, so I don't know if I would rename it. I would rename like uh we called Amazon AWS, Amazon Web Services, and now no one knows what web services are, so that I might rename a little bit, but um but I actually like the name of the cloud, so I'm not sure I would rename it.

Maybe Amazon Cloud Services?

Yeah, maybe.

Maybe. What's the part of your job you would love to outsource to an AI agent?

I try to outsource a lot of the parts to my jobs to AI agents that I uh if I can, but I haven't yet uh figured out how to outsource more of my kind of answering of day-to-day emails yet, and that's still I find takes a lot of my time that I haven't yet figured out how to do more efficiently, where I uh get the right information and get the right information out. Um but if I could figure that out, I think that would be great.

But have you, it sounds like you've tried. And I'm I'm curious about this, and I know we're supposed to be doing quick questions, but I was going to ask you a little bit about this later.

Yeah.

Tell me a bit about sort of how you've tried to incorporate artificial intelligence into your workflow, into sort of your personal professional life.

Yeah. I I think um uh there's a number of ways that I've done it. I think in particular for me, though, a lot of the benefits that I get uh in in my job in particular are taking a lot of information inputs and then kind of sharing those out to either the same or other people and kind of connecting a lot of those dots. And for my particular role, I haven't yet found a a huge uh shortcut into kind of being able to do that, particularly with regards to like the medium in which we communicate like email or other places like that, because I find that all of the shortcuts lose some of that nuance. Like there's some summary things that work. Um there's definitely some tools that allow me to summarize content more quickly or learn new content more quickly, which is I find that to be super useful. But um haven't found a huge time win for my role in particular where there's not as much kind of repetitive work or other things like that, and it largely is kind of knowledge um uh that I'm trying to get from a bunch of sources and then consolidate together to to send out to others.

That's actually very interesting to me and reassuring in a way because I sometimes feel like I should have found a bunch of shortcuts by now given how artificial intelligence is talked about, and I haven't either. So, maybe maybe both maybe both of us one day soon will. Um from one former AWS intern, that's you, to a future one, what is your best piece of advice?

I find that people always overestimate how much of technology has already been invented and kind of think that there's nothing left to do. And what I find is that we continue to be at the early stages of evolution, if as long as you're curious and looking and willing to try new technologies, new areas, that we're always at kind of that early stage of what can be invented. Uh and sometimes I run into interns who are like, "Yeah, you know, when you started AWS, it was small, but now it's a big company, and so there's not the same opportunity." Uh and I'd say that that's just that's just not true. I think there's just as much if not more opportunity than there ever has been.

Well, which leads me to my next question, which is, how do people know when you're unimpressed?

When I'm unimpressed?

Yeah.

Uh, largely would tell them.

Fair, so direct direct feedback.

Directly. Yeah. I I don't I'm not one to like that I mean, I think um, you know, I think uh largely though, it's not about impressing me. I'm more like uh, you know, I like when people are thoughtful, when they've come up with like the right sets of decisions. It's not that they're like uh like particularly, like every day, it's not like people are like coming up with something that's super um super novel. And it's and again, it's it's not about impressing me, but it is I like it when people have done the work so that we have all the information, whether it's customer input or data input or sales input or whatever, so that as a group, we can make thoughtful decisions. And so that I I like and impress when people have done that work ahead of time so that when we do get together, we we can make good decisions and and thoughtful decisions as opposed to like being um and and I'll often tell people it's not necessarily the the recommendation that I'm like going to be like impressed by, but I want us to have all the information so that as a team, uh we can move forward and make great decisions.

Last question, because you have a very big job. So, what is a hobby you wish you had more time for? I'm assuming it's not basketball based on what we just learned about you.

Uh, it's not. I've switched to golf. Uh, so I I I caught the bug a couple years ago and uh and quite love playing golf. I don't get to play as much.

Well, let me sort of set the stage a little bit. We're here to talk in particular about a bunch of announcements that you recently made around AWS and AI and a gentic AI that that Wired covered. Um, I'm biased, but I thought we did a great job with that coverage. Um, you know, just just very recently, but I also want to learn a little more about you in a professional context. So, tell me and tell all of us about your career journey thus far as now the CEO of of AWS, but you've had obviously a very long career at at Amazon before that and even sort of prior to that. How did you end up where you are now?

