Kevin Miller, Amazon

Clips
News
More from AWS Summit Washington, DC 2025

Kevin Miller

VP of Global Data Centers

AWS

Inside the AI shift in government: 11 perspectives on public sector ops transformation

Public sector transformation is entering a critical phase, driven by the rise of agentic AI and digital labor that doesn’t just automate, it adapts. But the biggest obstacle isn’t the technology. It’s the systems, silos and culture built for a slower, pre-AI world.TheCUBE’s John Furrier discusses public sector transformation.What’s really holding back progress in the public sector isn’t a lack of innovation — it’s the difficulty of adoption. From aging infrastructure to deeply rooted bureaucracies, systemic inertia remains the biggest obstacle. Meanwhile, AI is rapidly reshaping the role of government, raising the stakes for timely transformation, according to theCUBE’s John Furrier

AWS reengineers data center sustainability with AI and recycled water

Data center sustainability has become increasingly vital as data centers are highly energy-intensive, significantly contributing to greenhouse gas emissions and climate change .To address this concern, Amazon Web Services Inc. is pioneering innovative infrastructure design that dramatically enhances energy efficiency. The company is not only investing in renewable energy sources, but also implementing water conservation technologies, according to Kevin Miller (pictured), vice president of global data centers at AWS. Kevin Miller talks with theCUBE about how AWS is enhancing data center sustainability.“We’ve started to really use recycled water in the U.S., especially,” Miller said

play_circle_outline $30 billion spending on AWS infrastructure

play_circle_outline Water recycling and renewable energy initiatives

play_circle_outline Use of AI for operational efficiency and decision-making

Info
Transcript

Kevin Miller, Amazon

Kevin Miller

VP of Global Data Centers AWS

Kevin Miller, vice president of global data centers at Amazon Web Services Inc., joins theCUBE’s John Furrier and Scott Hebner during the AWS Summit Washington, DC event to discuss the company’s data center strategy. The segment explores AWS’ investments in AI infrastructure, sustainability and global expansion, with a focus on public sector partnerships.

Miller shares updates on the deployment of AI capacity through P6 instances and the use of advanced technologies such as liquid cooling and water recycling. The conversation highlights AWS’ efforts ... Read more

explore Keep Exploring

What was the total amount of money spent by Kevin Miller during the Amazon Web Services Summit? add

What are some examples of using recycled water in the US and other areas, such as Singapore, for industrial operations? add

Question: What are some of the impacts of data centers on rural communities and why are they beneficial for those communities? add

bolt Powered by CUBE AI

Kevin Miller, Amazon

>> Welcome back, everyone, to theCUBE, here in Washington DC for AWS Public Sector Summit, DC Summit, all one thing. They call it the Amazon Web Services Summit. Kevin Miller, vice president of Global Data Centers here. He's the star of the show, spending money like it's nobody's business. $30 billion in the past week. Got a lot of smiling. Looking good, good to see you.

Kevin Miller

>> It's great to see you again.

>> Welcome back.

Kevin Miller

>> Yeah, thanks for having me back.

>> So we're at the sixth month, almost halftime report for re:Invent. Last six months ago, we talked the game. Now you deployed the big numbers, and there's more. That's just US.

Kevin Miller

>> Yes. $30 million in the last week essentially of announcements.

>> What's the big thing? And you've got the international. What else is happening?

Kevin Miller

>> Well, I think when you look very broadly, we're deploying a lot of AI capacity. Of course, we've announced the coming availability of our P6 instances and continuing to scale up customers on Tranium, and so there's a lot. Even though there's only been a few recent announcements, there's been a tremendous amount of work all year to really continue scaling up that AI infrastructure, frankly at an unprecedented pace and that's been pretty exciting, keeping the team very busy.

>> I spoke with Prasad last week. He laid out the systems designs. Some air-cooled data centers coming. A lot of design-

Kevin Miller

>> And liquid-cooled coming very soon. Yes, yes.

>> So a lot of action happening. What's the focus on selection, the demand side? There's a lot of demand. We hear, "Come to my state." Is there a luring going on? Are you guys actively looking? Can you give us a taste of what's happening?

