In this segment from theCUBE + NYSE Wired’s “AI Factories – Data Centers of the Future” series, theCUBE’s Dave Vellante sits down with Rob Biederman, managing partner at Asymmetric Capital, to unpack a disciplined approach to early-stage investing amid AI-scale infrastructure shifts. Biederman explains Asymmetric’s founder-first model: writing $1–$10M checks (often via SAFEs), joining boards as they form and helping operators with go-to-market, operations, finance and strategy (not product/engineering). He shares why the firm avoided 2021’s lofty SaaS multiples in favor of backing proven builders earlier (single-digit pre-money), and highlights portfolio execution such as a cash-efficient LATAM e-commerce company scaling from ~$1-2M to about $50M in revenue. The discussion also explores Asymmetric’s subscale buy-and-build plays (e.g., pool cleaning in San Diego, sleep apnea clinics in Houston), where density, tech-enabled services and platform ops expand margins and enterprise value.
Biederman weighs in on AI economics as enterprises race to “AI factories,” cautioning that not every AI workload creates ROI and that overbuilt compute assumptions could face a reckoning. He argues that winners will prove a clear 10× value equation and avoid scaling go-to-market before product-market fit. Additional insights include early liquidity discipline (returning $0.20 on the dollar before the fund’s third anniversary), portfolio survivability (34 of 35 companies still operating; three positive exits), and guidance to founders: make your value proposition relevant, credible and differentiated. Tune in for candid perspective on how capital efficiency, ownership discipline and anti-thematic sourcing intersect with a world where GPU-dense data centers and AI-scale software are reshaping enterprise infrastructure and economics.
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.
Are you sure you want to remove access rights for this user?
Details
Manage Access
email address
Community Invitation
Sami Issa, Global AI
In this segment from theCUBE + NYSE Wired’s “AI Factories – Data Centers of the Future” series, theCUBE’s Dave Vellante sits down with Rob Biederman, managing partner at Asymmetric Capital, to unpack a disciplined approach to early-stage investing amid AI-scale infrastructure shifts. Biederman explains Asymmetric’s founder-first model: writing $1–$10M checks (often via SAFEs), joining boards as they form and helping operators with go-to-market, operations, finance and strategy (not product/engineering). He shares why the firm avoided 2021’s lofty SaaS multiples in favor of backing proven builders earlier (single-digit pre-money), and highlights portfolio execution such as a cash-efficient LATAM e-commerce company scaling from ~$1-2M to about $50M in revenue. The discussion also explores Asymmetric’s subscale buy-and-build plays (e.g., pool cleaning in San Diego, sleep apnea clinics in Houston), where density, tech-enabled services and platform ops expand margins and enterprise value.
Biederman weighs in on AI economics as enterprises race to “AI factories,” cautioning that not every AI workload creates ROI and that overbuilt compute assumptions could face a reckoning. He argues that winners will prove a clear 10× value equation and avoid scaling go-to-market before product-market fit. Additional insights include early liquidity discipline (returning $0.20 on the dollar before the fund’s third anniversary), portfolio survivability (34 of 35 companies still operating; three positive exits), and guidance to founders: make your value proposition relevant, credible and differentiated. Tune in for candid perspective on how capital efficiency, ownership discipline and anti-thematic sourcing intersect with a world where GPU-dense data centers and AI-scale software are reshaping enterprise infrastructure and economics.
In this interview from theCUBE + NYSE Wired: AI Factories - Data Centers of the Future event, Sami Issa, founder and chief executive officer of GlobalAI, joins theCUBE’s John Furrier to discuss the shift toward sovereign AI infrastructure. Issa explores the mission of Global AI in building dedicated, air-gapped environments that prioritize data security over traditional public cloud models. He emphasizes the critical need for enterprises to protect their "secret sauce," noting that because Large Language Models cannot "unlearn" ingested data, dedicated infras...Read more
exploreKeep Exploring
What is the mission of Global AI, and why was the company founded?add
What problem does your company solve for enterprises deploying LLMs, and why do they need dedicated, air-gapped AI infrastructure (rather than using their own data centers or shared cloud)?add
Does your company have good supply chain access?add
Do you see these centers as hubs or independent units, and will they be networked together?add
>> Welcome back. I'm John Furrier, host of theCUBE here at our NYSE CUBE Studios. Of course, we have our Palo Alto Studio connecting Silicon Valley and Wall Street. Technology is the market and you're seeing all the stocks, you're seeing all the capital markets. And we're here as part of our NYSE Wired Program, a CUBE original, AI Factory series. The technology is enabling all kinds of new things and the investors are confused, but as leadership continues to accelerate, we'll start to see visibility, agents, physical AI is coming. Sami Issa is here, Founder and CEO of Global AI. They're building the infrastructure enablement, accelerating AI. Sami, great to see you again.
