RAISE Summit 2025 | Mike Mattacola, Coreweave

Clips
News
More from RAISE Summit 2025

Mike Mattacola

CBO

CoreWeave

12 signs AI infrastructure is reshaping the IPO market: theCUBE’s RAISE Summit insights

Capital markets are being reshaped by artificial intelligence, with AI infrastructure emerging as a critical pillar of how companies scale, compete and go public. What was once considered a back-end concern is now front and center in IPO roadshows and investor decks.TheCUBE’s John Furrier talks about AI infrastructure at RAISE Summit.The shift is about more than hype. Investors are scrutinizing how companies embed AI into their business models — not just to drive innovation, but to ensure long-term operational resilience. It’s no longer enough to show top-line growth; firms need to demonstrate that their AI capabilities can sustain profitability and differentiate them in crowded markets

play_circle_outline Relationship with NVIDIA crucial for accelerating product launches and market presence.

play_circle_outline Shifting Sands: The Transition from Traditional Data Centers to AI-Centric Architectures and the Rise of GPU Clouds

play_circle_outline CoreWeave's Mission Control enhances observability and performance for AI workloads.

play_circle_outline Demand for GPUs increasing as companies progress in adopting AI technologies.

play_circle_outline CoreWeave’s Operational Model: Accelerating Modular Data Center Deployment and Scaling in the Competitive Cloud Landscape

play_circle_outline Customer demand for GPUs impacting choices in hiring and research collaboration.

Info
Transcript

Mike Mattacola, Coreweave

Mike Mattacola

CBO CoreWeave

In this interview from the RAISE Summit 2025, CoreWeave General Manager of International, Mike Mattacola, joins theCUBE host, John Furrier, to unpack how the GPU-cloud pioneer is accelerating the AI-Factory era. Mattacola details CoreWeave’s deep collaboration with NVIDIA and Dell as the company races to deliver H100, H200, GB200 and the newly announced GB300 systems nine months ahead of traditional hyperscalers. He explains how CoreWeave’s Mission Control observability stack and vertically integrated build process turn racks of cutting-edge silicon into prod... Read more

explore Keep Exploring

What has been the approach and progress of the collaboration between the companies involved in launching new products? add

What are the key changes in system architecture and market dynamics that illustrate the evolution from traditional IT infrastructure to modern GPU cloud systems? add

What innovations and features does CoreWeave provide to ensure observability and continuous monitoring for the GB200 and GB300 platforms? add

What factors contribute to the unique advantages of using CoreWeave for GPU deployment compared to traditional data center options? add

What strategies does CoreWeave employ to manage and increase capacity in response to high demand? add

What factors influence researchers and engineers when choosing their workplaces in relation to computational resources? add

bolt Powered by CUBE AI

Mike Mattacola, Coreweave

search

>> Welcome back everyone. theCUBE's live coverage here in Paris, France. I'm John Furrier, host of theCUBE for the RAISE Summit 2025. It's a two-day conference here on the international stage in Paris. Obviously global gathering, a lot of US a lot of European. It's a global community, as we look at the sovereignty and the AI infrastructure expansion, bounded by energy, the growth is incredible. Of course, CoreWeave is leading the way. Mike Mattacola, GM of the International Group at CoreWeave. Mike, great to see you. Thanks for coming on, theCUBE. Congratulations on the CoreWeave success. We've got friends that work there. Get the AI Factory pumping on all cylinders. First question is, you get all the latest 300s?

Mike Mattacola

>> Yep. Yep. So, we were just talking about Dell. So the first Dell Rack, we received that and it's now in a lab and we're running it as hard as possible and getting it ready for customers.

>> I've been in this business for over 30 years, at theCUBE, 16 years. There's this phases of data center gear lovers, and then let's talk solutions. Now it's back to speeds and feeds. But I remember back in the day when you go into a data center, it's like, man, that's beautiful rack, it's wired, it's beautiful. Now I see the pictures of Dell's AI Factory and it's like a cathedral. They're beautiful, right? I mean, so NVIDIA has really done an amazing job of building, aesthetically looking, of course the demographics want to see nice gear and they're loud, but we'll give them that, they perform well. But just the whole resetting of the thinking around what a computer is has been really fun to be part of that wave. You guys are leading the way with NVIDIA. Talk about that relationship and how that's enabling you guys to bring that value to the market because a lot of investments, you're starting to see the harvest coming. People here are seeing lines of sight on visibility of the unit economics, the value extraction, and the robust entrepreneurial ecosystem as well. It's not just the enterprise, it's builders across the board. Give us your view on the relationship with NVIDIA and how that's enabling you guys to be successful.

