In this theCUBE + NYSE Wired segment from “AI Factories – Data Centers of the Future,” Nebius co-founder and CBO Roman Chernin sits down with theCUBE’s John Furrier at the New York Stock Exchange to unpack how AI factories are reshaping enterprise infrastructure and the future of data centers. Chernin outlines Nebius’ two-track strategy: a multi-tenant cloud built for developer experience and managed services, and large-scale, mostly bare-metal deployments for hyperscalers and AI labs. He discusses the significance of Nebius’ Microsoft deal (described as “up to $20B” and set to become one of the largest single-site GB300 deployments) as both an engineering milestone and a way to feed scale and cash flow back into the core cloud business. The conversation explores why enterprises want “the baby of supercomputer in the cloud,” marrying cloud flexibility with supercomputing efficiency to minimize time-to-value without sacrificing performance.
Chernin details Nebius’ specialization in AI-centric workloads (large distributed training and inference at scale), a platform roadmap that moves beyond infrastructure into inference, fine-tuning and reinforcement learning as services, and a commitment to helping customers build on open-source models for control, cost and data leverage. He traces customer waves from foundational model builders to vertical AI companies and tech-forward enterprises, noting early traction with firms like Shopify and momentum in regulated sectors such as healthcare following Nebius’ compliance milestones. With roots in Yandex’s large-scale engineering culture and meaningful exposure to ClickHouse, Chernin also weighs in on the economics of AI-scale infrastructure (power and capacity as gating factors), hybrid orchestration and sovereignty, and why latency priorities vary by use case – from reasoning models to voice agents – as AI factories become the new unit of value in modern enterprise compute.
Forgot Password
Almost there!
We just sent you a verification email. Please verify your account to gain access to
theCUBE + NYSE Wired: The AI Factory - Data Center of the Future. If you don’t think you received an email check your
spam folder.
Sign in to AI Factories - Data Centers of the Future.
In order to sign in, enter the email address you used to registered for the event. Once completed, you will receive an email with a verification link. Open the link to automatically sign into the site.
Register for AI Factories - Data Centers of the Future
Please fill out the information below. You will receive an email with a verification link confirming your registration. Click the link to automatically sign into the site.
You’re almost there!
We just sent you a verification email. Please click the verification button in the email. Once your email address is verified, you will have full access to all event content for AI Factories - Data Centers of the Future.
I want my badge and interests to be visible to all attendees.
Checking this box will display your presense on the attendees list, view your profile and allow other attendees to contact you via 1-1 chat. Read the Privacy Policy. At any time, you can choose to disable this preference.
Select your Interests!
add
Upload your photo
Uploading..
OR
Connect via Twitter
Connect via Linkedin
EDIT PASSWORD
Share
Forgot Password
Almost there!
We just sent you a verification email. Please verify your account to gain access to
theCUBE + NYSE Wired: The AI Factory - Data Center of the Future. If you don’t think you received an email check your
spam folder.
Sign in to AI Factories - Data Centers of the Future.
In order to sign in, enter the email address you used to registered for the event. Once completed, you will receive an email with a verification link. Open the link to automatically sign into the site.
Sign in to gain access to theCUBE + NYSE Wired: The AI Factory - Data Center of the Future
Please sign in with LinkedIn to continue to theCUBE + NYSE Wired: The AI Factory - Data Center of the Future. Signing in with LinkedIn ensures a professional environment.
Are you sure you want to remove access rights for this user?
Details
Manage Access
email address
Community Invitation
Kush Bavaria, Ornn
In this theCUBE + NYSE Wired segment from “AI Factories – Data Centers of the Future,” Nebius co-founder and CBO Roman Chernin sits down with theCUBE’s John Furrier at the New York Stock Exchange to unpack how AI factories are reshaping enterprise infrastructure and the future of data centers. Chernin outlines Nebius’ two-track strategy: a multi-tenant cloud built for developer experience and managed services, and large-scale, mostly bare-metal deployments for hyperscalers and AI labs. He discusses the significance of Nebius’ Microsoft deal (described as “up to $20B” and set to become one of the largest single-site GB300 deployments) as both an engineering milestone and a way to feed scale and cash flow back into the core cloud business. The conversation explores why enterprises want “the baby of supercomputer in the cloud,” marrying cloud flexibility with supercomputing efficiency to minimize time-to-value without sacrificing performance.