Sami Issa
>> Pleasure, John.
John Furrier
>> Thanks for coming in.
Sami Issa
>> Thank you.
John Furrier
>> Okay, you guys, you're the Founder. So first of all, explain the founding and the mission for Global AI.
Sami Issa
>> Oh, absolutely. Dr. John Kelly, my partner, my Co-founder, and myself, a couple of years ago, realized that sovereignty is going to be necessary for generative AI growth, and that enterprises and nations are going to demand compartment of compute and infrastructure for them, dedicated for them. So we said, okay, this is a strong enough reason to start a company. We both bring in massive amount of expertise in the space, and we started Global AI, and that's what we do. We build sovereign, dedicated, not a cloud, AI infrastructure for the enterprise and for nations.
John Furrier
>> I want to get into the whole sovereignty thing. We've been saying for about a year and a half that sovereign AI and sovereign cloud are merging. This demand for different infrastructure, private and public clouds are merging into a programmable hybrid cloud, which is basically private cloud, if you want it.
Sami Issa
>> Yeah, I mean look, I think-
John Furrier
>> I mean, it's but if you're in the distributed computing network, you can slice and dice share this hybrid, private... yeah.
Sami Issa
>> Yeah, no, it's probably confusing a bit. So I think, look, I mean, people, we all have witnessed the capabilities of LLM. We all have seen what LLM did to coding. The LLMs had enough examples in a public domain to learn coding and master it, and now the business model of SaaS companies and coding-dependent companies is at question. So, that's a clear example of why you need your secret sauce to be protected. You need your secret sauce to be air-gapped, whether you're an enterprise or a nation. You cannot have it in a public domain, you cannot have it on a cloud. You have to be incredibly paranoid about keeping that information for you only, because one thing is unique about the LLMs is that they cannot unlearn. Once they have seen your secret sauce, it's embedded within their weights and it's near impossible to extract. And we envisioned that a couple of ago. We started the company to cater to fill that gap of creating these infrastructure dedicated for our customers, and that's what we do. It's single tenet, we're not a cloud, again. I keep telling everybody, we're not a cloud.
John Furrier
>> Yeah. Yeah, Sami, I love this conversation. It's multiple of my buttons, it pushes a couple of buttons. One is-
Sami Issa
>> I know you're a techie, John.
John Furrier
>> IBM and NVIDIA have a lot of similarities. Okay, you mentioned you guys have a deep relationship in NVIDIA, I want to get to that. But a lot of people think the data center growth is just an extension, but we've been trying to dissect from NVIDIA that they're building super computing factories.
Sami Issa
>> Yes.
John Furrier
>> This isn't just rack.
Sami Issa
>> No, no, no.
John Furrier
>> . It's completely designed. Okay, not a bolt-on.
Sami Issa
>> No.
John Furrier
>> They prove that interconnects and networking, their OS, I mean, we love the factory.
Sami Issa
>> You are spot on, John.
John Furrier
>> IBM has roots and big mainframes. There was a term that used to be kicked around in the '80s and '90s called Big Iron. Big Iron was a term that was for the mainframe because it was big iron in a glass house, data processing center, the information technology departments. So, you have big iron, that's a super computing mentality. So if you take the IBM DNA and if you see where I'm going with this, and NVIDIA. NVIDIA's essentially Big Iron. I mean, they're monster systems, they're super computing centers.
Sami Issa
>> That's brilliant insight. I wish John was here, he would have got a kick with that question, but you're spot on. I mean, 95% of data centers in the world are running on eight to 12, maybe 40 kilowatt racks. We are building a 150, 250, going to a megawatt per rack. These are super computers. Each GB-300 is an exaflop. A couple of years ago, you and I would dream of an exaflop machine, right? Now we have many, many exaflops within our data centers because each rack is an exaflop. So, you're absolutely right, these are super computers, super computers for AI, they demand enormous amount of power. And the beauty of... Who was the best company that sold super computers to enterprise or mainframes to an enterprise? IBM. My co-founder would agree with this. A lot of our team come from IBM. IBM did a phenomenal job in the mainframe space and I think NVIDIA is doing a phenomenal job in AI space.