Mike Mattacola

>> Yeah. So John, thanks for having me first of all. So yeah, look, the NVIDIA relationship goes back a long time now and one thing I would say is it's really hard and so ultimately we brought H100s to market first, about nine months ahead of the bigger players. We brought H200s to market first, GB200 and then as of last week we announced GB300 as well coming to market. So we've been incredibly busy. This doesn't mean that we're receiving... We may be receiving the racks and infrastructure at the same time as others, but we've really fine-tuned our ability to get hold of the infrastructure and get it online and running as quickly as possible as well. So I think that if you look at the relationship we have, obviously with GB200, 300, we've been working with Dell to get that launched to market and this has literally been a CoreWeave team, a Dell team, NVIDIA team working day and night in our data center to get this stood up and troubleshoot and online for customers as quickly as possible. And the complexities of launching Blackwell now, I mean every time something new comes up, you're going from hundreds of thousands of failure modes into the millions now, which you've got to have a software stack that is capable of understanding and observing everything that's going on and making sure it's running for customers. What we've done best is we've really fine-tuned the process of receiving the infrastructure, plugging everything in, testing it on our stack, and then get it in the hands of customers as quickly as we can.

>> What I love about your business is, is that one, and we'll get into the international things in a second, but what these GPU clouds have shown is that the system's architecture isn't yesterday's rack and stack and that there's-

Mike Mattacola

>> Very different. - ...

>> actually real density involved. HBM's a significant part of the success. And as you get stuff closer, getting that data as close to the processing as possible. Just as an aside, Intel's in free fall. Just in the past year, Intel's market shares dropped from 73% x86 to in the teens in one year. That's mind-blowing. That just goes to show you that the script has flipped. How would you share that narrative? Explain that to someone who's in the general purpose IT enterprise market who's like they're struggling with the generations of procuring gear or product and doing IT. I mean, besides the agents will be the new HR department, which by the way now believe that to be true because they are evaluated on tasks and they do a good job. But the engineering involved is significant. You guys have done the work early on and continue to with AI factories. What's different? How do you explain that to a common techie data center person or someone who's actually trying to leverage as much service to get that social property called the data working?

Mike Mattacola

>> So I think there's a couple of things. Obviously NVIDIA has been innovating at the speed of light. And if you look at the NVIDIA organization, we've all heard the stories, so it's very flat. There aren't lots of layers. Leaders have 10s or 20s of people working for them, and that just enables them to move really fast. Less meetings. Less one-on-ones. Just make a decision and go on the same day. So if you look at, you mentioned Intel and other legacy companies that have been in this space, they just haven't done that. So they're slow at the end of the day. But the way NVIDIA's adopted that is pretty unique, and we try and do that as much as we can as well and mirror that so we can move quickly. The other thing I would say, which I don't think it gets talked about often is the supply chain. So this is a gargantuan supply chain. So if you think of our world, we're working on acquiring power like years out, we have to, so particularly if we're building something from you, we need to get energy from the grid approved and then we need to get the site ready, build a data center. But the point you were making around infrastructure and how it's changing, obviously in some ways it's good because it's consolidating. So with GB200, you've gone from a diverse supply chain and now you're pushing that into A rack with a huge consolidation of everything going into a single rack. But managing that supply chain is really critical as well. So you can't just get the bright idea one day, Hey, I want to buy some GB200s and I'll order some, they're going to turn up next week. You need to be doing that two or three years ahead and making sure that everything's hitting the ground at the right time so you can move quickly for customers. So yeah, I think obviously the way NVIDIA does things, the innovation cycle, the speed they move out is really unique and their structure supports that. Obviously we do the same, but I think you've really got to be working multiple years ahead so everything can come together in the supply chain.

>> And NVIDIA's innovative too with NVLink and then KV Cache, it's really, they're engineering a lot of the fabric around the chips, which helps too. How has that changing some of the software paradigms? Because we're seeing now software stacks certainly at the platform engineering side, we've seen Kubernetes rise up certainly in the enterprise, become an orchestration layer for some of those platform DevSecOps, DevOps crowd. But now the AI stacks are emerging. How are you seeing that with your customers? What are some of the observations on how they're flexing? How are they standing up these factories? What do they run on? I mean obviously everything runs on Linux these days. But is there an operating system for the AI Factory? Do I load something? Is it pre-installed? I mean, I'm kidding, of course. But people are trying to figure out, okay, I got it, I want to run it, what do I do?