Chernin details Nebius’ specialization in AI-centric workloads (large distributed training and inference at scale), a platform roadmap that moves beyond infrastructure into inference, fine-tuning and reinforcement learning as services, and a commitment to helping customers build on open-source models for control, cost and data leverage. He traces customer waves from foundational model builders to vertical AI companies and tech-forward enterprises, noting early traction with firms like Shopify and momentum in regulated sectors such as healthcare following Nebius’ compliance milestones. With roots in Yandex’s large-scale engineering culture and meaningful exposure to ClickHouse, Chernin also weighs in on the economics of AI-scale infrastructure (power and capacity as gating factors), hybrid orchestration and sovereignty, and why latency priorities vary by use case – from reasoning models to voice agents – as AI factories become the new unit of value in modern enterprise compute.
In this interview from theCUBE + NYSE Wired: AI Factories - Data Centers of the Future, Kush Bavaria, co-founder and chief executive officer of Ornn, joins theCUBE's John Furrier to discuss how compute is emerging as a financeable, tradable commodity at the center of the AI supercycle. Bavaria frames the core challenge with a sharp analogy: buying compute today is like purchasing a private jet when most companies only need an airline seat. Ornn aggregates fragmented demand from startups and smaller enterprises, packages it into creditworthy off-take agreement...Read more
exploreKeep Exploring
What does ORN do, what is its vision and origin, and what challenges in the compute market are you addressing (including what you mean by "off‑take")?add
How does ORN's model enable companies to buy compute capacity without requiring large, creditworthy long‑term off‑take agreements?add
How did the company originate — what were the founders’ backgrounds, motivations, and mindset that led them to start it?add
What product are you building, how do the indices that track compute (specifically NVIDIA chips) work and where are they traded, and what are your long-term goals for the platform?add
>> Hello, I'm John Furrier with The Cube. We are here in the Cube's NYSE studio. Of course, we have our Palo Alto studio connecting Silicon Valley and Wall Street, part of the NYSE Wired program and community. This is our AI factory series. We highlight the leaders and the entrepreneurs who are making it happen. As AI factories continue to industrialize compute, storage, networking, and data, we're starting to see real change in the markets with AI on the horizon at full scale. These AI factories are pumping it out. We got an entrepreneur in the studio here who's part of this next generation. Kush Bavaria is the co-founder and CEO of ORN. Kush, great to see you. We had breakfast yesterday, MIT, CS, math, big MIT pedigree, but you're targeting a unique market opportunity, the intersection of technology and financial markets around these AI factories because they are industrial centers. Jensen Wong said at his keynote at GTC, monetization is going to be a big part of the tokens that are being pumped out of these factories, obviously enabled by them the supplier of GPUs. But you got AMD, but you have a whole industry forming. This is kind of an approach that you're taking. I want to get into it. Thanks for coming on. Appreciate it.
Kush Bavaria
>> Of course. Thank you for having me. It's wonderful to be here.
John Furrier
>> So talk about the vision, because obviously the world is now known AI factories. We were tracking this for two years. We were early on this. Actually, three years ago, we called them large scale clusters because the name AI factories wasn't around, but they're going to be everywhere. Centralized, big factories, Edge, Metro, you're going to start to see the compute paradigm evolve radically. And it's a business model problem and a technical challenge. Explain what you guys do. What's the vision? How'd this all come together.
Kush Bavaria
>> ORN mission is to help unlock intelligence for everybody. If you think about it, intelligence is going to be the most important economic input for the 21st century, but it's only as strong as the underlying compute infrastructure underneath it. And so what we do at ORN is we help turn compute into a financeable, transparent, and more reliable asset class in itself. Today, if you look at the market, it's very opaque, it's immature, and there's a lot of issues that relate to the compute market as we spoke about. And where we want to turn this fragmented demand into a long-term credit worthy off-take, so there's more intelligence for everybody in the world.
John Furrier
>> You mentioned the word off-take, you mentioned there were issues. Explain the challenges. What are the problems that you're solving? And what does off-take mean for the people who don't know what off-take means?