Yeah, after um, uh so I'll start at the beginning. I I had I worked for a couple of startups uh early on in my career. Um, none of them did particularly well, but I learned a ton from them, which was great. Um, after my second startup, uh my wife and I both quit our jobs and went to business school. Um, which was a fantastic opportunity. And uh as part of business school, kind of when I was there, um one of the things during my internship, um that I wanted to try to explore was what uh entrepreneurship looked like inside of a company, just because my my goal was always to go back and do a startup again. Um, and so uh as part of that, I looked at a bunch of different companies and uh I ran across Amazon and um and actually talked to Andy Jassy, and they were talking about building this technology services um capability inside of uh inside of Amazon. And I thought that's exactly what I wanted to see. I wanted to see what it would like for a a successful technology company to build, you know, to try to build something new inside of it, because I just want to see what that motion looked like and how you could learn from experienced um entrepreneurs and what was different than at a startup. Uh and so I did my internship for uh what turned into AWS uh in 2005 before we launched. Um, fast I was fascinated by it. I thought it was an awesome opportunity, and I said, "Great, I want to come back and work here for a couple of years, and then I would go back and do a startup." Uh and so then I started full-time in 2006, effectively as the product manager for all of AWS, like there was, you know, it was it was largely kind of um defining all of the services uh as we launched them. Um, and so I started a couple weeks after S3 launched, and uh before the rest of our services launched, and helped launch them and name them and price them and do a bunch of things. And uh and then I kind of kept getting more and more responsibility. I kind of focused on um EC2, which was our compute service. Um, started taking on engineering teams. I actually launched our block storage service, kind of wrote the PRFAQ for that, and hired the first engineering team and launched that. And then kind of grew to lead um most of our our our kind of core compute and networking and storage product areas. So, um all of the product and engineering teams for that, and um it was fun. I got to learn a lot along the way. Like, I I I'm not necessarily um kind of or I wasn't, uh originally kind of a deep technology person, but got to learn about hypervisors and kernel engineering and a bunch of like kind of these like really low-level kind of core technology pieces, which were cool. And it was super interesting for me to learn. Uh and it's just such a fascinating space. And as AWS grew really rapidly, um, you know, we we grew along with it, and we grew the team um pretty significantly. We were fortunate enough to work with some of the the best technology people in the world, um, as we built the service. And and a bunch of services and as the business grew, uh and then um, uh I can't remember the exact time. It was after about 12 or 13 years, um, so it would have been uh, like 2019, something like that. Um, uh Andy Jassy, asked me, uh, he called me in his office one day and asked if I would lead um sales and marketing, and I literally had nothing I didn't know anything about sales and marketing, and in fact, I was like kind of like looking around like if he was talking to someone else. Um, but uh, but it was a great opportunity to kind of learn that space. And it was it was a unique opportunity, right? I was basically handed one of the world's two or three biggest uh enterprise sales and marketing organizations, having never done any of those jobs before. Um, so it was you know, and I think Amazon's a bit unique in that we kind of trust people where, you know, you're smart, you know how to operate, you've you've you've you know the business, um, you don't necessarily have to know the exact thing that you're going into. Um, and the team was gracious and helped me learn that. And that was a great opportunity to get to learn how to how to run a field organization at scale and and really get to spend a ton more time with customers, which was awesome. Really understand the nuances of what a startup customer really want was looking for versus an enterprise versus the governments, um, uh and and kind of their their various different industries that work for. Um, and then took over a CEO, um, kind of a what was been two years ago, a year and a half ago. Um, and uh and but so I've spent almost 20 years here at Amazon all all in AWS.

Yeah.

And how big is the organization by employee size that you now run?

Uh, I don't know if I don't know the exact numbers, but it's, you know, it's in the hundreds of thousands.

Hundreds of hundreds of thousands. I mean,

Yeah. And a lot of that is, you know, we have large data centers where we, you know, we we we we run kind of very large operational organization. You know, we have data centers all around the world and, and pieces like that.

And I mean, I I run a team at at Wired. It's sort of in the in the low 100s, and I love management, and that's sort of for me was always it was clear fairly early in my career that that is sort of where I wanted to to go. Um, was that always clear for you, sort of the idea that, yes, taking on sort of larger and larger pieces of this enterprise is what I feel like I am sort of meant to do, what I want to be doing?

Yes, I mean, I like it, and I think that I'm, uh, reasonably good at it, I guess. So, I guess I took on

Well, I'll ask some of your employees. Some of your hundreds of thousands.