Mike Mattacola

>> Yeah, I mean look, if you think about the space we're in, so ultimately we want... The customer doesn't want to worry about infrastructure. So basically if you look at working with legacy clouds, you pretty much needed an infrastructure, your own infrastructure department or function in place to manage everything, make sure everything's working. For us our customers are really happy, they want to focus on launching new products as quickly as possible. It's a crazy competitive market and whoever is launching first and getting adoption is going to win. So they don't want to worry about infrastructure. So one of the things we've obviously done at CoreWeave and we have to keep innovating on with GB200 and GB300 now coming, a piece of our platform we call Mission Control and this is the observability piece. So you mentioned earlier the bill of materials for, there are so many, there's over a million, I think different items in a GB200, GB300 rack now, and we have to be able to see everything that's going on and monitor that continuously and make sure that we kind of get ahead of any potential issue so the customer has no interruption. So we do that through our Mission Control offering and we've really fine-tuned that. The reason we want to get everything as fast as we can is we can get it in the lab, we can run the hell out of it and tune our platforms-

>> And get into the customer's hands as fast as possible.

>> Yeah, exactly. - That service.

Mike Mattacola

>> So the customer doesn't care about all

>> of the failure modes on a GB200, GB300 rack, they just want

Mike Mattacola

>> to run their jobs right at the end of the day. So it's our job throughout Mission Control platform to-

>> I got to say my Telegram, my WhatsApp, my Threads are all blowing up in terms of direct messages. I want get your reaction to this too, the word is man, CoreWeave's got all the GPU action, so there's really clearly a demand curve. You guys are going up the S-curve big time. I'm sure you're experiencing it on your side. There's a huge amount of GPU. You guys are filling the void on GPUs right now.

Mike Mattacola

>> Again, I want to emphasize that, and this speaks to our uniqueness even more, lots of people are receiving GPUs. We've just built a stack. You know Chen well, you know Peter, your team's spent time with them. Their sole focus is on getting this infrastructure and deploying it as quickly as possible. And to be honest, we don't talk enough about the CoreWeave software offering as well. You hear about big data centers, lots of capital launching the latest stuff, but it's all made possible-

>> You're basically have an operating model, and that's what I'm getting to is that an enterprise has a choice build or go with CoreWeave. The service you guys have is getting great reviews on that front because you're fulfilling a demand and the time to value comes into not just GPUs because the software on the GPUs runs super fast and people are blown away. So we're getting a lot of signals, I just want to get your thoughts on that because that's a fast path to, one, hedging price, but also scoping what I want to do. So if I'm an enterprise, I got to scope what I'm going to be building or using and there might be multiple ways I do that consumption-wise. But at the end of the day, I don't want to wait around and there seems to be impatience and you guys seeing the same thing, I won't say, it's basically a bad word impatience, but-

Mike Mattacola

>> They should be impatience, the whole world is changing. The companies, large industrials need to adopt AI as quickly as possible. And a lot of the legacy... If I look at Europe, we're sitting in Europe now, it's a region that's pretty much owned by large industrials and they have lots of data and they need to adopt as quickly as possible. One thing I'm going to jump on a little bit here, when you talk about enterprise as well, if I take an example of the European market, my numbers might be slightly off, but typically it's been about a 60/40 split between on- prem versus cloud. So 60% on-prem and people building themselves. If you're trying to build on-prem with today's technology, we're at 136 KW per rack, we're going into the 200s pretty quickly, versus legacy would've been 6KW. It's like a piece of Lego, right? Compared to what we are doing today. If enterprises trying to do this on their own, then they're not going to have time to market. By the time they may have got a GB300 online, we'll be deploying Rubin, we'll be way beyond that. And they're just losing a lot of performance that they're not saving the money they could be. So particularly for enterprises, there's going to be a big change for them, I think particularly in Europe as well, moving towards working with cloud partners and they don't have the time and resource, the expertise to run this kind of infrastructure as well.

>> I think that's my point about the demand curve.

