CoreWeave closes $8.5B investment-grade debt facility at SOFR+225bps, its largest ever

coming on uh trolling the haters. Keep building. Keep shipping those lines of code and we will talk to you soon.

Have a good one. Goodbye.

Catch you.

Let me tell you about Plaid. Cloud powers the apps you use to spend, save, borrow, and invest. Securely connecting bank accounts to move money, fight fraud, and improve lending now with AI. And we are running over time. But let's bring in our next guest. Brandon from Core Wee is in the ream rating room. Sorry for keeping you so long. Thank you so much for taking the time to come talk to us on TBPN. How are you doing?

I'm doing great. Thank you for having me on today.

Of course. And uh could you uh introduce yourself and and kind of uh take us through some of the news? I mean, there's so much going on, but I know that there's some big announcements more recently.

Yeah, absolutely. My name is Brandon McB. I'm one of the co-founders of Core Weef. I'm the chief development officer. So, I lead all things debt, equity, M&A, ventures for the business. This is where all the money is raised within the organization.

Yeah. And and this new uh this new project, this new vehicle, how are you positioning it? How are you explaining to people who uh are familiar with coreweave as a stock that they can buy and they understand that you might use uh debt to finance uh you know the purchase of a of a facility. They're familiar with the mortgage. How are you introducing this product?

Yeah. So that this product is one that we brought in the market about 3 years ago. Okay. And it was with our first facility DDTL1. And what we've introduced today is the continuation of just continuing to build upon and execute within that original product.

We're effectively taking this GPU infrastructure along with revenue associated with the GPU and these long-term take or pay contracts,

put them together, and then take them out into the debt market and finance them. And I would say it's a financing strategy that has occurred across many different sectors, right? Like this is how LG facilities or power plants get financed, right? You have the infrastructure in one hand, you have these long-term contracts in the other hand, you put them together, marry them, and you can stand up very scalable financing mechanisms around it. Um, what's really important about the one that we announced yesterday, this is our largest transaction at 8.5 billion.

Um, and it's our first transaction being done. Thank you. Yes, it was awesome.

It's a big moment. um huge number, massive, you know, I sometimes like forget the numbers that we're working at sometimes. Um and it was done at an investment grade cost of capital. So this was done at soer plus 225 or you know approximately 5.9% cost capital which is incredibly important to us uh given the capital intensive nature of our business. And what what uh what gave the market what gave the rating agency or you the confidence to put this in the investment grade bucket? How are you uh how are you providing the data and the back and and the and like the backbone of that strength?

Yeah. To to your point, this all comes down to market validation. the the clearest signal out there is pricing and confidence grows as cost of capital comes down, right? Doesn't really work the the other way. So, being able to drive down cost of capital here just shows the amount of demand there is for the core weave credit product

in the market. How much of this is uh how how important was sort of beating the depreciation gate allegations to put it sort of silly but a lot of people were wondering like oh those those H100s they're not going to they're not going to age well and what we've seen is that there's a lot of data points out there that seems like they're aging like fine wine from what we've seen. Uh but take me through your your uh did you always know that that was going to play out the way? Was there any un uh unexpected surprise to the upside? And then how important has it been to have more data about how these GPUs monetize over a longer length of time?

The depreciation subject has always been very interesting to us. Um we've been very consistent. 6 years is the appreciable life of this infrastructure and I would say if anything we're expecting for the infrastructure to have useful life beyond six. I mean you you see that today and like late 2010 SKs that are available in market on cloud. like they're not running that infrastructure unprofitably,

right? And it means that there's client demand for it. But as we look towards like Ampier and Hopper generations, you know, I think in our last earnings, we even noted that our pricing on AER went up during 2025 and our pricing on Hopper stayed within 10% or so of where we started the year. Why is that happening? I I think it's largely driven by inference, right? And ultimately this concept that not every workload only needs the latest greatest piece of compute right which we're incredibly well known for running but you know there are different types of workloads that need different types of infrastructure and the market is simply efficiently pairing the right compute with uh with the workload that they need. So we see incredibly robust demand um all the way back to the Ampear generation which I believe is six years old at this point from its um original skew launch date and we're we're really not seeing any deterioration of these older skew demand profiles um while also seeing accelerating demand for latest generation infrastructure as well. Uh I think everyone's heard the stories of electricians being flown around in private jets and and kind of the chaos of of actually you know building out uh data centers over the last two years. Any any funny stories from from the core weave team recently around uh around all

Yeah. on on all the effort going into uh you know building out these clouds.

