Lukas Gentele, vCluster Labs | theCUBE + NYSE Wired: AI Factories - Data Centers of the Future

Palo Alto studio connecting Silicon Valley and Wall Street. >> I'm John Furrier, host of the Cube here with Dave Vellante, my co-host. Hello, I'm John Furrier, host of the Cube here in the Cube's NYSE studio, of course. We are Palo Alto studio connecting Silicon Valley to Wall Street. This is part of our NYSE wired program and community. We're talking to all the leaders. This is our AI factory series, the data center of the future. We talked to the leaders who are building out the technologies to bring in more AI infrastructure to faster to enable more intelligence in the enterprise and throughout the world as sovereign cloud and many other things are emerging. It's all cloud native meets AI native. Lucas Gentili's here, co-founder and CEO of V Cluster Labs. Lucas, thanks for coming in. You've been working very hard with your team on what we call the Kubernetes KubeCon community world. Um those cloud native days were really, really strong. We saw the rise of AWS, the hyperscalers. Who thought there'd be more clouds? But now we have neo clouds and AI clouds, but a lot of that early cloud native work done from, let's say, 2013 to 2020 was building on the building blocks of cloud native technologies, higher-level services. Great. Check the box. In comes hybrid cloud. Now you have distributed computing. So all the things that you were doing are relevant for all the AI infrastructure hyperscalers emerging cuz they have to stand up massive scale so fast. So explain what you guys do. I want to get into this cuz this is a market force we're seeing with the rise of the neo clouds and the neo cloud labs and among other infrastructure providers. >> Yeah, who would have thought, uh you know, even just like five or six years ago that, you know, it wouldn't just be a couple of large hyperscalers, that there would actually be 200-plus new infrastructure providers in the world. And I think there's new ones popping up every day, to be honest. What our company really sets out to do is help them build the software layer on top of the hardware. Obviously, you know, they're buying amazing hardware from folks like Nvidia, for example. And then the question is, how do you deliver to customers? And you know, some of these new labs and AI native companies, but also traditional enterprises were starting to run workloads on GPUs to fine-tune models, for example, open-source models is is a major trend trend these days. And you know, they typically want an AWS-like experience in the sense they don't want to start from the metal with BMC access. They want to start with at least SSH, you know, provisioned machines and >> Start with the basics. >> Yeah, be able to relaunch a machine and relaunch a machine. >> neo clouds. Let's get into it because this I think this is a really going to be a great opportunity. I wrote a blog post on Friday that said, you know, the money goes where the constraints are. The biggest constraint we're seeing in your world is the ability to orchestrate workloads in a single tenant-like experience on a multi-tenant cloud cuz all these clouds are multi-tenant, meaning multiple people are in there, companies. But now you have each company with multiple workloads. You can almost apply some tenancy to that. So, you got to orchestrate this stuff. You got to manage resources. And it reminds me of microservices. You got to like know what's going on with the agents. So, there's a lot of intelligence at the infrastructure level in the new AI infrastructure architecture. Unpack that for us. Explain what's going on cuz that's where the action is. >> Yeah, the scale is just so much larger, right? You're seeing so many nodes in a Kubernetes cluster or in a Slurm cluster, for example. And you know, if a node fails, you know, your training job might be at risk, right? So, so you really got to make sure that you're checkpointing correctly and like automatically trying to heal infrastructure. And then when you know, a lease with one tenant is over, for example, you might need a resize and you know, restructure your network and and all of that needs to be driven by automation. >> Talk about the economics cuz you're obviously as co-founder you're in founder mode right now. You got a great opportunity. Um but the economics in this AI area isn't just about cost takeout like the cloud native game and squeeze more efficiency and enable the IT and developer shift left all those things we covered. This is revenue. You're starting to see real economics. And Jensen's earnings this past week last week highlighted the fact that they're now calculating revenue at Nvidia on the gigawatts that they're serving. Okay. So now you're starting to get into the weeds of unpacking what's inside the data center. So I'll give you an example and I want to get your reaction. Little things matter. A picojoule here. Um put some photonics in there. Training node not dying. There's consequential revenue impact. Explaining this important nuance. It's very nuanced but I think people are squinting through all the high-level stuff and going, "Okay, I can tell if something's going to break. I got to mitigate that cuz I can quantify the revenue." >> Yeah, I mean you just have to look at, you know, the prices that uh that new clouds are able to command. Uh you're looking at a a Nebius and a CoreWeave and the reason why they demand such premium prices is they're proven to work at scale. They have a high degree of automation and they feel, you know, as resilient as a hyperscaler in some ways but they don't have the traditional, you know, I think benefit of having to run an availability zone where literally you you have duplication. With GPUs just imagine somebody would have to pay twice as much, right? In order to get that level of redundancy. That's just not the world we live in anymore and automation really needs to help you um to optimize, you know, what you can deliver in terms of SLAs to your customers. >> If you look at the cloud reference architecture NVIDIA puts out, I mean, they call it out front and center. Now, they have a very dense architecture. There's a lot of KV cache and Dynamo going on. There's a lot of networking. But, if you look at their reference architecture to