Kush Bavaria
>> Yeah. So if you think about it this way, let me give you a good analogy. Today, if you want to buy compute, it's very much like buying a private jet to fly from New York to San Francisco. What ORN does is instead of buying a whole private jet for one person to fly from New York to San Francisco, we let you buy an airline seat and a large scale cluster. So we turn this long-term off-take where if you're a data center builder, you need to have a credit worthy tenant such as Microsoft, Google, Amazon, or Oracle come in and basically off-take your cluster from you, and then they will go in and sell that cluster to different startups and companies on the other side. What ORN does is it turns this sort of fragmented demand on one side with startups and other people that are not creditworthy in itself. And we bring that together, make it credit worthy and off-take it that way. So instead of just buying a full private jet, you could buy an airline seat as a company.
John Furrier
>> Yeah. I wanted to get that out there because I think this is an important nuance and it highlights the opportunity that you guys are going after and others. That off-take means that I'm a data center. I could maybe been doing Bitcoin mining for a decade, made a lot of Bitcoin, but I got the energy. I got a bunch of GPUs. Maybe I get some more from NVIDIA, but then the big hyper-scalers have huge demand. So they need to go get that seat, if you will, which makes sense. And they got big wallets, but startups don't either. So you're democratizing access to allow the builders, the people who are deploying the heavy infrastructure to do their thing, which has its own economics, and they're going to hit the demand curve on the supply side. Is that right?
Kush Bavaria
>> Of course. Yeah. I think there's a lot of people that want to build models post-trading, pre-trading. There's different stacks in this sort of AI model ecosystem. And we want to make it accessible for everyone. Whether you're a small startup that's 10 people, 100 people, 1000 people, no matter what the size of your company is, you should have access to these accelerated compute units. Because as I said before, intelligence is becoming ... It's the most important economic input for the 21st century, and we need ways for more and more people to contribute to this economy.
John Furrier
>> And what's the current situation? If you had to kind of put the classic old way, new way, what does this solve? What's the alternative? Without you guys, what does the startup, what does that large enterprise or even hyper-scaler need to do?
Kush Bavaria
>> Yeah.
John Furrier
>> Provisions, a lot of grunt work. What's the value of having that service?
Kush Bavaria
>> I think ORN brings more of that financial layer and transparency into the market. If you look back a hundred years ago, farmers wouldn't plant corn or before they knew that they could sell corn at a predetermined price in the market. So Wall Street came up with futures contracts, right? And if you think about today, the farmers in today's world are the data centers themselves. And so what we do is the chips are the corn, and we let data centers pre-sell in a futures market their compute capacity the same way farmers were selling 100 years ago.
John Furrier
>> It's interesting. I love that farm analogy because, one, it speaks to the financial side of it which you guys are going after. We'll get to that in a second. But also they have workers. They have farmers. Those are agents now. So tokens out of these factories are fueling both the unit of work and the land itself.
Kush Bavaria
>> Of course.
John Furrier
>> AKA the data centers. Okay. So now translate this into what that means for a business. Why is a business model opportunity? Is it a technical challenge? Is it both? Explain kind of the impact of what you're doing if this continues and it will, what does that look like?
Kush Bavaria
>> Look, we want to bring transparency and create an exchange for compute as a commodity. We believe compute will turn into every other commodity has existed. If you look at oil markets, electricity markets, any other commodity that exists has a market. It's traded right behind us. But compute doesn't have a sort of economy that exists around it. And so for ORN, our job is to build out this whole economic layer and this whole compute as an asset class infrastructure that allows people to hedge, finance, and trade compute.
John Furrier
>> All right. I'm looking at a press release from October. You guys announced your seed financing, modest around 5.5 million, not a monster round, but I like what you guys put in here. I want to get your thoughts about what's changed. Back then, you said the raise supports your mission to turn compute into a tradable commodity. Is that still the mission, has it changed at all, and can you share some momentum?