John Furrier
>> And so I brought up private cloud and hybrid cloud earlier because I wanted to get to a point there. When you add intelligence, so AI factories in these super computer centers, there's an implication. There's an assumption is that, I think we all agree, that they're an intelligence factory-
Sami Issa
>> Correct....
John Furrier
>> with tokens. Things happen, software gets better, systems work better, systems and software working together. So when you have that kind of intelligence, you have the network up and down the stack, doesn't AI factories eliminate the need to call things private, public, on-prem? I mean, you guys are essentially not cloud, but what does that even mean? You're a distributed computing node in a hybrid network.
Sami Issa
>> Actually-
John Furrier
>> So, if factories have all the intelligence, there's no distinction-
Sami Issa
>> Well-...
John Furrier
>> where workloads are, or is there? How do you clear that up?
Sami Issa
>> Yeah, it's a great question. So okay, everything we see today, everything we've seen from Anthropics, OpenAI, et cetera, is based on LLMs digesting or ingesting and digesting publicly available data. But the value is really in the enterprise-protected data, right? This is where your secret sauce is. Again, we've seen the impact on coding. There was enough examples of coding in a public domain. LLMs mastered coding in a two year, right? And now they are in essence disrupting the business model for SaaS companies. You can argue, well, this disruption is going to be big, smaller, medium, we will find out, but there is no debate that they're disrupting it. So now, this is different than cloud. This is intelligence in the cloud. And do you want your secret sauce to be LLM'd? Because in the cloud you have the right to be forgotten. In the LLM world, you do not have the right to be forgotten. You cannot unlearn, you cannot convince an LLM to forget the learning, it's near impossible. So, we started that company with that key insight, that enterprises are going to realize that you want to protect your secret sauce way more than you did before, and you are going to demand that compute to create the creativity and productivity enhancement. You cannot have it in your own data centers because it's really complex. So, we will build it and dedicate it for you and air gab it from anybody else. And we have customers who demand to be the only customer in the whole data center. They say, "We don't want anybody else small, bigger or medium. We want the entire data center for us only." Not just the data hall, not just the compute, the entire data center.
John Furrier
>> You remember the old days you had intranets, extranets, DMZ, demilitarized zone, which is a network nomenclature for like an area where you have to play both sides? You apply that to the whole cloud notion, it goes away because what you guys are doing is essentially building AI infrastructure for enterprises-
Sami Issa
>> Correct....
John Furrier
>> to treat as a node or an extension of their premise.
Sami Issa
>> Exactly.
John Furrier
>> Whether they're in the cloud doing data processing, you are a trusted node in their network-
Sami Issa
>> Exactly....
John Furrier
>> as a factory node, and NVIDIA's powering that.
Sami Issa
>> 1005, Perfect, perfect summary. We are your compute that you couldn't have on your prem, because for the reasons that we just described. It's too complex, it's too expensive. And by the way, you don't have to cap ex it. We cap ex it for you and it's op ex to you, which is, CFOs love that.
John Furrier
>> Yeah. And so, now I want to get into what this means for the customer. So, let's just take through a scenario. I've been writing and commenting about the mode, strategic modes. In fact, just IBM stocks dropped because someone wrote a post that said, "Anthropic can do COBOL," which yeah, I can do COBOL, but I can't build Z, Z frame. So, you have this unlock, what are some of the use cases, because agents now are end-to-end workflows? So, I want you to talk about the importance of having this private supercomputing, whatever you call it, AI infrastructure, because when you start getting into agents that has the whole intra, DMZ, extranet issue of cross-domain traversal. So, if I'm writing agents for an enterprise, so that's happening. And then the unlock of how much of that private data has not yet been AI-ified, meaning-
Sami Issa
>> Yes, not been LLM'd yet.
John Furrier
>> It hasn't been trained. So, how are enterprises thinking about their architecture, knowing that agents are going to be going all crazy, doing good things?
Sami Issa
>> Yeah, I mean-
John Furrier
>> Data's going to be unlocked on-prem in a secure environment.
Sami Issa
>> I mean, look, your tech background is refreshing because you ask deep questions and I like that. I think, look, I think what we are hoping to provide to our customers is the safe environment where they can build, train on their secret sauce, and create their agents, their tools, their flows, that benefits from the wisdom that they have embedded in their own private networks for decades. It is difficult for me to imagine that a drug discovery company is going to want its data in any of the public LLMs.