>> People are appreciative of the fact that they can get the service they need in a timely manner without all the heavy lifting. I mean, neo-cloud, specialty cloud, I mean it's a hyperscale environment for a service where their IP is on-prem. And that's happening, we're seeing that clearly. But let's talk about Europe because we're international, we are in France. It's interesting conversation. I love coming to Europe because I've been preaching to the choir, maybe to myself actually, about sovereignty. Not sovereign ai, that's a trendy topic. But sovereign cloud is a legit thing. And on-prem private cloud is happening. We're starting to see countries on-country kind of vibe where it's like on-premises seeing countries looking at their infrastructure, whether you're a multinational or a government. So you're starting to see sovereignty cross over the tech conversation into geopolitical where I was just on stage with founder of UiPath who's a Romanian, proud of his European heritage.

Mike Mattacola

>> Yep. - - Daniel was talking about,

>> when I asked him about his founder position, he was kind of preaching to the European crowd. Like come on guys, we got to go faster. So I think there's a huge opportunity. I'd love to get your thoughts as you look to expand out CoreWeave I'm sure you've got a green field of opportunity. What's it like? What's your plans? What's your growth strategy? How are you going to build out the international piece? Because it's ripe, I could smell this opportunity. You can see people leaning into the urgency of refactoring, whether it's the political scene via having an economy, entrepreneurship, I mean the enablement is off the charts.

Mike Mattacola

>> So sovereignty, love to have this discussion, it's really important. It's really important for Europe too, because if it's not addressed quickly, then Europe's just going to slow down and won't catch up with everybody else. I spend a lot of my time with governments all over the region. You kind of have to. The world we play in, lots of power, massive deployments, big investments in countries. When we're building a new team, deploying data centers, it's in the billions very quickly. So we typically talk to governments around that and how we go faster. The sovereignty piece, so the way I think about it, there's a lot of noise around sovereign AI. That's fine. The way I would break this down is into percentage of demand. So basically there will be certain workloads and it's going to be a lower percentage, it really should be truly sovereign. So in country, managing country in a true sovereign environment. The reality is true, true sovereign so you'd need local chips, so you need French chips, you need French only team, you need French networks or EU, whatever you want to call it.

>> It's hard to pull off. - If you go down that path

Mike Mattacola

>> completely, the capital doesn't exist.

>> It's hard to pull off. - The assets don't exist. It won't happen.

>> The ability to execute is almost nil.

Mike Mattacola

>> Yeah, yeah. I will offend some people when I

>> say this sometimes, but I still consider myself a European,

Mike Mattacola

>> but Europe's very good at talking, now it's time to do stuff. We need to move quickly, otherwise Europe's just going to get further and further behind.

>> If they stare at their navels, as they say, the expression goes, you'll be basically like, okay, nothing happened. The nirvana use case of sovereignty as you point out, the degree of difficulty is almost 100%. So I think if they can get over that, what happens next? Because then you now have an enablement. Andy Grove used an expression, we quote on theCUBE, let chaos reign, then reign in the chaos. In a way I like that philosophy because you got to put the guardrail, hate to use that word, it's overused, but just get it out there and see what happens.

Mike Mattacola

>> Yeah. I just said it, I'd say again, now is not the time for talking, it's for doing. So ultimately I'll tell you our story a little bit over the last 12 months as well. So we-

>> Yeah, go ahead. Please do.

>> We had no team, no data centers.

Mike Mattacola

>> Around a year ago I announced we were launching our

>> business in Europe. We committed to deploy about a billion pounds in the UK and $2. 2 billion in the rest of Europe. Fast-forward to today and we have six state center regions online. I planned for about four, we've done two more than we'd planned because demand has been pretty high. So what we did when we were approaching Europe, so we basically set up a footprint. If you look at legacy clouds, typically they were looking at the lowest cost place to run a workload in the far north of the Nordics as something that's cheap energy, but they weren't really designed for AI and for latency going forward as well. So we set up our footprint in the south of Norway, a little bit near Stockholm, and then Barcelona and Spain and then the UK. So we could basically service any kind of AI workload with low latency roughly within 2000K of a large city, like 20 milliseconds roughly. So we've done that and we've delivered it quickly. We are basically sold out. So we're adding more, we'll be adding another region about 90 days. So back to the point around sovereignty, and the point I was making earlier, there's going to be enough demand that we'll keep serving customers for typical workloads and then there will be some sovereign workloads and governments need to work with private and make that happen. But I think we've got to be really careful that we don't over blow this and think everything needs to be because it doesn't and we'll just be 10 years late on everything.

>> Yeah, that's totally a great point. Mike, that's a great point. Now let's get into some of the, I love the distributed aspect of it, and if you don't mind I'll drill down on that. You mentioned the latency was a design criteria. You're basically distributing hubs of super computing and you're sold out. How do you GM that? You got to do more, you got to get more. So what's on your focus, what's on your agenda? Get more space, procure more power, what's in your plane?

