Kristin Myers: Now, every headline about artificial intelligence is about the same five companies and the billions that they're spending on data centers. But not every data center is being built by a trillion-dollar company.
In fact, there is a whole tier below that: smaller operators building smaller facilities closer to where the data actually gets used. They have real assets and real contracts, but what they don't have is a lender.
Michael Abbate and Charles Buck are co-founders at GreenWulf Asset Management, and that gap is their business.
And then coming up later, we have Joe Benarroch of the New York Stock Exchange here to talk to us about how founders tell their stories right here in this building.
But first, let's hear a bit more on GreenWulf Asset Management. The road from an idea to the opening bell starts right now.
Everyone is talking about the massive data centers powering the AI boom, but Michael Abbate sees an opportunity beyond the hyperscalers.
Abbate is co-founder and Chief Investment Officer of GreenWulf, an asset management firm focused on opportunities being created by the AI infrastructure buildout.
And its pitch is simple: smaller and emerging data center operators need capital too.
So GreenWulf is targeting companies that may not have investment-grade credit and can't always access financing on the same terms as the biggest technology companies.
The firm is reportedly targeting $1 billion for an asset-based lending fund designed to finance capital-intensive AI infrastructure projects.
But GreenWulf isn't only betting on growth. Its investment strategies are also designed to find market dislocations created by disruptive technologies like AI, looking for opportunities as the industry evolves through different market cycles.
And Michael brings experience across traditional finance and digital assets.
Now, as billions pour into AI infrastructure, GreenWulf is making a different bet. Can financing and investing in the disruption surrounding the AI boom become an opportunity of its own?
And Michael Abbate and Charles Buck, co-founders of GreenWulf Asset Management, are here with us at the desk.
Mike, Chuck, thank you so much for being here with us today.
So I want to start with perhaps a little bit of some background for folks that might not know. What is an edge data center, if you can connect some of those dots for everyone? And how is that different from the ones that, you know, Amazon, for example, are building?
Michael Abbate: Sure. So both Chuck and I have experience in the credit markets. And one thing that we observed as soon as we started participating in the AI data center build is that there is a plethora of capital on one end of the spectrum, which were for the venture capitalists, the startup companies, where that type of capital was focused more towards, you know, growth opportunities and demanded a huge return on that capital for the risk that they were taking.
And then, obviously, there's a tremendous amount of capital that's being earmarked for projects that are backed by what we call investment-grade companies, high-quality issuers.
And there was really nothing in between. And so we started to focus on that market.
And more importantly, when we looked at those companies that were looking for the capital, we realized that they had assets that you can, as a lender, underwrite to.
And the biggest issue in terms of the understanding of the market was people's inability to underwrite to the duration of these assets because they are mostly GPUs. And it's a new industry, a new market, new supercomputers. People don't really know how long, you know, they'll last for.
So they are immediately just dismissive. Traditional credit investors are dismissive of assigning any value to them. And that's why the capital skewed towards the bigger, larger, higher-credit-quality companies.
And that's left the smaller companies in a little bit of a lurch, having to go back to their venture backers for the expensive capital.
And so we see a huge opportunity, really, to invest based upon the asset, based upon the computer, based upon the GPU itself.
We think we're particularly super well experienced and able to do that, and providing a need to the market that doesn't really exist right now.
Kristin Myers: So can you talk to us a little bit? Obviously, you're saying why the traditional lenders are really heading more for the Amazons and the Microsofts. And obviously, there's a lot of opportunity in that sort of gap.
How big is that gap, though? Right? What is left over after the really big companies are spending so much money on some of these buildouts? How big is that opportunity for some of these smaller guys?
Michael Abbate: Well, the total market started off as estimating about $3 trillion of AI data center spend. That number just keeps on getting bigger and bigger, right?
So, you know, now we've seen reports of $6 trillion, $7 trillion in intensity, even mentioned $10 trillion on one of its conference calls.
But, you know, we're just focused on, if we literally just take the publicly traded companies who are not investment-grade rated, go through their financials and understand exactly what their funding needs are, we estimate that's $400 billion to $500 billion in higher issuance just from the publicly traded companies alone.
Charles Buck: But it's the private marketplace, it's the sub-investment-grade tranche of this ecosystem that's dramatically underserved.
So where we focus, again, it's on the GPU itself. And one of the reasons people are apprehensive to lend against this is because there's a misunderstanding about the useful life of a GPU.
And sort of the conventional understanding or thought is that it's somewhere between four and six years. We have insight and belief that it's significantly longer.
And so that's where we focus for our loans.
And one of the reasons that we take comfort in that is people think of NVIDIA as just, it's a hardware company, but it's a software company. And that software that sort of enabled modern AI is CUDA.
CUDA itself is very powerful and can actually prolong the life of these GPUs.