Kush Bavaria
>> Yeah, I'd say the mission is still the same. We still want to turn compute into a tradable compete class. I think what's changed the team is we've experienced hyper growth. The team is expanding. We're constantly hiring and we've even moved to offices twice already as a company. And so I think it's been incredible to see how the market is sort of shaping into what we believe and what we've thought is like training compute as an asset class, and how the market has sort of proved us right and our thinking from that. And I'd say that, look, today we're growing. The company's growing. I can't share the financials. Soon, but-
John Furrier
>> You got a big smile. I think this is the funding in your future. But we'll wait on that, but I want to get into the momentum and what that means for the market. And like I was saying on the intro, there is huge demand on the supply side, the build outs are clear. You see the CapEx. And even I was interviewing the Caruso CEO, he said, "We need more money." And everyone already thinks there's too much money in there, but he's not wrong either. There's a massive value proposition for the value that's being created. And Jensen, in his GTC keynote, showed the Pareto chart with the new Vera Rubin, which is on the higher end, obviously the delays and greatest, but you're starting to see a portfolio of the classic product positioning, entry level, mid-range, high end, and that's translating into portfolio thinking. Okay, don't query the AI for what's the weather report in New York City today on the Vera Rubin. Use that for other things. And so this is like policy. This is tech concepts. Maximize resource, link, load, schedule. That's like an operating system thinking around this whole concept of this utility.
Kush Bavaria
>> Yeah. I think those words you just said apply to electricity market itself, those words of like, "Okay, how do I schedule? How do I hedge this demand?" it's very similar to we look at the same way as we look at electricity. I think also another thing that people like to mention is, look, a lot of people think that the most constraining factor today is powered land or compute, but in our world, I think finance of these data centers is just as much of a constraint as other resources are. If you don't have the money, if you don't have the tools to get a loan, to get financing for your data center project, there's no way that's going to be built. And I completely agree with what Chase said before. Chase is incredible from Crusoe as well.
John Furrier
>> Yeah. So the farming example comes back now. So futures that actually enable that market, we don't know what the GPU market's going to be in five years, but yet these contracts on the data centers are being locked in from longer terms, but yet the buyer's going into probably two, three, maybe five max contracts and they probably don't even know what their capacity needs are going to be in five years.
Kush Bavaria
>> Yeah, of course. I think what happens today is there's a lot of different contracts that get rolled out. People build a data center assuming the life of it will be 10, 20 years. And that's sort of like shaped out in the past. But today the rate of accelerate is 100X, 10X every year. Every time NVIDIA releases a new chip, you just look at the frontier as you saw, it is 10X, 100X better than the previous generation that token input, and the models are only getting better and better and more efficient to run on the past generations strip. So there's a lot of uncertainty in the market. And when there's uncertainty, there needs to be created a market so people can express their own opinions onto that space.
John Furrier
>> Well, certainly in a build out mode. I want to ask how you got into this. MIT, I was kind of talking, kidding, but in a good way, MIT pedigree. But MIT, world-class organization, great computer science, great physics, math program. All great skills for this. Are you guys sitting around drinking beer at the Muddy saying, "Hey, let's attack the commodities market." Take me through the mindset because this is kind of a trend we're seeing. We're seeing a lot more brains going into solving hard problems that actually have commercial societal impact and isn't your classic build the new app.
Kush Bavaria
>> Yeah.
John Furrier
>> It's a whole nother level. Explain the origination, how it all came together.
Kush Bavaria
>> Of course. So my co-founder Wayne and I started the company back in August of last year. Wayne was a quant trader before this and I was working in semiconductors and venture before this. And so it's sort of a mix between the two worlds. And as you mentioned, I think MIT does a great job of teaching you about very different disciplines. And for us, we wanted to do something that has a real impact on the world and can help people in all kinds of spaces. And I think giving access to compute is just as important as doing anything else because without compute, we can't treat intelligence as an asset class, right? Intelligence is going to be the most important thing that people have access to. It's a tool that gives people a way to move up social ladders. It's a way for people to sort of create businesses off of. And I think intelligence, the whole AI boom is going to bring the world into an abundant state. And so it's helping sure that we do our part accelerating that.
John Furrier
>> And the electricity example is a good one because if you look at what's going on with AI infrastructure, it is truly industrializing kind of the new way to do things, create jobs, solve some societal problems. And so this clearly is an economic force. Again, back to monetization, take us through some of the things you're working on now. How does it work? Where is the beachhead that you're securing in your business? Is it financial? Is it technical? Take us through some of the mechanics of how everything's working.