Kevin Miller

>> Well, we're actually always looking, always developing options, because in the data center area, the timelines are actually, you have to think long term because even if you have a parcel that's ready to go and has energy, it takes a while to construct the data center. And when you're talking about building at the scale we are, which is hundreds of megawatts or gigawatts of capacity, you can't build that overnight. So you have to really be thinking long-term. So we're always looking at what options exist, where we could go to add additional capacity, and we look for really a combination of where's energy available, where's there land available? And then also very importantly, where is the right kind of economic environment within a state or within a country to really support the growth? Because we want to come in and really be partners in delivering obviously what we need to meet our customer demand, but really also helping communities meet their interests, and so we're looking for that kind of partnership.

Scott Hebner

>> In terms of the curve of demand for AI compute and where you guys are heading, let's say over the next five years, do you feel you're getting ahead of that curve or is it always going to be tracking or is that curve still to be defined? Or is it just like AI and all this new stuff coming on board?

Kevin Miller

>> There's many curves. It's a question of which curve do you-

Scott Hebner

>> It's counting clouds on a cloudy day?

Kevin Miller

>> Yeah, that's right, counting clouds on the cloudy day. So look, we're adding capacity really at an unprecedented rate right now, and we feel good about the capacity we have and that's coming online in the next year, year and a half, but it could be off. We could need more, and so we're always keeping a close eye on demand and looking for opportunities where if we need to, we can expand.

>> When you guys look at the technology side, like power and cooling. Oh, we think we talked about that at re:Invent. Are there new innovations? You mentioned liquid cooling. What are some of the advancements you're seeing and how does that affect the state or the region that you guys align with? Take me through that long-term process, because once you make the selection, you're married basically.

Kevin Miller

>> Yeah. Once we've chosen that we're going to build in a particular campus or a particular area or state, liquid cooling or the specific equipment that we're deploying in there doesn't really change the need to build a data center to do it. Now, certainly with AI computing, one of the things we're seeing is that the density of the racks is going way up, because the whole point of AI is you want to get as much GPU compute or accelerator compute in as small a physical space as you can, and so that really drives the densities up, which drives the need for liquid cooling. And so ultimately, that may allow us to simplify the data centers in some way, but really right now, our focus is on having flexibility to make sure that we can deploy whatever comes, because we have a rough idea of the designs six or 12 months from now, but we know that that's going to change.

>> It's a lot of work. I'm not going to say gamble, because there is probability theory involved here. Which processes will be available? Supply chain, architectures.

Kevin Miller

>> That's right.

>> How do you guys handle that?

Kevin Miller

>> Well, I think really the key is to maintain that flexibility and to know that when we think about the timeline to deploy something, we're always looking to reduce that timeline so that we can really make decisions as late as possible. So we know we're going to build a data center and we can start doing that, but what percent of liquid or what percent of air is it going to require? Our goal is to really push that decision making as late as possible, because the later we go, the more certainty we have about exactly-

>> Risk management is better.

Kevin Miller

>> That's right. So we're always just evaluating that and then just understanding as things change, because they're changing every week.

Scott Hebner

>> What are some of the new technologies you guys are looking at in terms of energy?

Kevin Miller

>> Well, of course, last year, we were investing in X-Energy as a generation four small modular nuclear reactor company, and I think that's not going to happen overnight, but that is something that we think has a lot of promise to provide 24 by seven base load support on the grid long-term in a very safe environment. But of course, we are the world's largest corporate purchaser of renewable energy and continuing to buy solar and wind. We're also doing more with battery as well, and we think that battery in the right places in the grid can really provide some really meaningful benefit, particularly at those peak times of the day.

Scott Hebner

>> More than backup?

Kevin Miller

>> Yeah, provides support for the grid when it needs it the most, right?

Scott Hebner

>> Yeah, booster.

>> The peaks and valleys.

Kevin Miller

>> Right, mostly the peaks, is to be able to take care of when all the air conditioners are running on a hot summer day. Usually, that peaks in the late afternoon, and if you provide a few hours of support, that can really change the dynamics quite a lot on the grid. So we're doing more with battery and really just keeping our ear to the ground on all the technologies that are emerging.

>> So you guys announced you're expanding your water recycling piece. Can you explain that a little further, what the stats are on that?

Kevin Miller

>> Yeah. We've started to really use recycled water in the US especially. We have some pilot projects that we're starting to expand, particularly coming out of Virginia. And it's actually a technology, it's an approach that we've used in other areas for a while, including in Singapore, where we've actually made use of what they call industrial water, which is really harvested rainwater, and also the outflow from sewage treatment plants where it may not be potable water. It's not potable water, but it is perfectly acceptable for use in industrial operations. And so of course, we take that water, we make sure it's clean and safe, and then we use it in the data center. And so it just avoids consuming potable water when you don't need to.