Mike Mattacola

>> One of the things the CoreWeave's teams done incredibly well is we've had crazy demand, but we've always delivered more capacity quickly. So back to the point we talked around our stack, it's been fine-tuned to be able to build data centers, test them and hand them out to a customer as quickly as possible within literally weeks, weeks and months. So in terms of I'm not too worried we may be sold out, but I'm adding more capacity very quickly.

>> You are adding more, you're focused on that.

>> Yeah. So we'll always stay ahead of the curve there.

Mike Mattacola

>> And the beauty of how we've built the company is that we can add increments, modular increments of capacity very quickly. So we may have a large chunk of power, but we'll do six deployments to fill it out. So we'll deploy to demand.

>> You have a blueprint for building out these computer centers. It's not buying real estate and throwing gear in there. You guys systematically come in here-

Mike Mattacola

>> Yeah, we're vertically integrated so we don't contract anything out. That would be a terrible idea.

>> Yeah. Of course. - So if you look at what we did in

Mike Mattacola

>> Europe, so we literally hired people, sent them to the US, they joined a build. The engineers worked together, got comfortable with the stack, and then pretty much they had a little bit of help to launch the first regions here. But now they're self-sufficient. So we built a team in less than a year that can deploy large, large- scale data centers very quickly. So we have the automations, we have the tech to do that.

>> So I have to ask a personal question if you don't mind.

>> What's the day in the life for you? You wake up, work out and probably not have time for that. But take us through, in all seriousness, what's it like right now? Because again, the demand is you're selling out, so it's like you can't build them fast enough. The demand's high. I know you guys are working hard. You mentioned Chen Goldberg, she's the engineering lead now, what's it like on a day of life for you? What goes on? You talking more politics, you do more engineering, customer procurement, customer delivery? What are some of the mix of your time? How do you spend your time?

Mike Mattacola

>> Yeah. I mean, look, so for me, I wear a lot of hats and I took this challenge on personally because it was exciting to build something from nothing.

>> Of course. - And I'm not the guy, to be honest,

Mike Mattacola

>> to sit in one department and do that for the next 10 years. So I can't sit still, it'll be a disaster and everyone would hate working with me. My days, they're really, we do a lot of this, right? Obviously we've been working pretty hard to get our brand out in the region as well. We like to do it through execution and then talking about what we've done versus promising everything and then trying to catch up. So we put our money where our mouth is and we've executed. So my days, I get up, I work out most mornings. Again, I'm a hectic guy. So if I didn't do that, I'd be a disaster to work with.

>> You're busy. You're super busy.

Mike Mattacola

>> Yeah. But you'll be looking at,

>> is there something slowing down delivery? Sometimes I'll need to jump into a supply chain. Sometimes I'll need to talk to the engineering team and work through things with them. But wherever you need to be just to remove constraints and go faster.

>> We appreciate what you do. It enables an entrepreneurial scene here, helps companies, the economy impact is huge with what you guys do. So we appreciate what you do and a lot of people do. I have to ask you about supply chain and NVIDIA because you mentioned that earlier. What I love about NVIDIA right now is that they're executing really flawlessly on the supply chain. Jensen puts out the roadmap. He brags about it, of course, as he should. On the other side of the marketplace he's got the ecosystem. Could you share your thoughts on how that ecosystem's developing? Because he's lining up a lot of enterprise scale on ecosystem because now you've got supply chain on one side and ecosystem on the other, that's going to set the table for enterprise and other kinds of connective tissue into the systems.

Mike Mattacola

>> Look, I think I go back to the point I made earlier, a couple of things. So the next round of demand for us and for the market is enterprise. So in Europe it's really more of an enterprise play anyway. You just don't have the VC money and the growth capital you do in the US. So it's more of an enterprise play anyway. But in general, we've got to adjust to that and be able to work with, we've been building 20, 000, 50, 100,000 GPU clusters for large labs and now it's going to be more helping a large enterprise take two machines and get comfortable using them, run a POC, get some results, sell it to the CFO, and then grow from there. So that's a completely different business. In terms of the supply chain, that's not their problem. That's our problem with NVIDIA as well.

>> You take it on. - So ultimately that we are the interface

Mike Mattacola

>> with the platform they're running on.