I guess that's not for me to say. But um, I liked it, and and the more I did it, the more I, you know, it was a ton of learning. I what I really love is when I get to take on roles where I get to learn more um and get to stretch myself, and uh and frankly, I love building and love having an impact on what the business is doing and what our customers are doing. And um, you know, scaling beyond yourself, it's hard because you can't do all of the, you know, you don't get some of the joy of like actually physically kind of getting to build the thing or deliver the thing yourself.

Yeah, or like write the you didn't write the story in my case. Yeah. Yeah.

That's right. But um, but you can write a lot more stories, right? in in that case, and um, and you kind of uh learn to do that through others, which is also a an interesting and useful skill. Um, and then you learn how to communicate to teams, you know, first through direct management, then through layers of management, then through, you know, through uh, you know, mediums like talking and and things like um, you know, all company meetings or or other kind of mechanisms like that where you have tens of thousands of customers or or employees that you might be talking to. Um, and so you got to think about how do you build mechanisms, and this is one of the things that I enjoyed learning, which is, how do you think about building mechanisms that allow you to help those individual contributors make some of the the right kinds of decisions in the strategy that that you're trying to drive for the team or the business or the company. Um, and you know, I've quite enjoyed kind of learning how to how to leverage some of those mechanisms at different scale. Um, and that's been fun to fun to do too.

Because you just talked about sort of going from managing people to managing managers, I swear I'm asking for a friend, but do you have any sort of particular mechanisms that that you have picked up in those 20 years that stand out to you as particularly effective strategies because I will say there is something very specifically different about managing someone who then manages people who then manages teams of people. Um, it's like it it has the potential to be a very unproductive game of telephone, and I'm curious about the mechanisms that you employ to make it much more effective than that.

Yeah. I I mean, look, everybody has their own way of of doing this. I think for um, so you a little bit have to find what works, and I do think that as your team and your organization gets larger, you have to change some of those things. And I think that's one of the common pitfalls that I see people fall into is that they will assume that the thing that worked great when they were a line manager, managing a team of six to 10 people, will work the same as when they're managing a team of 100 people. Um, and and those things same things won't work, and then, you know, I think there's another shift that's like when you don't know the name of everyone in your team or your organization, which I can't unfortunately know today, um, and usually that breaks somewhere around, you know, 100 to 200 or some somewhere in there, you just you'll run into people who are in your team that you don't know or or don't know their names. And um, and you just have to think about all of those things differently. Um, and where they are. It's really hard, I think, to have you want to give the broadest set of people in your team as possible mental models on how you would make decisions in their place as opposed to what the decision actually is, because if you have to like send down edicts all day, one, it's not as empowering to your team, and two, like there's chances that whatever you say gets miscommunicated. But if you can have, and this is actually a lot of the power of culture uh in a company, but also what we think like these mental models, which is like if I was going to approach this situation, this is how I would think about it, um, or these are the the ways in which I would make trade-offs or think about kind of decisions, then you can empower tens of thousands of people to go make decisions. And then you can focus on how do we make sure that we hire smart people and um don't punish them when they make bad decisions, but course correct. And that's how I've kind of learned at scale where your real leverage points are ensuring you have those right mechanisms at place, having a mechanism that does allow you to kind of find where things are going well or where they're not, and when they're not, where you can dive deep into that and really get into the details to understand really at the very core level, like then you kind of switch modes, you're like, "Now I'm line manager again, I'm understand the very details of exactly what you're doing," get things kind of sorted and then kind of jump back out again and and and stay at a high level. Um, and and look at mechanisms to how you can do that, and um, and that's the best way I found to do it.

Well, I appreciate the free management advice. I'm sure many of our listeners do, too, so thank you. But I want to now ask you about about AWS in a broad context. I think a lot of people listening I would describe Wired's audience as sort of curious generalists. Obviously, they're listening to this because they're interested in sort of technology and where it's taking the world. I'm sure all of you listening have some sense of what AWS is. You have some sense of how important it is to what you do every single day, but they probably don't know sort of in brass tacks just how big AWS is and sort of how vital it is in terms of just infrastructure. So, I'm hoping can you explain it to us, maybe not like we're five, but like we're we're 21, we just graduated with a humanities degree, and we're really trying to understand this thing that you that you are in charge of.