>> Anyone who's played golf knows, we've seen when you go to the golf course, recycled water to get the grass to be green. You don't drink it.

Kevin Miller

>> Yeah, no, you don't drink the water, but it's perfectly safe for use in the industrial context.

Scott Hebner

>> And you're doing this across dozens and dozens of sites, right?

Kevin Miller

>> Yeah, we will over time. We've got a number of pilots and we're expanding it now, and over time, it'll be dozens of our sites.

Scott Hebner

>> Do you try different things at different sites in terms of pilots and try to get best practices?

Kevin Miller

>> We do, and one of the things to understand about our... A data center is not just a data center. In some parts of the world, we build entire water treatment facilities to really clean the water, and in others, we build sewage treatment facilities to make sure that the water is safe to discharge, and so we have a pretty deep understanding of how to be responsible with water. Of course, we have our goal of being water positive by 2030, and so it's not one size fits all. In some areas, we're going to be able to use recycled water. In others, we do different things with water, but all towards that 2030 goal.

Scott Hebner

>> Define water positive for the audience.

Kevin Miller

>> Essentially to return more water to communities-

Scott Hebner

>> Back to?

Kevin Miller

>> Yeah, back than we consume in our operations. Our data centers are actually very efficient at their use of water. It's something like 0.18 liters per kilowatt, which is really an industry leading in terms of the low amount of water use that we have. And in a lot of data centers, they use no water for the majority of the year. It's really the hottest days in the summer when they need water to reduce the incoming temperatures a bit, but actually most of our buildings run most of the year where it's just using free air. There's no water involved, aside from drinking fountains.

>> Yeah, that's what I was getting to. All right, so your goals now going forward, you mentioned some of the ESG stats 2030. Talk about the sustainability side of it and also the build out side. What are some of your goals?

Kevin Miller

>> Well, from sustainability, we remain very focused on our 2040 goal to be net-zero carbon, and so that underpins a lot of our purchase of renewable energy and more than 28 gigawatts of renewable energy in wind and solar, as well as our investments in long-term carbon free energy through things like our X Energy deal and our relationship with Energy Northwest to actually deploy a small modular reactor in the state of Washington. And our sustainability goals, it's really about continuing to focus on that. We've also done a number of work with reverse logistics and really doing as much as we can to harvest... At this point, actually 99% of the parts that come out of our data center, it's not waste. We actually recycle it and it gets either recycled directly into our servers as spare parts or it gets handled responsibly in another way, recycled or reused. And so we're continuing to focus on that kind of circular economy type principle. And then outside of sustainability, we're just continuing to deploy to keep pace with demand, make sure our customers can continue to scale and be successful.

>> What do you do every day when you wake up? You come up to the office, you sit around and say, "This place, how are we doing? Hey, we're doing good on sustainability." It's a huge task. You've got a lot of people working on projects, economic models. You've got a huge staff and I'm kind of just speaking a little over the top to make a point, but it's complicated.

Kevin Miller

>> Yeah, that's right.

>> Can you frame the scope or the magnitude?

Kevin Miller

>> Yeah. No two days are ever different, but when you're talking about deploying gigawatts of capacity around the world in a given year as well as making progress on all of these goals, it does require to stay really attuned to the data. We have data from all various data sources, and it's actually one of the interesting things we're doing within my organization, data center organization, is we talk in a lot of AWS services about building data lakes and really tapping into that data for AI. We're doing that. This year, I asked my teams recently to come up with all of the areas that we're using AI to improve what we do day-to-day, and last year, the team had maybe two or three items. This year, it was two pages, and I had to actually prune some of the items off when I'm summarizing it. And so it's across the board, from working with our general contractors and giving them RFI responses faster, because we have GenAI learning from what we've said previously. It's that to tapping into our data lake that we've assembled across my entire org to make better decisions and better insights about where do we need to optimize a particular data center? When you have a fleet of hundreds of data centers, at any given time, you might have one that is doing really good and others that may not be doing so good, and what lessons can we learn from the way these are operating well to make sure that the whole fleet can operate effectively?

Scott Hebner

>> Do you envision agents, AI agents, agentic systems beginning to do that?

Kevin Miller

>> 100%. Yeah, absolutely.

Scott Hebner

>> Down the road?