>> But we have an incredible team. We work closely with NVIDIA to make sure the supply chain is humming every day, and that shouldn't be anywhere near them. They just want compute. The big challenge with enterprise coming in as well, it's a completely new skill set. So if I'm in an enterprise right now, if I want to deploy AI in my business, I'd hire like 21, 22 year olds that are comfortable with it, using it every day and build a little function and let them loose and build stuff. I think it's a new model that-

>> It's interesting.

>> First of all, it's great that you guys have that I call it operating service because it's hardware, software, Mission Control. I have to ask you, just kind of last question is I'm seeing a lot of reactions to folks when they use these super cluster, super computer AI factories. It's almost like they've been starved all their life and they see a big meal. And this is a research class emerging. You're seeing a lot of the big dollars go towards people who are waiting for the world to get to a point where it just was unreachable before. So you're seeing new use cases. So share some examples of where... And that's where the research classes emerging, I don't mean academic, I mean hard problems. It could be related to climate. It could be related to life sciences. Tons of examples where it's like, oh my God, I have to wait in line for super-computing power. The HPC market's been kind of slowly inch by inch, and all of a sudden boom. Share some stories around aha moments, some fun examples of real needle-moving examples?

Mike Mattacola

>> Look, I think, so to your point earlier, I think people, some of the best researchers, engineers out there are deciding who they join based on how many GPUs they have. We hear that all the time. So it's good for us. So I think where things are going, we've been working with large, some of the largest labs in the world and building some of the largest super- computers in the world as well. And obviously they've been training LLMs and things are going to move to agents, they are moving to agents, agents is kind of next and that's going to change all of our lives and how we do business every day, how we live, connecting all the dots between different data sources. None of us really know what it's going to be like in two or three years. So I think that's the next phase of what our customers are... I know that's what they're working on now, but they need so much compute to do this and the type of compute as well. So if you think about how we build our clusters, we build them for inference and training. We don't want you to move regions, but that's great for us as well because as we move towards test on compute and agents and other things as well, we're ready to go.

>> I was talking to a customer of yours already, probably, or a prospective customer. I won't say his name, I just gave away the gender. But he used the word at NVIDIA, he runs the products for a major region outside the United States, in charge of the entire buildup of the country. And he said to me, and I want to get your reaction, we'll close it out after that. He says, picking a provider is picking a university. The culture matters. So what's the culture of CoreWeave? How would you describe culture of CoreWeave? Because I kind of see his point. You have certain philosophies, it's kind of like, hmm, do I go to that school or that school? They specialize in that, this is the vibe, it's a party school, whatever. It was kind of interesting because research is a big part of the innovation cycle of the providers. First of all, do you agree with that and then how would you describe CoreWeave as a culture?

Mike Mattacola

>> Yeah, look, I first really agree with that. So the reason we've been able to grow so much and so quickly with our customers and with some key customers as well because it just works. Our infrastructure works. To be blunt, we've worked with customers and they've come back to us and didn't have a great experience and they want more CoreWeave, right? So we've really built something that, back to the... If you are a startup and you've just raised money, 50 to 70% of that's going to go to your cloud bill at the end of the day. So you want to spend it wisely with the right partner. And we have a pretty unique, to touch on the culture side of things, probably the key value in our business is act like an owner and people truly do. It's not just wallpaper. So know our CTO, Peter, who some of your team has met Peter before. So he is in customer Slack channels. He is answering tickets. He's challenging everybody to move faster and fix things. If you look at what our founders did in terms of in the capital markets and finding creative ways of raising large amounts of debt so we could build this business as fast as we did, that's pretty unique as well. So there's a lot of innovation there we probably don't get credit for. And then just the whole company in general, we have about 38 data centers in production as of this month. We had-

>> You're executing culture. Ownership driven.

>> When I joined a couple of years ago,

Mike Mattacola

>> I think we were in the early teens and we've deployed, but these are not small data centers.

>> They're monster data centers.

Mike Mattacola

>> So we execute fast

>> and people are having a great time, to be honest, they have a lot of fun. It's not boring at CoreWeave. And we've put a pretty unique team together.

>> Not for the faint of heart and you got to be smart.

>> I mean, people like to work for companies that's solving big problems. And also you're doing that and your customers are. Thanks for coming on. I really appreciate your time. I know you're super busy here-

>> Great to meet you. - ...

>> in this international event.

Mike Mattacola

>> theCUBE is global, of course in Paris. A lot of challenges, but the opportunities and the demand for super-computing power, the AI factories, H100s, H200s, GB300s are out, Dell AI factories, of course all the actions coming. Harvest time for AI, that's the big story here. Thanks for watching.