how they certify who's going to run Run Vera Rubin or whatever, the word Kubernetes is everywhere. I mean, it essentially reads cloud native to me. >> 100%. >> Explain the importance of that and why that's so mission-critical and why NVIDIA and others are making that choice. Is it because it's stable? I mean, this is like it is not like cloud native and AI native. It's all one thing now. >> Absolutely. When when we started talking to CoreWeave in the very early days, CoreWeave was like maybe 20 people at the time and we were we were even smaller. We we might have been like 10 people. >> Hey, fellow fellow travelers. They're doing pretty good. You can follow their [laughter] path. I'm sure you'll be very happy. CoreWeave's got great team and then they got they've had explosive growth. >> Yeah, the the reason they started talking to us is Kubernetes was so front and center to their strategy. And at the time, just like everybody else in the Kubernetes space, we were focused on Fortune 500 companies, large enterprises, you know, nobody would have foreseen that a CoreWeave might have actually the chance to become larger than some of these companies and more significant in terms of the structural relevance to you know, our everyday lives today to be honest. AI is so entrenched in everyone's life and and work life as well. Um so, these these infrastructure companies become a central backbone of the entire economy ultimately. >> They're AI infrastructure clouds, basically. They're IaaS for in- AI. >> Yeah. >> Um it's interesting. If you ask me, I mean, CNCF and the KubeCon event we've been through many times, uh both you you and you and I, that was because OpenStack failed. If people don't know what OpenStack is, check out OpenStack. That was the that was the whole open source effort to try to replicate and compete with AWS. That started around 2010. Actually, when Cube started, we were involved in those early discussions with Rackspace. >> So, I was going to say a lot of OpenStack and some of the neo clouds today, though. >> Yeah, oh yeah, oh yeah, and telcos. Because the bones that they built in that structure yeah, were legit. Then in comes KubeCon, which is an ecosystem opportunity that aligns with the hyperscalers. Almost a perfect storm for a CoreWeave to emerge because you have a lot of OpenStack open source stuff available. You know, it's a lot of bare metal, a lot of kind of core principles, but you bundle in kind of cloud native maturization of Kubernetes. >> Yeah. >> And all those the work of the Linux Foundation, which that was a beautiful model that a thousand flowers bloom, let the best win. You now have all the ingredients for CoreWeave to saying, "Hey, I can compete with AWS in this very narrow growing space called AI training inference." >> Yeah, and that was a good point. >> me in 2015, would there be another AWS? I probably would have said, "I don't ever think there'll be another hyperscaler." It's just too high of a bar to build out. >> Yeah, that's what I thought at the time as well. This is this sounds like a wild plan to try to compete with AWS on GPUs, but they proved the market for everybody else, and you know, I think Nvidia actually has a really smart strategy there in also fostering that ecosystem. You know, we've been on Jensen's slide at the GTC keynote a couple of times, and they're really lifting up and highlighting the partners cuz they're not trying to you know, own everything. They're trying to provide the building blocks and support folks, but they're leaving enough room for everybody to to flourish and grow. >> I think Lucas, I think we just illuminated something that's never been written about, but I would just we'll just call it out here and just we capture it. The work of the open source community around OpenStack and Linux Foundation built the AI infrastructure of in what Nvidia Nvidia is actually just everything. Not copying, but they're implementing the similar principles. >> Yeah, they're definitely launching something >> play, it looks a lot like CNCF for Nvidia. >> Mhm. >> Everyone's in there. So, open source >> Yeah, Nvidia is open source and a lot of other >> at Open Compute. Let's highlight another one. >> Yeah. >> Between OpenStack which became the fertilizer in some cases, you know, piece parts for cloud native. And you combine Open Compute. There would be no rack-scale system without Open Compute. >> Right. >> Cuz they basically created the format for the rack scale. >> Mhm. >> So, open source is a huge part of the innovation. What's your reaction to that? What's your thoughts? >> I mean, that's the beauty of open source. I think everybody understands the power of open source. And in the age of AI, there's some discussions how open source might change and how it's so easy to wipe code things that it might not need open source. There's some challenges in open source as well with, you know, maintainers being spammed with, you know, auto-generated, non-really valuable contributions and credit for that. >> code for open source. Also, vulnerabilities. >> 100%. Vulnerabilities get detected much, much faster, of course. >> Thank you very much. >> They might get introduced in a in a much sneakier way, right? Like, they it's very, very interesting. >> What's your take on the Hugging Face and video news that hit today? We we reported on Silicon Angle last week uh that was coming. Um We expect that to be in the plan when they started open source. Um I mean, Hugging Face has become a basically a direct site for projects. Now, they're calling it the registry. Um of open source projects. What's your take on that? good for the industry? Good for Nvidia? Good for everybody? What's your take? >> I I believe it shows another commitment of Nvidia really endorsing open source. In this case, really clearly open source models, right? And and inference on open source models. And they've been investing in open source, particularly this year, very heavily uh with DSX and all the DSX OS tools, for example. Uh Nico, which is an infrastructure controller to stand up uh servers. Uh there's a lot of observability tools that Nvidia has open sourced a Rex scale management system recently and those are all great building blocks to build on top of and that they open up the conversation with the community