Kevin Miller

>> This year, we'll have agents that are doing that. Whether it's optimizing based on metrics from the data centers or just helping the decision-making around renewable energy projects. There's just so many facets where at the scale we operate, we can absolutely have agents have a huge benefit to the day-to-day.

Scott Hebner

>> Do you see it becoming more and more autonomous as years roll by or human in the loop?

Kevin Miller

>> Yeah. No, I think for some things, it will absolutely become more autonomous, and our data centers actually already have control systems that of course we've specially designed, but they are autonomous. So for example, if we have a power outage in a data center, the control system is automatically reconfiguring the electrical distribution in a data center without any human being involved, and so there's already a degree of autonomy, but I think we can take it to the next level by then applying AI systems. And obviously we have to do it safe and carefully, but that's all given. We will do that.

>> You have workflows and mechanisms you know.

Kevin Miller

>> Absolutely.

>> So you can have agents do that.

Kevin Miller

>> Right. We can have agents take well-trodden paths where we already know the guard rails, and we can be very confident that they're going to be successful.

>> All right. What's an area that you're excited about this year?

Kevin Miller

>> Well, I think the rapid advancement of AI is hard to undersell, and so everything I just talked about. It actually is really fun for me to be able to see how, yeah, we're deploying a lot of the capacity, but we're also able to really make use of it to be more efficient, to be able to deploy more predictably. And so really, it's a flywheel that's helping us and helping our customers.

>> And Roger was just talking about the economics. It's almost like a planned community. We're going to have a data center, all this impact.

Kevin Miller

>> Oh yeah.

>> But it's really first generation, large scale for a lot of these states to have a major technology enabler.

Kevin Miller

>> That's right.

>> Your regions were really well intentionally picked.

Kevin Miller

>> That's right, and increasingly, as we just announced in Pennsylvania, it is quite rural, and also when we announced in South Bend last year, South Bend isn't quite as rural but it's still largely in an agricultural community. And the University of Notre Dame's there of course too, but those communities, one of the things I think about is for these communities, data centers are really bringing in a source of dollars that are coming in outside, and it's not just then dollars circulating within a local economy. It's actually an incredible source of new dollars continuing to come into those communities, and so I think there's just a lot of development that can be done there.

Scott Hebner

>> How much of your time are you spending on regulatory issues? Is that more or less? Do you hope it goes down?

Kevin Miller

>> Well, I have an amazing team that actually, a public policy team that works night and day. Yeah, Shannon Kellogg and his team are amazing, but certainly things are changing rapidly, not just in the US, but as you think about what's happening in Europe and the reaction to both what's happening in the US as well as changes in Europe that are happening in terms of the governing and governments there.

>> But Kevin, you're open for business right now. You're in business development mode.

Kevin Miller

>> Absolutely.

>> You're in growth.

Kevin Miller

>> That's right, and we work with governments to be successful in doing that.

>> So take me through that targeting. Again, you're open for business. There's a little criteria involved. What's the selection criteria? As you do business, has that changed at all this past year? Has it opened up wider in terms of doing business with your data centers?

Kevin Miller

>> Yeah, so one of the things I say is actually, one of the best things that helps us in being successful is when we have a government that's really engaged and interested in having us grow and really puts people on point to help us drive all the various agencies. Because in every government, it doesn't matter, the names might be different, the exact functions might be owned in different places, but every government has multiple departments that all have to be involved from a construction, environmental review, economic development point of view. And so having people that are really on point to help coordinate across the whole of government to make these projects be successful and move at pace is incredibly valuable to us. So that's one-

>> Frustration.

Kevin Miller

>> Yeah.

>> It's an operating system.

Kevin Miller

>> That's right. And just knowing-

>> Agents will do well there.

Kevin Miller

>> Just knowing that it's a focal area can be incredibly helpful.

>> Create the bid.

Kevin Miller

>> I mean, someday it might be an agent to do that, but who knows? We'll see.

>> Kevin, great to have you on theCUBE, and again, congratulations on the 30 billion news that hit. It was super exciting. 20 in Pennsylvania and 10 in North Carolina.

Kevin Miller

>> That's right. It's been fun.

>> Onward and upward

Kevin Miller

>> And more to do, so thanks for having me.

>> Thanks for coming on.

Kevin Miller

>> Yep.

>> Bringing down the story, really an expansionary, modern era of infrastructures coming all over the United States and around the world. We didn't even get into sovereign cloud. We'll wait for in July to talk about that, but again, it's all changing here in DC. It's theCUBE bringing it to you. Thanks for watching.