In this segment from theCUBE + NYSE Wired’s “AI Factories – Data Centers of the Future” series, Chris Stephens, field CTO at Groq, joins theCUBE’s John Furrier from the NYSE to unpack how AI factories are reshaping enterprise infrastructure and sovereign compute strategy. Furrier notes Groq’s momentum (~$750M raised, a valuation approaching ~$7B, and a new partnership with McLaren) as Stephens outlines why inference is the “killer app” and now a market of its own. He details rapid standups of sovereign AI inference “factories” (~51 days in the Middle East, ~40-some days with Bell Canada, and ~30 days in Helsinki) and explains how telcos are leveraging trust, data and national footprints to deliver AI-scale services.
The discussion explores where value is accruing across the stack: from the physical build-out to the software layer that operationalizes Groq’s LPU-based system. Stephens highlights GroqCloud (launched ~18 months ago), native MCP service support and a Compound-powered research product, all aimed at simplifying deployment and enabling secure, standards-driven agent communications. He also digs into real-world use cases (customer-facing agents and workflow automation), cross-site/sovereign interconnect considerations, and why “joules per token” is becoming a defining metric for scaling reliable, low-latency inference within power-constrained data center designs.
Forgot Password
Almost there!
We just sent you a verification email. Please verify your account to gain access to
theCUBE + NYSE Wired: AI Factories - Data Centers 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: AI Factories - Data Centers 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: AI Factories - Data Centers of the Future
Please sign in with LinkedIn to continue to theCUBE + NYSE Wired: AI Factories - Data Centers of the Future. Signing in with LinkedIn ensures a professional environment.
In this NYSE studio conversation for theCUBE + NYSE Wired: AI Factories - Data Centers of the Future, Argentum AI founder and CEO Andrew Sobko joins theCUBE’s John Furrier to cut through the noise on AI infrastructure and explain why the new bottleneck is no longer silicon but site-ready power and data center capacity. Sobko details how Fortune 500 AI teams and hyperscalers are racing to secure compute at scale – and why Argentum AI’s global marketplace is becoming a practical “easy button” for dialing up GPUs through diversified partners.
The discus...Read more
exploreKeep Exploring
What is the current state of demand and supply in the AI industry, particularly regarding compute resources?add
What is Argentum AI and what services does it provide?add
What are the primary demands and trends observed in the GPU market, particularly from large corporations and hyperscalers, and how has the company responded to these demands in terms of partnerships and infrastructure?add
What recent trends and insights have emerged regarding the importance of compute as an asset class in investment conversations?add
>> Welcome back here. I'm John Furrier, host of theCUBE at our NYSC studios here in New York City. This is part of theCUBE's Wired program. A CUBE Original, of course, we have our Palo Alto studio connecting Wall Street and Silicon Valley, tech and the capital markets. That's the focus of the program. This is our AI Factory series. We talked to the leaders. We're making it happen, building out the next generation infrastructure so AI can move to autonomous, fill in the agents, and then get the physical AI, of course. Cloud technology, the neo clouds are here. We got another great guest coming back. Andrew Sobko is here, he is the founder and CEO of Argentum AI. Good friend of theCUBE, has been on before doing some amazing things. Great to see you. Thanks for coming back.
Andrew Sobko
>> Thank you for having me again.
John Furrier
>> Let's talk about AI infrastructure. Obviously, the last time you were here, we saw the surge in demand. You see constraints. Amazon's looking for more horsepower. Everyone's looking for more capabilities. Enterprises are onboarding. The agent wave is hot. We just saw CES, NRF, a retail show, physical AI, a lot of agents. A lot of reality is hitting the world on the fact that there is value, but the demand is there. You're filling the void. Take us through what's up.
Andrew Sobko
>> Look, so first of all, AI demand is very real and people speculating what's happening in the industry, but we've seen it's very real. We've seen all serious teams AI teams in Fortune 500 companies are in the massive shortage of compute. And the winners are trying to secure compute at scale. And we've seen an interesting trend where people were very dependent on any single one cloud or centralized cloud and now they're kind of pivoting and trying to de-risk themselves and kind of partner up with multiple players like us.
John Furrier
>> Talk about Argentum AI, what you guys are doing, because the service you have is unique, it's in demand. Talk about the service.