At a high level, our idea behind AWS hasn't changed in the last 20 years, which is there was a bunch of pieces of technology that was hard and non-differentiating that companies had to do for a long period of time. And that was used to be they had to build a data center, they had to go find servers, they had to take care of the servers, when a disk drive broke, they had to go fix it, they had to set up their networks, et cetera. There was a whole bunch of work that they had to do before they could ever, you know, write um Netflix, that's like a, you know, that could actually like write a cool application that would stream a movie to your end customer, or, you know, write Airbnb that would think about logic of how you could connect individuals with people who had rooms that they wanted them to to to stay in. Uh whatever your application was, right? And so, our goal was, what if we could do that work for companies so that they didn't have to do that? Um, and so that that was that was the thesis when we started, which is, what if it was as simple as somebody could come and make a an API call and and and simply say, "Great, give me servers, give me storage, give me databases, give me whatever," and we'd provision them for them, and then, you know, through an internet connection, they can have access to that. Um, uh and so it turns out that was a very powerful idea. Uh and I think we did a good job executing on some of the abstractions that made it really powerful where a lot of companies um went and built their applications on there. And so, the companies I mentioned, Netflix and Airbnbs and Pinterests, are all some of these early customers that we had that built their business from the beginning kind of on AWS and and the cloud and don't own data centers. You know, they they largely just run inside of AWS. Initially, when we first launched the business, we thought this was going to be incredibly compelling for startups and some technology companies, which which it was, but as we grew, we found out that enterprises and really large organizations were equally compelled by this value proposition, eventually. And we had to build a lot more capabilities for them, whether it was like encryption capabilities or audit logs or abilities to hit particular compliance things or whatever it is, but now we have customers like Pfizer and and JP Morgan and um the United States government and the intelligence agencies. We that was a a big a big win for us when we we kind of convinced the US government that we could build a top-secret region and that they could run intelligence workloads inside of AWS.

And and were you in those meetings? I would have loved to have been a fly on the wall in those meetings.

Yeah. Yeah. And and we uh we walked through some architectural questions that they had and and And what does it take to convince the United States government that they should they should run on the back of AWS? I mean,

Uh, you know, I mean, like there was a whole RFP process, there was a lot of work that we did, but it was also just some whiteboarding where we kind of walked through like how would it work and how would you, you know, and and it at the end of the day, you know, it's not it the cloud sounds like it's a magical technology, but it is, you know, it's it's data centers and it's networks and it's it's servers and and other things that we run at very high reliability, at very high security. And um, and we kind of and we found some forward leaning technology folks that wanted to figure out how they could get the benefits to the government because it turns out, if you find the right person in an organization, even in somewhere that's as as large and bureaucratic as the US government often is, um, you'll find people who want to lean forward, they want to go faster, they want to deliver value for citizens, and finding that right person and then being able to collaborate with them, you know, their eyes light up just like they do for a startup company, right? And so, so today now it's across almost every country and every industry that you that you think about. You know, Nasdaq trading markets run on AWS, it's financial services companies run on AWS, hospitals run on AWS, media and entertainment is, you know, whether it's live broadcasting or streaming broadcasting or or any of those things, we we power a lot of those that technology um across the board, really. And so, we're uh we're excited that we have millions of customers all around the world, and you know, we've grown the business now to be about a $132 billion run rate business.

Wow.

Um, and it's but it's still growing 20% year over year on on that large of a base.

Just when you thought you were done learning on the job, enter artificial intelligence, right? Which obviously has been around as a technology for a very long time, but we are in this sort of new era and this new sort of challenge for you and your organization, which brings me to this this recent keynote um at the Reinvent conference that you held recently. You also streamed it on Fortnite for the first time, I might add. But you announced some pretty significant changes to AWS, to your mission, to your priorities, and to what you you would be offering um to your consumers. And I'm hoping you can sort of talk us through that that transformation, if you would describe it as a transformation, maybe you wouldn't.

I think technology is always iterating and going through these kind of transformations, so I think for us, staying at the forefront of of every technology innovation is incredibly important. And I think there's been almost no technology leap since maybe the cloud and the internet before that. And so, um, we've been investing in AI and in AWS and Amazon for the, you know, the last decade plus, two decades, maybe. Um, but um, but definitely with the the leap forward from generative AI over the last three years, we've just seen a massive change in what's possible for customers. And so, we've had this vision that it's not just going to be AI is over on one side, and then the rest of your business is going to be over on the other side. But, basically, AI is going to be built into what everyone does. And in order for that to happen, number one is you have to have all of your data in the cloud world, and then we've built this whole platform of tools that then allow you once you have that data in the cloud to to deliver differentiated value to your customers. And so, some of the things um that we launched and and particularly I'm quite excited about at Reinvent are really around AI agents. And the difference between kind of the first generation of of AI tools that were really around summarization and content creation, right? And then we're all quite excited and got a lot of value out of those, but there's only so far that goes. I think the next stage is these agents. And agents, the the real value is they can take access to your data, they can still do some of those summarization and content creation, but they can go actually accomplish tasks, and they're able to reason. And when you have these agents that can go reason and accomplish tasks on your behalf, all of a sudden you can kind of force multiply what you're able to do. And we launched a couple of things. One was called Nova Forge, where we allow customers to actually take their data and integrate it in at the early stages of training one of these frontier models, which we call Nova. And so, that gives uh enterprises one of these AI models that deeply understands their data and their domain. And then we launched a number of these frontier agents that allow customers to really go and and deliver big bodies of work, whether it's in coding or operations or security. And we've spent the time to really build this platform where it could actually deliver value. And I think broadly people kind of understood now what that long-term strategy was. They really get it, and as they see projects really delivering into production where there's real value, they see that kind of AWS is that platform that they want to go do that. And that's that's what customers are telling us over and over again this last week.

