SpaceX is gonna need to borrow some more money. Several tens of billions more, reportedly.
And that makes sense. You think it’s cheap to colonize Mars? Or build the Starship Enterprise? Not likely. We gotta get off this rock before it becomes uninhabitable (I know, I know: “Too late!”), and much as it pains me to admit this, Elon Musk is probably our best hope.
So, sure. Let’s loan the company some more money. Assuming, of course, it’s earmarked for Martian settlements and other things you’d expect a company called “SpaceX” to be pursuing, not — and I’m just spitballing here — Nvidia chips.
You probably know where I’m going with this. If not, you don’t read the Financial Times, and I can’t blame you: It’s $75/month for God’s sake, which makes me cringe every time I see it on my debit card statement. That’s enough to buy six pounds of organic salmon at Harris Teeter and let me tell you: That’s some good-ass salmon. (“How come you didn’t tell me how good the salmon was?“)
Anyway, according to the FT, SpaceX is working with friend-o-Musk Apollo to arrange up to $40 billion in financing for the sole purpose of buying Jensen Huang’s latest chips. $30 billion of that would be IG debt, the rest loans.
I mention this for what I hope are obvious reasons. First, it underscores the notion that SpaceX, for all the interstellar hoopla, is a data center REIT. A neocloud on steroids. (And ketamine. Allegedly.) Recall from SpaceX’s first earnings report as a publicly-traded company that the space segment accounted for a mere 12% of overall revenue last quarter, down from 18% the prior year.
Second, it suggests that notwithstanding fruitful efforts to develop competitive AI models with cheaper hardware, there’s still no substitute for Nvidia’s bleeding-edge chips. That’s good for Huang and Nvidia shareholders in the near- to medium-term. But it raises uncomfortable questions about the ROI on massive outlays for the company’s technology.
Third, it’s a nod to the notion that high-grade corporate debt supply to fund the AI race will continue apace, with ramifications for government issuers, who are now competing for capital with the largest, most creditworthy, companies in the history of capitalism.
Assuming the new SpaceX financing eventually gets done, it’d be the second time SpaceX sold bonds since going public. The first deal, June’s $25 billion offering, was heavily oversubscribed. I imagine the same will be true for the $30 billion in IG debt at the center of the new raise, which is expected to close in 2027.
The figure below shows you the history of hyper-scaler “plus Nvidia, plus SpaceX” debt sales since Oracle kicked off the borrowing binge a year ago last month. It’s updated with a couple of new Amazon and Alphabet offerings (a sterling deal and a Kangaroo sale, respectively).
Factoring in the loan portion of the prospective SpaceX financing, the deal would count as the second-largest AI-related raise yet, behind Amazon’s dollar-and-euro sale in March.
The news comes just a week after Bloomberg said Broadcom’s lining up as much as $60 billion to help Anthropic buy chips.
I don’t know when we’re going to start classifying SpaceX as a hyper-scaler, but… well, it’s a hyper-scaler ain’t it? What else is it? A rocket company, supposedly, but at this point that’s about like calling Bobby Bacala a train conductor instead of a mobster.
“I wonder what H is doing right now? He’s probably eating salmon at a bartop. If they don’t tell him how good the salmon was, he’d be mad.”
Well, we can wager on a tsunami of breathless mega-bullish AI commentary from Goldman and Blackrock once (or if) this goes through. “If” just in case this proves to be a canary in the coalmine event. Hah, it’s AI!
i come for the analysis, but i stay for the pop references.
President: Well, I would hate to have to decide who stays up and who goes down.
Dr: Well, that would not be necessary, Mr. President. It could easily be accomplished with a computer. And a computer could be set and programmed to accept factors of youth, health, sexual fertility, intelligence and a cross-section of necessary skills. Of course, it would be vital that top government and military men be included to foster and impart the required principles of leadership and tradition. Heil! Actually, they would breed prodigiously, yeah? There would be much time and little to do. With a proper breeding techniques and a ratio of, say, ten females to each male, I would guess that they could interact their way back to the present gross national product within, say, 20 years.
DR. STRANGELOVE. I hasten to add that since each man will be required to do prodigious service along these lines the women will have to be selected for their sexual characteristics which will have to be of a highly stimulating nature.
ALEXI. I must confess you have an astonishingly good idea there Doctor.
DR. STRANGELOVE. Thank you sir.
One of my favorite flicks!
Can we say there’s a new debt count, re-set to zero (SPCX bringing the first 40 bil) now that we’re talking SUPERINTELLIGENCE borrowing?
Arsenio getting hostile in that skit always killed me.
The subtle change in Dave’s face when he says “he’d be mad” is what still gets me after all these years.
Idk if SpaceX is a hyper scalar. Do they offer public cloud computing like the others? They feel like a data center reit,
Does Meta offer public cloud computing? No. They might soon, but they don’t now, and they haven’t for the entirety of the (short) period the term “hyper-scaler” has existed. Also, I specifically called SpaceX a data center REIT in this article, and indeed I was among the first to use that description for the company once it merged with xAI. The point is: If a hyper-scaler is first and foremost a company that’s spending massively on capex to fund the AI buildout, and using a lot of debt to do so, then that’s what SpaceX is on its way to becoming. Or at least that’s what it looks like.