So when we hear four to six years, well, there are GPUs six years old, the Amperes, three generations ago, that by many measures are working better now than they did at launch.
So until we see someone take an Ampere and throw it in the trash, we believe most are wrong, right?
Michael Abbate: And I just want to point out, Chuck is very qualified to make that statement because his brother wrote CUDA.
Kristin Myers: Oh, wow.
Michael Abbate: So, yeah, there's some family insight, familial knowledge and insight that's happening there.
Charles Buck: There's some empirical data that shows that the power of CUDA and what it can do.
Basically, every major innovation in AI has had to use CUDA. And what's interesting about that, and why it's such a powerful moat, is the more you use it, the stronger it gets.
Kristin Myers: So obviously, it was never a small bet. It's always been a big bet, but growing bigger by the day, really taking advantage of an opportunity that it looks like some of these other lenders are really missing here.
So I have a question, though, around collateral. What does that actually look like?
If one of the operators fails, you know, after a year or three, what do you actually own?
Charles Buck: Well, there's a long line waiting for these GPUs.
If we make a loan to a business and something happens to their business model and we take possession, we will know, we'll have insight as to when this will happen and where these can be redeployed because there's such substantial demand.
Michael Abbate: And that's why it's really a category of specialty finance. It takes a special skill set to be able to draft a contract, understand your security, perfect that lien on the collateral, and then have the operational wherewithal to repurpose if you're in that unfortunate situation of having to foreclose.
Kristin Myers: So what you were talking about before is that interconnection queue, because there's so much, as you were mentioning, specialty finance. There's so many phrases and acronyms that I'm obviously becoming more and more familiar with every single day.
So that interconnection queue, as you were mentioning, there's so many people that are really in the market for all of these products.
So tell us about that interconnection queue and how long that is and how that fits.
Michael Abbate: So the interconnection queue is slightly different. That is reserved for the power industry, right?
And so we were talking about, you know, financing the actual silicon, the actual chips. Obviously, you need the power to power them.
But, you know, the interconnection queue is obviously super topical right now, given everything that's happened in ERCOT, in particular with their Batch Zero process.
But just to throw a couple of, you know, numbers out there, in Texas there was a 495-gigawatt queue.
To put that into perspective, the peak load in ERCOT in, I guess it was 2025, was 85 gigawatts, right?
So multiples, right, of people that were applying for power relative to the absolute size of the grid.
They went through a process, they weeded out a whole bunch of requests, but they're still down to 66 gigawatts of demand in ERCOT versus that 85, you know, peak load.
So the industry is going to have a tremendous amount of generation that they're going to need to bring on in order to service that demand so that when these chips get delivered and financed properly, there's a place to plug them in.
Kristin Myers: And you actually called that foundational constraint this energy problem.
Michael Abbate: Yes. That's the constraint in the industry right now.
Kristin Myers: How does that change, or does that change, the math for you guys? Because you guys are looking out and doing some of these lending deals.
Charles Buck: Well, we have backgrounds in power, so it's nice. It became super relevant again.
So you have to have insight as to what's coming down the pike, where regulation is going to be changing.
But there is so much demand for debt capital in the space. We're just going to focus on a high-grade pipeline, getting into the ecosystem itself, having the right relationships with the right high-growth companies.
Again, these were VC-backed companies that are now [unclear in original transcript], at the top tier of [unclear], where even some of the best companies that are growing, you know, 10 times year over year, still can't get financing.
And that's a great opportunity for us because of our comfort, knowledge around residual values and useful life and so on.
And these loans can be in the shorter term. But being one of the few debt providers in the space has given us a great opportunity.
Kristin Myers: Okay, so my producer is going to kill me, but I have to ask you this one last question very fast. Let's try to do this in 30 seconds.
So a lot of folks are talking about an AI bubble, I'm sure, as you've heard. But if demand flattens, what does that do for you guys going forward?
Charles Buck: It's more opportunity. So that's an overbuilt cycle that's happened before. That's the early 2000 IPP, okay?
Where, if you were an investor in infrastructure due to the overbuild, because banks were lending to all these infrastructure plans, you could capture gold-plated, brand-new power assets for massive discounts because they were available.
And then you hold them, and the overbuilding gets absorbed, and then they become usable again.
So if there is a distress cycle that's similar to that, that's something great for us.
Michael Abbate: It comes down to execution, right? We've spent a lot of our careers trying to become infrastructure investors and project financiers.
So both Chuck and I spent a good amount of our careers in project finance, and you have to have the experience to be able to navigate through that.
And, you know, those who can, that's an opportunity.
Kristin Myers: All right. Mike, Chuck, from GreenWulf Asset Management, thank you so much.
Michael Abbate: Thanks for having us.