You know, it's interesting, I feel like 2025, which we are, you know, thank God, almost done with, um, was a was like a confusing year in the narrative for AI. And I say that because I I feel like in January, you know, I went to some conferences, talking to some people, this is the year of the agent. A gentic AI is here. It's about to change everything. You know, that was in in January. That was almost a year ago.

Yep.

Am I using a gentic AI right now? Absolutely not. You know, has that sort of come to fruition in the way that I think people were talking about in January? No. And at the same time, you've you've seen sort of, you know, reports from, you know, MIT, for example, finding that 95% of gen AI pilots in companies are failing to yield the productivity that I think those those corporate leaders thought they would see. How do you make sense of all of those narratives coming together? Well, we just Were companies just moving too quickly? Was the promise just just too fast?

If you if you jump forward and don't kind of build that strong foundation of I have my data, I know the workflows, I really know how how some of these things are going to tie together, um, then you're not going to get any value out of it. Like you're going to have just a chatbot, which is kind of cool and looks neat, and then everybody has a chatbot, and then what? And so, it it is those differentiated workflows. You know, and they're kind of less sexy and interesting for people to look at, but a workflow that can help you automate insurance claim processing is super valuable, right? And you can actually like get people their claims processing faster, make sure that you cut your costs, have a better customer experience, make sure you have better accuracy. Like, you can deliver all these things. And, you know, some of these technologies weren't available in January. Like, this technology is moving so fast that the capabilities are are much better today. And so, I will tell you, you know, at Reinvent, I sat in a room of executives for a broad set of companies. I asked a show of hands of who is either now now starting to see positive ROI to their AI investments or see a clear path to meaningful positive ROI in the next six months, and I think it was 90% of hands went up um to people in the room. So, so it is like people are starting, and I think that's not the answer that I would have gotten a year ago. But it's because we've done a bunch of this work, and it's because we've done a bunch of this like work together with customers to understand exactly what they want, how do we solve their problems, and how do we deliver solutions that do deliver them real value and not just, you know, clickbait headlines that that sound good.

And on that note, you know, I will say candidly, Amazon has not been a key part of a lot of these AI narratives, right? There have been other companies that have been out there making announcements. It feels like once a week. I mean, I can tell you leading Wired, it's sort of this constant stream of of news, this model, that model, we're doing this, we're doing that. Does that worry you? Do you worry about Amazon not being in that narrative, or do you again just feel like you took your time for a reason?

Yeah. I think both of those things are true. Uh I do worry about it because I don't want customers to kind of think that we're not innovating or or driving the latest technologies that need, and I want to make sure that it's not just kind of headline grabbing stuff, and it's actually great value that we're delivering for companies and and value that we're delivering to the business. Jeff Bezos used to have a saying that you have to be willing to be misunderstood for long periods of time. And and for us, like I I think that's what maybe some of the last two years was. And so, you know, I think a lot of that narrative has changed now. And if you talk to lots of analysts, if you talk to folks in the press, talk to customers, they no longer kind of think that. Now they're saying, "Look, actually, AWS has by far the strongest agentic platform to go build on. They have the broadest set of models that I can build on. They have the broadest set of security controls and compliance controls that actually if I go put these agents in production, I can actually audit, know what they're doing, control what they're doing." Um, and that is what we're seeing. And and it's not to take anything away from, by the way, like chatGPT is incredible consumer application, but it's just a different thing. That's not our business. Our business is to make sure that banks and healthcare companies and media and entertainment companies and and energy companies can drive their businesses and deliver more outcomes for their customers or cut costs or whatever they want to do. And so, I'm quite pleased with where we are now, and um, and I think we've already seen that narrative largely shift.