John Furrier
>> No, that's clear. It's private LLMs for them.
Sami Issa
>> But then every, every enterprise has some secret sauce that is the shareholder value creation is built upon. None of them, drug discovery is an obvious example, but I think none of the enterprises with secret sauce want to have their secret sauce absorbed in an LLM. So, what do they do? What do they do? They cannot build it themselves because it's too expensive, too complex.
John Furrier
>> And time-consuming.
Sami Issa
>> Very time-consuming. And by the way, Jensen is doing a refresh every 12 to 18 months, which is insane to keep up with. I mean, we have the GP-200s and GP-300s running. Vera Rubin is coming up in April, and we are running like mad to keep up.
John Furrier
>> Talk about the relationship with NVIDIA, I think this is super important because what they're investing in, it's very clear they're not holding any cards back. They're like, play it all out there. The enterprise is huge for them, and that's new for NVIDIA. They never really had that. They had the gaming, they had Bitcoin mining, they had that supply and ecosystem outside, but now you have enterprise AI, whole nother ballgame for NVIDIA as they think about the enterprise. Of course, why wouldn't they? They're going to knock it down, NIMs, NeMo, all good stuff, we covered that. Omniverse, great for training physical AI. So clearly, NVIDIA gets it, but they're kind of new to the enterprise. How is your relationship with NVIDIA? Are you an early partner? Are you co-designing with them? Radical co-designing, as they say?
Sami Issa
>> We are very, very close to NVIDIA. They have been a fantastic partner to us. As I mentioned, we can because we are technically strong, we have a very strong technical team. We can build these liquid-cooled data centers ourselves. We can bring up the bleeding-edge technology online as I've mentioned, GB-200s and 300s.
John Furrier
>> So, you have good supply chain access?
Sami Issa
>> We have the supply chain access, but also more important, we have the access to equally important, the access of the brain trust within NVIDIA where we can, because these machines are complex. They're unlike anything that I or you have seen before. Like you mentioned earlier-
John Furrier
>> Large scale....
Sami Issa
>> they're super computers in a rack, so they're not easy to bring up. You need very, very qualified people to spend enormous amount of time to bring these machines up. And without NVIDIA's brain trust helping us and Supermicro, our partner, we cannot bring them. I was talking-
John Furrier
>> And you guys are building the physical plant, too? It's not just you're dropping it in a NVIDIA box.
Sami Issa
>> No, no, the whole thing.
John Furrier
>> You're engineering the facility, the power.
Sami Issa
>> We fit it out. We put the pumps in place, water pressure in place, electrical. Everything from the ground, vertically integrated. I was talking to an investor a couple weeks ago and he said there are 300 individuals in the world, 300, in the world who can actually build at scale GB-300s today. And that is the best, the most active description of the market today. This is not 3,000, 300 across planet earth.
John Furrier
>> All right, talk about the demand side from your business. You guys got great momentum. Obviously, you got a product-market fit, I'm sure there's demand for that. Private, on-prem AI, extended on-prem, it's an extension so it's basically on-prem in the category.
Sami Issa
>> Exactly, exactly right.
John Furrier
>> Talk about the growth. What's the demand look like? Just share some anecdotal and then other approaches. Some people are taking the whole, "We're a neocloud, we want GPUs here," and they're basically white labeling data centers. What's the difference in this market?
Sami Issa
>> I mean, look, demand is unlike anything I've seen in my entire career. Demand is unbelievable. If I go to the market with 500 megawatt, it'll be gone within a couple of days. If I go to the market with a gigawatt, it will be gone within a week. And I'm not talking about-
John Furrier
>> I mean, the hyperscale is buying too, everyone. Meta, Amazon.
Sami Issa
>> I mean, exactly. I'm not talking about speculation. I'm talking about 3, 5, 6 years contracts with investment-grade companies that need that demand for the growth and for customers. So demand is off the charts, anything and everything. We are running like crazy trying to bring as much supply as possible to the market. So, demand is off the charts. And I think-
John Furrier
>> So, you're building like crazy.
Sami Issa
>> We are hovering power across the nation and we're looking outside the-
John Furrier
>> Share the footprint and what your goals are and what your growth strategy is.
Sami Issa
>> I mean, look, we are looking at building about a gigawatt in the US by 2029. A couple of gigawatts here and outside. We have partnership in KSA, in the Kingdom of Saudi Arabia, with Humane.
John Furrier
>> You're building a global network.