Yeah. It would take a lot of private planes to to fly all those electricians through. I mean, we're talking hundreds to thousands of engineers on these sites. And, you know, I think that's it's a great point you raised because I I believe it's an underappreciated aspect of what is being done here and the true scale of the engineering effort out there in market. Like, you truly just have to go see one of these sites, right? They are massive, massive engineering feats that are being executed on with an incredibly intense supply chains all the while moving in a extremely high velocity technology sector. Um it it it really boils down to rich uh risk management which is the DNA of our management team. Um it feels like in terms of inference demand we've seen a number of sort of exponential jumps when we went from uh LLM to reasoning models the number of tokens generated to answer a question uh increased uh exponentially saw something similar with agents coding agents uh are you guiding or or expecting for just a smooth trend to continue are you thinking that this is more discontinuous is there some sort of next next generation that you're already looking forward forward to in terms of increased inference demand.

It's it's really hard to predict

where the technology goes since the technology itself is meant to be so disruptive.

Yes. Um, I think what's super important for us is our client base and our clients represent the um the organization that are at the front lines of pushing this technology as far as it can go and we're consistently supporting them and it gives us this really interesting information flow understanding where they're going in their organization so that we can be orienting ours to be moving with them in lock step. What do people misunderstand most about coreweave right now?

Um, you know that I I would say there's three real pillars to our success. It is having the best product in the market for running this infrastructure. Um, our product is purpose-built for AI and supports today the leading AI organizations on the planet. um you have to have the right physical infrastructure capacity, the ability to navigate the supply chain, right? And that that is not something uh anyone can just stand up and do. It's incredibly challenging and detailed. But today we have over 43 data centers in active operation. Um that that's a key component of how we've been able to pull down our cost of capital so aggressively is just simply through execution. And that final piece is financing,

right? You can have pieces one and two, but if you can't raise the capital needed

to uh uh to bring that infrastructure online, it's it's sort of meaningless. And so we've spent an immense amount of time focused on the credit market and driving an ability to navigate it effectively and scalably while driving down our cost of capital. and thus why this instrument that we announced yesterday is so incredibly important to us and will serve as the blueprint for future transactions.

Are you in a rough agreement with the broad consensus that I'm feeling in the AI industry around uh a chip bottleneck being more of a gating factor uh factor in scaling AI progress over the next four years than power constraints? you know, I I would qualify it as cloud connected infrastructure. Incredibly important. So, it's it's really marrying power with chips, getting them online delivered to clients.

Yeah.

Um that that bottleneck within there is delivering power shell capacity. Um at least within our business, right? We can bring online uh uh more capacity to our clients as we have more powered shell within the market. So, we're consistently uh scouring the market for incremental capacity to be able to market to our consumers.

Interesting.

How uh how have the sort of mom and pop data centers done over the last two years? I'm talking about people maybe like 18 months ago that that were realizing, hey, this AI thing is going to be pretty big. We have we can find some power here. Let's throw up a shell. Let's get some GPUs and try to bring it online. How have they fared? Because I think there was a lot of there's obviously so much demand but a lot of there were a lot of

questions or like micro neocoud.

Yeah. Just just somebody saying like hey we we can get some power here let's jump on this. How have they fared? I imagine some of them end up being kind of rolled in to uh you know a platform like coreweave or the other other uh neoclouds or or they end up um but but how have they fared?

You know it it's really tough to comment on on their businesses. We don't see them in market too frequently. I think our product looks a lot different than um the other offerings that are available out there.

Yeah. And and these people were like, "Oh, it's like it's like real estate. You just like put up a box, throw some GPUs in it." And I would say like no one no one you none of the leading Neoclouds are saying like, "Oh, it's just like real estate. You just like throw up a box." So there was that disconnect there.

I I I would agree with that. Um there is a a a chasm of execution between uh uh you know signing a contract to deliver infrastructure and actually bringing it online and making it revenue generating and that's where the core product sits.

Yeah.

Well, it's fantastic progress. Thank you so much for coming and breaking it down for us. Appreciate it.

Uh happy to meet you guys. Congratulations on on the deal. We'll talk to you soon.

Have a good one. Goodbye.

Let me tell you about Gemini 3.1 Pro. with a more capable baseline. It's great for super complex tasks like visualizing difficult concepts, synthesizing data into a single view, or bringing creative projects to life. Uh Tyler, do you know how to count cards in poker?

In poker.

In poker. Do you know how to count cards?

Uh no.

You have exactly one day to learn because you have to go to Sam Blond's Monaco Invitational. You have to bring home the $100,000 cash prize.

You got to win.

There's it's a no buyin poker tournament. founders from all

Is counting cards a thing in poker? That's for blackjack. No. Am I

You're going to learn how to apply counting cards from blackjack to poker? You're going to invent newulate new card counting technology.

I think it I think it does work in Oh, maybe you're right in blackjack. Maybe it doesn't work in poker. I think it does work in poker. Anyway, you're going to learn you're going to cheat and you're going to win. That's the plan. You're going to cheat. You're going to win. You're going to get thrown out of uh Monaco. You're going to get thrown out of San Francisco. Sam Blonde will be putting hands on you, but it'll be a good time.

Put out an article that says the Aremis