I wanted to ask you a little bit more about Nova Forge, which was particularly interesting to me. And what was particularly interesting here is this idea of custom pre-training as opposed to the idea of fine tuning as sort of a way that a company can take a model and and really make it their own. Can you explain that distinction to everybody before I sort of dig in a little more?

Sure. It's not new that people have thought, "Okay, there's these out-of-the-box models that I want to customize." And the best mechanism that we've had to date to customize these are these open weights models that um that Meta was great in first kind of release with their first Llama model, but now we've seen um a variety of these, whether it's Mistral or DeepSeek or Quen or whatever. There's a number of these. But what they are, they're still black boxes, right? You know, you basically get a a fully kind of pre-trained model, and then they open the weights, and you can do either tuning of the weights to kind of focus in particular areas that you wanted to focus on, or there's kind of new techniques like reinforcement learning where you can start to send more information to these models that train them after the fact. But we find is that if you put too much new data into these models in the later stages, they forget the early stuff. And so, they forget kind of what made them great at reasoning, or they they they actually forget some of that data because they get what's called like overtrained on some of the data you're giving it, and then they lose what was valuable in the first place. And so, there's only so far that that can go. We also find is that those techniques are very ineffective if the model wasn't already trained on your domain. And so, if you try to go teach one of these open weights models about protein folding and they know nothing about protein folding, it doesn't work, because it doesn't know how to inherently reason about that thing. And so, what we found is that if you can train them earlier, and I use this analogy in my reinvent a talk about, you know, the human brain, you're able to learn new languages early when you're younger, much easier. Like, now if I try to learn a new language, it's it's it's much harder to do. Um, and so, models are somewhat similar to that. And so, what we've done, which is a unique thing, we kind of refer to it as open training, but, but really the idea is that if you can take your data, you have this corpus of data that is from your domain and your particular company and the ways that you do things, and if you if you're able to insert it into the pre-training stages and then mix a bunch of the data that was used to originally train that model and then finish pre-training the model, the model is then when it's done, it actually now inherently knows all of your your stuff. It knows about your data, it knows about your company, it knows about your domain, and anything that you do about fine tuning or post training after that is actually much more effective because it already knows that. It's just never possible before. One, because open weights models never would expose their data, and and there was no kind of mechanism for them to be able to go do this. And so, we did this with with our Nova models and our Nova 2 models. And and we what we do is we open them up and we say, "If you're a, you know, taking a financial services company, you can take all your information, inject your data, mix it with an Amazon curated data set." So, we this is the data that we that we have that we have proprietary used to to use, mix, and we'll we'll give you tools to easily mix that together, and then you finish pre-training the model and you effectively have your own custom frontier model that understands your business that you were able to train for, you know, a couple hundred thousand dollars or or whatever the cost is to finish that training versus millions of dollars of doing all the research to actually go um and build a a frontier model.

And you know, with any sort of new opportunity, right, there is risk. And I'm curious about risk in the context of of Nova Forge in a few different ways. You have, one, obviously, you already deal with companies who have very sort of confidential proprietary information, right? You just mentioned financial services. You know, inputting all of that data into this this model and into this training. There's there's sort of that piece of it. You also offered a really compelling example off the back of of Reinvent from Reddit, right? Which is using Nova Forge to develop a model that can be used for content moderation, which essentially means training a model that breaks a lot of conventional rules around how these models are are typically designed, right? They would be designed to avoid offensive or violent content entirely. Reddit on the other hand needs a model that can like lean into that kind of content in order to do the moderation. So, those are two sort of I think very distinct examples that I just offered, but I'm curious what kind of risks exist when we start down the road of this kind of custom pre-training, and and are there distinct risks that that might be different from what we've seen historically?