Andrew Sobko
>> Yeah. So Argentum AI is a compute marketplace. We are a global aggregator. We've partnered up with hundreds of data centers globally and our kind of core focus to give and simplify access to compute from different startups to also large Fortune 500 companies, in many cases, even very large hyperscalers are using Argentum AI. Today, Argentum AI already has over 120,000 GPUs available on the platform for this year for long-term rentals, which gives us significant scale. And again, we are on the way to a million GPUs in the next year and a half.
John Furrier
>> So your customer is someone who wants infrastructure, capability, compute, GPUs. What's specifically the number one thing you're selling right now?
Andrew Sobko
>> Yeah. So number one, our biggest demand comes today is from some of the largest Fortune 500 companies and some large hyperscalers, simply because we have a lot of GPUs available on the platform today. And we've seen, again, from historically, we went from this interesting kind of point in the industry where people were talking about it's all about the chips, now it's all about the power. Who has the power? Who secure the power and where you can actually place chips and operate quickly. So what we did effectively over the last, I would say, six to nine months, we partnered up with a lot of actual infrastructure partners. We work with a lot of neoclouds globally where we bring some neoclouds, and of course, our job is to give access to actual compute, to GPUs, to bare metal, to some very large corporations.
John Furrier
>> So you're the easy button because we've been covering this and there's a learning curve for getting the infrastructure up and running. Certainly on the physical side, getting the systems, deploying them, operationalizing them, you're taking a different approach, just almost like dialing up GPUs. Is that right?
Andrew Sobko
>> We're kind of simplifying it. I like to say we're kind of building almost like a DoorDash experience where you can get compute on demand at scale in less than five to 10 seconds.
John Furrier
>> And what's the use case? People who are doing projects? I mean, I see the need like, okay, we got this thing, let's move it in and let's stand up some GPUs. Wait, I got to go buy them. Where do I procure? Which service do I use? You provide what? The simple path?
Andrew Sobko
>> So, your simple path, you come to us, let's say you're a scaled startup, we partnered up with a lot of startups that recently raised hundreds and hundreds of millions of dollars. You need to quickly stand up infrastructure and typically it takes, you got to go partner up with somebody and need a very strong balance sheet. We are fully taking that friction away from it and we can help stand out infrastructure again. We have a lot of stuff available now and we can help you get stuff at scale in the next, let's say 60 to 90 days. We have a particular client right now that's actually looking for 20,000 GPUs in Q1, which some large companies will tell you it's impossible to get right now and we stood up this kind of infrastructure with our partners, we give them access to 20,000 GPUs in Q1.
John Furrier
>> And first of all, explain that because I think that's the false narrative we're hearing, "Oh, it's short supply," mainly because people have been buying them. What's the real deal here? Where's the hype versus the reality?
Andrew Sobko
>> So first of all, AI is very real. Hype, I like to say it's like who gets paid for it, right? So that's where we start. So on the supply side, there's definitely massive supply shortages. We went from, again, from the chips, now chips are not an issue anymore, now the biggest issue is actually who has the power, who has the infrastructure, has data center capacity available today. That's the biggest shortage today. So what we were effectively able to do, we were, again, for the last many months, we were aggressively secure in some of that data center capacity for let's say Q1 and Q2, and now we're securing aggressively for Q3. So we went ahead of everyone else. We grabbed the capacity under some sort of exclusivity to us. We were able to stood up the infrastructure and GPUs, with some of our neocloud partners. Again, we are a marketplace. We're just a platform, but we had to go down into deployments and help our partners to stood up large infrastructure so we can simply give access to some of our largest clients.
John Furrier
>> So for people who want to experiment, look at the risk analysis, but don't want to lose momentum, you're a good path for that?
Andrew Sobko
>> They come to us and we recently had a lot of conversation with a lot of large investment groups and they tried to give their startups simple access to GPUs. In many cases, we can give access to your startups tomorrow. We can give them access to hundreds and hundreds of GPUs tomorrow.
John Furrier
>> So you have your finger on the pulse of the real demand. Supply side, notwithstanding, certainly there's a supply chain, but there's a real need. Can you comment on this AI bubble narrative? First of all, you know where I stand. I think there's no AI bubble, but I don't want to lead the witness, so to speak. What's your view on the AI bubble? What are people getting right? What are people getting wrong? Where are we?
Andrew Sobko
>> Yeah. So there is some speculation there could be some sort of AI bubble in maybe some valuations of some companies without any cloud and clear economics and kind of understanding what exactly they're doing, how they're going to simply maybe make money and what they're doing. But what I'm seeing, there's actual AI has been applied in actual production use cases at scale from fraud detection, customer service at scale. We are talking right now to a very large Fortune 500 company, which is not your typical AI native firm and they're looking to get up to 15 million GPUs in the next two years. So they're fully redesigning the entire company, which is one of the most-
John Furrier
>> This is enterprise....