Kush Bavaria
>> So today the product looks like we have indices that track the price of compute. These indices will track the price of NVIDIA chips to be more specific and they're structured the same way other indices are in the ... If you look at the oil markets or natural gas markets, right? We do the same sort of approach when creating our indices. These indices are then tradable on some of the CFTC regulated markets, such as . But we also do our own structured financial trades on top of these indices ourselves. I think the goal for us is to build a whole host of an exchange. Today it's very financial, but we're also thinking of making it more and more technical. If you think about electricity markets, it's the same way. When you trade on a PGM sort of grid in electricity markets, right? You trade on nodes. And for us, it's the same concept of applying that. These are markets that existed before. Compute is just a new commodity. The market structure should look relatively the same, at least and apply the concept of compute into that market.
John Furrier
>> How are you guys thinking about the system side of it? Because all these AI factory conversations all point to one thing and it's well documented. Every keynote, every vendor, mainly NVIDIA and all the chip vendors. It's not just the GPU or computer XPU. There's memory involved, HBM. You saw the prices of memory are skyrocketing. Take me through the mindset of how you guys factor in the subsystems around these large scale factories.
Kush Bavaria
>> Of course. I think there'll be markets that exist for every single component. I think as you mentioned before, HPM memory prices today are increasing at 100X, 300% from what previous years have been. And so it's incredible to see that all sorts of parts of this whole data center economy have sort of increased in price, decreased in price, and I think everything itself is a asset and if you look at it, right? And I think another thing that we'd like to talk about is sort of the whole economy in general. There's a stack of infrastructure that a lot of people I think miss because it looks more of like construction work and it looks like, "Oh, why do I care about this?" But I think the whole AI boom and all these application companies are supported by the infrastructure layer, which in our opinion is the largest sort of part of the stack.
John Furrier
>> What are you guys seeing in terms of the engineering build out? Give your opinion on the progress, where are the hotspots, what's working, where's the need?
Kush Bavaria
>> Yeah. I think a lot of the AI application layer and the model companies have done an incredible job where we use most of the models today internally to help develop our product, and I think it's just incredible to see the rate of progress that's happened. I think in previous generations and previous tech cycles, the rate of progress is very slow. You look at the 2000s and internet boom, and most of the apps only came out after 2003, 2005, 2006, but AI started in, you can say 2023 was when the ChatGPT-4, ChatGPT-3 moment came out. And after that, it's only been three years, and just look at the incredible rate of progress that's happened in the world.
John Furrier
>> And they got the security models coming out. It's a lot of chaos, but as Andy Grove once said, let chaos reign, then reign in the chaos. Clearly is a growth curve. Kush, great to have you on this AI Factory series. Appreciate what you do. Final question, what's some of the things you're working on now? What are you optimizing for? What are your goals?
Kush Bavaria
>> Look, I think for now we're looking at building out the exchange for compute. It's still the same mission that we started with and we want to help unlock intelligence for everyone. We're actively hiring. So if there's anyone that's watching that wants to join the team-
John Furrier
>> At which position? Put a plug in.
Kush Bavaria
>> The whole team. We are hiring across every single position across the whole team. We're trying to expand and-
John Furrier
>> Go to market.
Kush Bavaria
>> Go to market, engineering, sales, the whole stack. And I think it's important that the company's growing at a rapid pace and we just need to keep up with that. As I said before, agents can only do some of the job today and hopefully they'll do a lot more, but I think the world is still very people focused.
John Furrier
>> How are you guys thinking about AI native and ages, just while I got you here?
Kush Bavaria
>> We use every single probably tool that you've thought of and we sort of benchmark. Look, we're an early stage startup. We have the luxury of picking between the best product that exists in the market. So the tools that we use are what works best for us. We're not locked into a vendor. And so it's incredible for us to see the rate of development and we have absolutely no switching costs. I can tell the team that, "Hey, we're switching from OpenAI to Anthropic tomorrow." And the whole team can switch very easily, so we will always use best in the class and we're much faster than any incumbent in the space.
John Furrier
>> All right. Well, good luck. We'll keep track of it. Thanks for coming on and sharing. And again, love the mission. Appreciate it.
Kush Bavaria
>> Of course. Thank you for having me.
John Furrier
>> All right. I'm John Fury, host of The Cube here at the NYC Wired program. It's a Cube original, of course. The NYC Wired Network, it's open to everyone where the leaders are coming in sharing what they're working on in AI factories. It's the hottest area. As the industrial base of AI needs that core infrastructure, as that continues to thunder along on the build out, that will enable more agents, more tokens, more change for how we live, work, and play. We're doing our part to bring you the coverage here in The Cube. Thanks for watching.
John Furrier