I don't know that there's any particular different risks. I think from a from a data protection point of view, um, we have all sorts of data protections around this this happening kind of in this data still living kind of in customers' domains and in their VPCs and um, and them having exclusive control over that. And I think that's one of the things that we pride ourselves on and have over the last 20 years is really protecting customer data and making sure that it's isolated and protected. You know, I think from when people are building their own model, this is true no matter what. If they're fine tuning it, if they're doing any of that work, companies have to think about kind of what is the output, and they have to own the output of their own models, whether they when they even if they're using an off-the-shelf model, by the way, you have to own the outputs of that. And I think that's one of the key pieces is that you can't just give up responsibility for whatever your technology is doing, right? You can't give up responsibility because a database makes a a particular join and you're like, "I don't know, it's just how the database did it." Like, AI is no different than that. And um, and we deliver powerful tools to customers, but they have to to own those outputs and think about them. We still have safety classifiers, by the way, for things that are are really like, you know, you can't go pre-train something and then like create it to go build a bomb or things like that. You know, we still have a lot of those safety controls that are like real safety controls are are absolutely still in place, and there's no circumventing those, but with regards to things like like you mentioned, content moderation, you know, that that's a that's someone else's choice, and they can make a choice, and they own kind of what that looks like, and they have to be sure that they um, by the way, they that's why they're doing it is because they want to make sure that they make great choices for the content that's on their website that's appropriate for their site.

And so, we're talking about sort of this this this premise whereby AI agents will be much more integrated into enterprise settings, right? And I'm curious about how you think about that in the context of the workforce. Obviously, there has been, again, we're I'm looking back at 2025 and remembering sort of comments that, you know, other AI executives have made around, you know, job disruption, you know, um, cuts to to the workforce, et cetera. You actually made an interesting comment, um, where you said, you said that replacing junior employees with AI is, quote, "one of the dumbest ideas you've ever heard," which made me laugh. Um, I'm curious if you could talk a little bit more about that and more about how you see artificial intelligence and a gentic AI changing the workplace in the years to come, because I think you have a maybe a point of view on this that, um, some people might find reassuring, and that I think is different than what we hear from a lot of a lot of other leaders.

Yeah. And that point, in particular, by the way, was just it was specifically around software developers, but but I think it applies to lots, which is there there is a there there was this kind of thought that like you'll just replace all of your junior engineers and all of your junior employees, and you'll just have the most senior, most experienced employees and then agents. Uh and number one, my experience is that many of the most junior folks are actually the most experienced with the AI tools. They're actually most able to get the most out of them. Number one. Number two, they're usually the least expensive because they're right out of college and they generally make less, so kind of like they're if you're thinking about cost optimization, like they're not the only people you would want to kind of optimize around. Um, and and really three is that at some point, that whole thing explodes on it on itself if you have no talent pipeline that you're building and no junior people that you're mentoring and bringing up through the company. We we often find that that's where we get some of the best ideas. We get the new new like fresh blood into the company from fresh hires out of college. There's a lot of excitement, there's a lot of new thoughts, there's a lot of new ideas. And so, my thinking was just like, you you've got to think longer term about how you think about the health of a company and and just saying, "Okay, great, we're never going to hire junior people anymore." That's just a a non-starter for for really anyone who's trying to build a long-term company.

What does this mean though, sort of for the workforce broadly? I mean, what does it mean for Amazon's workforce? Like, what does this look like as AI agents sort of infiltrate, which is not a generous word, but it's the word that comes to mind, sort of infiltrate the way we the way we work and live.

Yeah. I I think one of the things that I tell our own employees, your job is going to change. Like, there's like, there's no two ways about it. Like, I I I'll if there's the only thing I can promise you is that the way that you did your job four years ago is not how you're going to do your job next year. And you're going to be able to have a bigger impact, you're going to be able to do more things, you're going to be able to to to have a broader scope of responsibilities, and it's just not going to be the same things that even made you successful five years ago may not be those things. You're going to have to learn new skills, you're going to have new learn new ways of working. We may have to organize our teams differently, we may have to go after problems differently. So, so people are going to need to be flexible. There is for sure going to be disruption and how work is done because jobs are going to change and industries are going to change, and if they don't, you will most likely get left behind by people who move faster and um and do change. So, there is going to be some disruption in there for sure. Like, there is no no no question in my mind.

And and when you say disruption, I mean, certainly the way we work, but but but disruption in the context of sort of of job loss, sort of in a big picture economic way.

You know, that I think is uncertain to me. I I like I am very confident in the the medium to longer term that AI will definitely create more jobs than it than it removes. Like, for sure, I think anytime you find opportunities to create new economic prosperity or build new experiences or things like that, there are there are more jobs that get get created. And so, you know, in the short term, but they will be different, and there are jobs that will be um eliminated as part of it or reduced almost for sure. And you think about jobs where in particular where there are some jobs that get automated away, just like all kind of waves of technology that has been true, right? There are some things where you just no longer need quite as many people to do a particular job. And so, our job is to also provide training and upskilling so that we can retrain so that there are other roles that that some of those folks can do. And and not all people will want to do that, and there may be some churn in the in the short term as people either are hesitant to to to learn new skills or don't want to learn new skills or other things like that. So, there there is going to be some, but but this is true about almost every single technology change. This is true about, you know, when personal computers came around, it was true about industrial automation in the 1930s. Like it was, you know, was true when the internet came and and it will be true for AI, too. And there will be some jobs that were, you know, that that that there won't be as many of. That is true, and I think there will be new jobs.