Andrew Sobko
>> well known company enterprises, it's not your typical AI company, completely redesigning all of their stores globally and kind of refocusing again, customer service. I'm not going to say too many comments, but potentially trying to get from the customer service, payments, everything else replaced with AI.
John Furrier
>> So basically the enterprises are looking at AI native as the outcome that they want.
Andrew Sobko
>> Yes.
John Furrier
>> Is that fair?
Andrew Sobko
>> Yeah. Some people may partner up with some large language models companies, very effective, but again, they still need a lot of GPUs. They need a lot of infrastructure. So again, it all becomes around compute.
John Furrier
>> You mentioned valuations. First of all, I would agree, I think valuation is a little bit over the top, but I think there's a real deal here for sure. What are investors missing? I mean, a lot of people, I see the narrative on TV, on cable, certainly not here in theCUBE because I think we have a good call on this in my opinion. Happy to debate with anyone. I don't think there's an AI bubble. I think there was a little bit of a euphoria and enthusiasm, but definitely it's not like the internet bubble at all. But from an investor standpoint, there are a lot of money chasing deals. What are the investors missing and what are they getting?
Andrew Sobko
>> Yeah. So a few interesting data points here on the investor side, and we have in a lot of those daily, daily conversations. So again, we went from huge model, talent, shortage, right? Investors were stepping up and helping their companies to secure talent. Now people clearly realize it's all about the compute. Can you get reliable compute at scale? And interesting data point, just recently went to a couple conferences, we just came from Abu Dhabi Finance Week, and what I'm seeing specifically, it's coming from the Middle East and of course US, compute is becoming a new asset class. And our job, what we're trying to do, we're trying to make it very tradable. We're bringing liquidity to it and also kind of helping people kind of stand up some of those infrastructure and bring huge demand at scale immediately on demand. But the biggest ones, again, it's all about the compute and investors are finally realizing it. And some investors, again, were so focused again on AI, they were thinking about it's all just about the software, but it's no longer just about the software.
John Furrier
>> I mean, it's multifaceted, but I think you bring up a good point. It's not one trick pony, there's many ways to slice the revenue model. In this case, your marketplace is a service, you make money on that. This whole concept of liquidity and asset class is interesting too, because now you can play there too. Is that right?
Andrew Sobko
>> Yes, that's what we do. We help companies, we'll help, again, brand new data centers, brand new neoclouds, kind of bring immediate demand, which means we're bringing immediate liquidity to them.
John Furrier
>> So I have to ask this. You can answer it or you can try to answer it or pass or say no comment, but are investors banging your door down because you have a very asset light model, you hit the demand curve, you got optionality on the revenue side. What's the capital markets like for you?
Andrew Sobko
>> Oh, capital markets is very interesting. We having a lot of conversations with investors sort of all over us. I like to explain our business model very simply. We are basically asset light . We are fully decentralized. We can stand up. We have a lot of partners globally. We're focusing on those cross-border transactions. So we're getting a lot of investor demand coming from global players.
John Furrier
>> And you haven't really raised big monster rounds. You essentially have momentum and visibility to unit economics and you got the supply demand side down and not a lot of capital raise, right?
Andrew Sobko
>> Yes. So again, our business model is very asset light. We're a marketplace, but again, we're still focusing on the very aggressive go to-market strategies and global expansion and continued aggressive security and supply because it's all about the supply. There's definitely, again, supply shortage and we're sort of being ahead of everyone else. So we are in some late conversation with some major investment groups for some next round. We're having those conversations, but again, at the same time, our focus is continue solving problem for our largest clients, which we're doing and we're announcing some major deals then.
John Furrier
>> Andrew, you're a marketplace. First of all, everyone knows theCUBE, they know I'm a huge fan of marketplaces like AWS Marketplace. Google's got a marketplace. Databricks has a marketplace. Those are products, procurement like marketplaces. Simplifies the buying process, but you got that and you got the financial side of it nailed down. So you got buying and procuring the service and you got the financial side of it.