around what should be the open standards to build these neo clouds and I think that's fascinating. >> I actually was one of the people that was kind of pointing out to Nvidia that they that their homogeneous system they didn't like that um they're not really homogeneous but I was just saying if you go down to Nvidia >> Mhm. >> you become the mini computer, the proprietary engine. Uh, they listened not to me but other people as well. They're all about open now. >> Yeah. >> So, and within 2 years Nvidia shifted to open. Why? Because they got the best product. Makes a lot of sense. That kind of squashes the whole Nvidia is an open argument. >> Mhm. >> So, open source will continue to thunder away. So, my question to you is where do you see the next constraint that people going to put either direct capital at or entrepreneurial effort? What's the big areas that need the most work right now that are constraints? >> I believe there's lots of areas that are constrained today. Um, I think we can have a small contribution on our part to make sense of all of these open source solutions and help folks turn them into a concise platform uh, to actually hand out to customers. We really you know, pull them together as an experience layer for the end customer. That's our contribution on that end and people definitely are strained on that. Most neo clouds they might have one person, you know, working on the software stack or two people. A lot of them are 50 people company. >> you mentioned CoreWeave, Nebius, this N scale, there's Argentum. They're all doing billion dollar deals. They got to stand this stuff up, build I should say build out and then turn on as fast as possible. They don't need to have a team of six people go figure out Kubernetes. They need operating help. This is kind of where you guys come in, right? >> 100%. I think for maybe the top 10 really leaders in the space, we become point solutions that they can build on top of. Nebius is a good example. They're great customer of ours. We helped them in a very specific area with the Nebius token factory and that's amazing that we can have a contribution to this. But then there's other providers. You think of a Boost one or Corvex, right? They don't necessarily have the scale of a Nebius yet, but they're also signing deals left and right and becoming very very successful neo clouds. And for them we can do even more. We can become really a platform to be to build on top of to do a lot of things for these guys. >> All right, talk about your company, the momentum you have and why people are using you and what's your headroom in terms of more growth. >> It's growing like the whole space. I think you know, if you look at our revenue growth, it's through the roof. I just had a board meeting yesterday actually and the investors were like, oh wow, the momentum is continuing. This is incredible. >> Keep going with what you're doing. >> Absolutely. And then you know, our team has I think doubled or tripled in the past 12 months. We really we can't even hire fast enough. Yesterday in the board meeting I told my investors we're pretty much cash flow positive right now and we didn't plan on doing this. We're just like >> We did it. >> We have a hard time catching up with hiring enough people. So it's an it's an incredible time to be in this space. >> Who you looking for right now? Put a plug in for potential hires. Um Where are you on the funding letter? Where are you guys at? >> So we raised the series A from Khosla Ventures. They were our lead investor. That's a little over 2 years ago at this point. Very exciting to have an investor like Khosla on board. They were the first check in Open AI. They've made so many huge bets in the AI space. >> They know infrastructure. >> They do know infrastructure. >> and know infrastructure. >> 100% yeah. >> [laughter] >> He's a tough tough investor. He also understands the founder's role. >> 100% yeah. Just look at this past companies and where he's where he was involved. >> He knows he he knows infrastructure. That's a great partner. All right, so in terms of hiring, what roles do you have open? Put a plug in. What kind of people you looking for? Is it Kubernetes, cloud native, AI native, scientists? >> Yeah, anybody who can operate on the BMC layer, anybody who's operating on the networking layer, anybody definitely with Kubernetes experience. Uh we also have a CFO role open and a CMO role right now. So we're building out the leadership team very clearly. [music] Uh yeah, I think we have 35 open job >> And you know you know you going to do a series B soon. >> Yes, that is another topic. >> get some more cash. That'll help. >> [laughter] >> Yeah. Um well, congratulations. Now what are you focused on now? Obviously great momentum. Um growing like crazy. Love the opportunity. And again, I think this is a great example you you guys are doing. There's any scale sold to N N scale. That was a sign that okay, we're we're going to start to see that. I won't say middleware cuz it's a bad word, but like a operating layer of organizing those resources become mission critical and you can quantify it. >> I think our big focus right now besides, you know, the 150 plus clouds that are at a sizable scale, we're also focusing really on the rising stars. Um Grok is for example a customer of ours, and those have potential to become massive players in this industry because they have, you know, huge amount of capitalization. Uh they have a huge amount of uh power committed and and data center sites, and they're bringing chips online like crazy. We're talking about dimensions of 50,000 plus chips in the next 12 months. Uh that's incredible to be part of those journeys, and we want to make sure we're a partner for them from everything we've seen in the industry and enable them to build on top of us and with us and with our expertise. >> Well, listen, congratulations on all the momentum. Again, you're in a good spot. Uh we've been covering it. We see it. Thanks for coming on. Appreciate it. >> Thank you. >> See AI Factory series the data center of the future part of the Cube and the NYC wired program and community here in New York City. Of course, we have our Palo Alto studio connecting Silicon Valley to Wall Street. I'm John Furrier, host of the Cube. Thanks for watching.