I'm I'm curious about how you're thinking about all of this from an environmental perspective. I think one of the sort of notable pieces of of a gentic AI is this idea of of having them run continuously for hours or days. I would imagine there is sort of a sustained energy demand in that context. We're already talking about a very energy intensive technology. You know, Amazon right now is the single biggest purchaser of new renewable energy contracts in the world, at least for the last five years, if I'm if my facts here are correct.

That's right.

How do you sort of put that commitment together with this this drastically increased need for for energy?

Well, look, I think we have to keep and and so part of what that is is us investing in, you know, that is not just going out and buying existing things. That is investing in new projects and bringing new renewable energy projects online. And and that is it's a huge commitment for us. It's something we do every single year and we will continue to do, um, because we think that's important that um, from an environmental impact point of view and and and frankly from an energy availability point of view is important. Um, I I do think we're going to have to keep looking at other energy sources. I think that nuclear is one of those that's going to be um, very important for us to look at in the in the medium to longer term so that we make sure that we have enough kind of carbon zero energy out there. But, we need to look at all of those sources of energy. And um, and it doesn't mean that like no one's going to use natural gas in the intermediate term. I'm sure some of that is true, too. Um, and for our goal is to how do we continuously period over period, year over year, reduce the carbon intensity of the the energy that we consume with a goal to get to zero. Uh and so, that is where we're working super hard. We're still committed to that as a company, and we spend an enormous amount of time on that.

And how do you how do you sort of respond to criticism of of this venture, which is, you know, an enormous undertaking internally? I'm sort of curious about the management piece there. I mean, you had a few weeks ago you had over 1,000 Amazon employees, granted, I mean, it's a company with seven figure employee base, if I'm correct, but they described the company's, quote, "all cost justified, warp speed approach to AI development." They say that it will cause, quote, "staggering damage to democracy, to our jobs, and to the earth." That's quite a statement from Amazon's own employees. So, I'm curious sort of when you hear that read back to you, how do you address it?

Look, I think when you have uh, number one, is we we we encourage our employees to to have their own thoughts and um, as to how they're thinking about things.

Well, that's I mean, that's that's good because a lot of tech companies these days are are not happy to see that happen, so

Yeah. But, you know, I I will say also say that, you know, when you have an organization of any size, you have viewpoints from a lot of different places. And I would say that that is not the majority opinion from our employees or even close. And I think most of our employees are excited about the technology we're building, excited about the um, value that we're giving customers, and um, are excited about the potential and and like the climate pledge that we have and and the the path that we're on there, too, which is also important and we're equally committed to. And so, you know, I think that's okay. I think as long as those concerns are respectful, we're willing to listen to them. But I I think they're far from the majority. In fact, they're the the very small minority view.

And now, last question, I mentioned sort of at the top of the of the conversation that in January 2025, everyone promised me this was the year of a gentic AI. And they were wrong. And so, as we look out to 2026, I'm curious for your prediction in the context of AI, what is next year all about for us? And then in 12 months, I'm going to call you and we'll see.

Just to be super clear, I'm awful at predicting the future. So, you can You can You can call me. I just probably won't be right. But, I will say that, uh look, I think one of the big things that we are going to be incredibly focused on for our customers is delivering real business returns for them. And so, whether that's through agents, whether that's through customized models, whether that's through like scalable infrastructure, whether that's eliminating tech debt through using Transform to get them off of mainframes or get them off of legacy databases, like, that I think is going to be a big focus. It's not going to be about going and experimenting. It's not going and trying new technology. It is really delivering value to the to the end customers and to the business.

So, it's brass tacks time for the P&L to look a little bit different in the next year.

Yep. And that's where look, and I think that's where that's where customers have relied on AWS to help them improve for the last 20 years, and it's what we're particularly great at.

Well, Matt, thank you so much for your time. I really appreciate it.

Yeah, absolutely. Thank you for having me.

This show is produced by Jessica Alpert with help from Audreyana Tapia and Sam Egan. Sound design, mix, and original music by Prân Bandi. Kate Osbourn is our executive producer. And I am, of course, your host, Katie Drummond, Wired's Global Editorial Director.