Andrew Sobko
>> Yeah. I truly believe, and I mentioned this a few times already, I truly believe when you bring into some sort of from idle GPUs to second life GPUs to a brand new GPUs, you bring liquidity, which means you bring a lot of demand and you sometimes use some of your balance sheet to make those transactions more effective, way faster. You make industry way better and you make it more accessible and you kind of simplify how people access to it. Today's startups are actually struggling getting access to compute and GPUs. It's a massive problem and we are simplifying it.
John Furrier
>> I mean, what they say, big TAM, good economics, profitability is looking in line with asset light. That's awesome. Congratulations. We'll see how that goes. But I got to ask you, you mentioned decentralized. Some people are looking at some of the architectural features of the players, centralized versus decentralized. What's the pros and cons? It could be a risk factor or it could be an opportunity. What's your take on this whole centralized risk factor?
Andrew Sobko
>> So I'm seeing the people finally realizing what recently happened with Cloudflare from AWS. If one data center goes down, the entire global ecosystem goes down. So banks, payments, everything goes down, the whole infrastructure. In our model, because we work with so many very stable partners, if one data center goes down, it becomes a local issue and we can redistribute data to a different location really quickly. That's how the system is designed. So I do believe that the future is in decentralization. Also, the only way to determine the real time price for GPU per hour is being fully decentralized is working with everyone, working with all neoclouds, with large data centers globally. That's how you determine the real price if you .
John Furrier
>> So you've got a couple of benefits there if I get this right. Correct me if I'm wrong. Hedging and hedging on the service, so price, availability, and then service, quality of service.
Andrew Sobko
>> Quality , yeah.
John Furrier
>> We just saw Verizon Mobile just crashed. I was talking to someone who builds edge devices with humanoids in cars and if a car's on Verizon, you got to have a failover, you can't rely on one provider on the network there. There's an obvious example that happened yesterday, but you're looking at it from more of, hey, if I'm cranking out AI outcomes, I want to have diverse optionality on service levels, speed, and you could program that from the customer.
Andrew Sobko
>> Exactly. Yeah. So we truly believe that the next cloud future, the biggest cloud will be at the marketplace. It's not going to be the typical centralized cloud that we know today, which was usually typically built on a 2010 infrastructure for your typical SaaS companies. And now with this huge demand in AI, they just cannot handle it.
John Furrier
>> So since we last talked, you were at Abu Dhabi all around doing meetings, customer growth, you guys got some good growth. What's the takeaway in the past few months with your business? What's the outlook? What's your perspective?
Andrew Sobko
>> Yeah. So the people finally started realizing that compute and diversification of different providers. We've seen a lot of demand coming from our clients who historically were focusing on just one cloud, one partnership, now thinking about working with us because we have full diversified portfolio of suppliers and different, from data centers and neoclouds globally. That's number one. People are finally realizing the importance of decentralization, again, as we just discussed from Verizon and everyone else. And our biggest advantage today, we were ahead of everyone else and we took some risks and we took some balance sheet risk and we secured tons of capacity for 2026 ahead of everyone else. We went out, and again, as I mentioned, we partnered out with about 100 neoclouds and data centers globally. We have over 120,000 GPUs available today, and that's our single biggest advantage today.
John Furrier
>> Well, the demand, people are hungry for it. Congratulations. What's on your agenda this year? What's the goals?
Andrew Sobko
>> So the goal is, of course, is to scale our supply side, hopefully to 500,000 GPUs. And in the next two years, I kept saying that hopefully we'll get to a million GPUs. So imagine in the marketplace we're having a million GPUs without actually owning single one on your balance sheet.
John Furrier
>> Well, we appreciate what you do. The AI infrastructure certainly is the gating factor for how fast agents will come on the scene. And then ultimately behind agentic is the physical AI. That's the edge. It's all going to be in demand for capabilities. It's like electricity. You're powering everything.
Andrew Sobko
>> That's exactly what's becoming.
John Furrier
>> Congratulations. Thanks for coming on theCUBE.
Andrew Sobko
>> Thanks for having me.
John Furrier
>> I'm John Furrier here at theCUBE. AI factories and our AI infrastructure series, these leaders are making it happen. The AI infrastructure has to continue to thunder away and fill those gaps and provide the power, the supercomputing power to power those data factories, the AI factories, the edge, and everything in between. This is a 10-year cycle that is going to power agent technology, agent infrastructure, I call the middleware, and of course, as AI physical comes in, autonomous vehicles, autonomous retail, autonomous healthcare, everything's going AI and autonomous. Of course, without the horsepower, it won't happen. So keep an eye out for this area. Super hot. Thanks for watching.