Sami Issa
>> That's exactly.
John Furrier
>> Hey, Global AI.
Sami Issa
>> Global AI. Hence the name, hence the name.
John Furrier
>> But you see these centers as hubs or independent units. Are they networked together? I mean, imagine you're basically building private factories.
Sami Issa
>> Exactly. They are not networked together, they're dedicated. And even if we have... From experience, which is the case, if you have a customer that has a footprint here in data center A in location A and location B, they don't want them connected in the background. They're going to connect them at their hub. So, they're not connected yet.
John Furrier
>> So you guys are running hard, funding, feeling good about that?
Sami Issa
>> Funding is great.
John Furrier
>> Market talk.
Sami Issa
>> We went to market a couple of times last year. We raised money super quick, few weeks, combination of debt and equity. Debt is against contracted revenue so we don't do any speculation building. Everything has a contract, has a customer, has contracted revenue, everything's investment grade. So, it's a fantastic place to be.
John Furrier
>> One of the things that we have a vision on, we're tracking obviously the AI factory, central factories, distributed nodes, whatever node it is, big, small, medium, edge is coming and then metro, we see metro factories coming. So, when you have these connected intelligence devices, supercomputing, enterprises start to behave differently. They're essentially redesigning their network, so I, as an enterprise could say, "Hey, I got to have a global footprint, tap Global AI, and run my own workloads in a distributed computing fashion." That's pretty much where you guys see the puck going.
Sami Issa
>> Yeah, I mean, look, the end state, I think the end state is going to be, AI intelligence is going to be utility, like electricity. You're going to plug into the wall and give intelligence, right? Where it comes from, does the execution happen 50% on the edge and 50% outside of the edge, or 20/30? Depends on the workload, depends on the latency needs, depends on the compute need. The compute, it still takes enormous amount of time. I mean, if you look at six seconds, probably three quarters of that is compute, not latency. So, we're still not at the latency edge point yet for most big workloads that we have seen.
John Furrier
>> Well, Sami, you guys are doing a lot of hard things. One, identifying locations, building out these data centers/supercomputing centers, AI factories. For the people that don't know the complexity, scope what it's like not just to get real estate and plug in the power in the wall and spin it. Right? You guys are designing AI factory systems. So every piece of material, every piece of power you're squeezing out of it. Talk about the design that goes into the engineering of that. Just scope the order of magnitude for people who don't understand, why is this so hot? What's the big deal? Just get some power, plug in a rack. Talk about the difference between how hard it is.
Sami Issa
>> I wish it was that easy. I'll give you an example. We're building a gigawatt data center. Gigawatt is a big number, right? Now, and here's the scary part of this. These chips are burning so much energy and producing so much heat that we have to cool them with liquid. That liquid that has to reject that one gigawatt worth of heat, comes all the way to the chip. So, you have millimeters, less than millimeters between a gigawatt electricity and water. And make one mistake, and single mistake, and the end result is catastrophic, so.
John Furrier
>> The density, the heat. .
Sami Issa
>> Density, heat, the proximity of liquid and electricity is something that we've never seen.
John Furrier
>> Yeah. I mean, these are like substates, like electricity. These are highly engineered power-
Sami Issa
>> These are power plants, these are power plants.
John Furrier
>> Token power plants.
Sami Issa
>> Yeah. A gigawatt is a typical nuclear reactor output. Right? A nuclear reactor is in a 1.2, 1.4 gigawatt.
John Furrier
>> Yeah. Well, Sami, you got a great venture. Looking forward to meeting your co-founder from IBM and getting more deeper dive on that. For the last minute we have, put a plug in for the company. What are you looking to do? You're hiring? What are you are you looking to hire? Give a plug.
Sami Issa
>> We're hiring like crazy. We're looking for talent, entrepreneurs, intelligent people across the spectrum. We're growing. I mean, we're adding, I don't know, 20, 30, 40 people a month now. Building like crazy across the nation. We're looking outside the country. We're looking for power. If you have power, call me.
John Furrier
>> You're open for business.
Sami Issa
>> We are open for business.
John Furrier
>> All right, Sami, appreciate it. Global AI, again, this is the wave, it's a global infrastructure. Sovereignty is where people want to have their customization, get their privacy, private AI, super AI, it's all happening. Enterprises want their own infrastructure and it should be computing. We're doing our part in theCUBE, bringing that to you. I'm John Furrier, the host. Thanks for watching.
John Furrier