$40 Billion in Chips, Funded by Rockets

SpaceX just signed the largest GPU deal in history.

And the strange part? It wasn't the AI that got the money. It was the rockets.

On Saturday, Musk posted that SpaceX had closed a $40 billion equipment lease with Apollo Global Management. The chips are NVIDIA's Vera Rubin processors. They'll power SpaceX's terrestrial compute buildout.

Let me explain.

The deal is structured as a master finance lease. An Apollo-managed entity buys the GPU pods. Then it leases them to SpaceX over five years. SpaceX pays rent each quarter… roughly one-twentieth of the total cost per payment, plus a fixed rate on the balance.

That fixed rate is where the story gets interesting.

Semafor reported the lowest-rated slice of the deal carries about 8.3% interest. A safer slice, backed by the chips, prices near 7.75%. Both are more than a percentage point above where SpaceX's existing bonds trade.

That's expensive. So why did SpaceX get the money at all?

Because lenders trust Starlink. SpaceX booked $14.6 billion in Q3 revenue. That's up 161% from a year ago. Net income hit $4.4 billion. The company is investment-grade. It has cash coming in every month from millions of broadband subscribers and dozens of rocket launches.

Musk said it plain: “Debt markets have almost entirely shut off access to money for any company that doesn't have substantial revenue independent of AI data centers.”

In other words, if your only business is renting GPUs, you can't borrow right now. The credit window slammed shut. Pure-play AI infrastructure firms… the so-called “neoclouds”… are locked out.

But SpaceX isn't a neocloud. It's a rocket company that happens to want chips.

That's the contrarian read here. Wall Street spent the past year pricing SpaceX as an AI stock. The Q3 earnings beat came from compute and Starlink growth. Analysts call it a tech platform now.

And yet the $40 billion loan closed because of the old business. Launches. Satellites. Subscriptions. The stuff that sounds boring on an earnings call but looks fantastic on a credit application.

Now, I know what you're thinking. Forty billion dollars in new debt sounds risky. And you're right to ask.

But the lease runs through a special-purpose vehicle. That means the debt sits off SpaceX's balance sheet. It doesn't show up alongside the company's existing borrowings. And it doesn't threaten the credit rating that made the deal possible in the first place.

Chamath Palihapitiya called SpaceX's debt “better than most sovereign debt.” That may be a stretch. But the logic holds. When lenders can't trust AI revenue alone… they trust rockets.

SPCX closed at $146.90 on Monday. Down less than half a percent.

The market shrugged. It shouldn't have.

Elon just created a device he believes will be "the biggest product ever."

He thinks it could 70x investors' money.

And he's about to make a major announcement…

By the end of this month.

Maybe even tomorrow on X.

He's going to make this game-changing device available to the public for the first time.

He has to sell 1 million to become a trillionaire.

Would you bet against him?

Click here to find out how you can claim a stake in this 70x AI agent.

Notice how the money is shaped. Rather than borrow $40 billion and buy the chips, SpaceX had an Apollo-managed entity buy the GPU pods and lease them back — a master finance lease, with rent paid quarterly as hardware arrives in tranches through 2027. Two advantages fall out of that. SpaceX conserves cash, paying over five years instead of all at once, and the obligation sits in a separate vehicle rather than piling onto its own balance sheet. The lessor, Apollo, is really betting on two things: that Nvidia's chips hold their value, and that SpaceX keeps paying. Watch whether other hyperscalers copy this lease-not-buy structure.

Read the subtext of Musk's victory lap. Over the past two years, a wave of “neocloud” startups — companies whose entire business is renting out GPUs — raised enormous sums betting on endless AI demand. Now debt markets have turned cautious, and lenders are refusing to fund firms with no revenue beyond AI data centers. Musk predicts most of the 200-plus neoclouds will fail or get absorbed. SpaceX, with rockets and Starlink throwing off cash, can borrow where they can't. The lesson is old… in a credit squeeze, the diversified survive and the pure-plays starve. Watch for the first big neocloud failures.

Follow the cash going out the door. SpaceX's quarterly capital spending has surged toward $20 billion, up from under $7 billion a year earlier, and the CFO signaled $35 to $40 billion more in the fourth quarter alone — a spending pace that even SpaceX's huge cash reserves can't sustain from cash flow, which is exactly why the Apollo lease exists. This is the real reason the story matters: the AI buildout is so expensive it has to be financed, and only a company with proven, non-AI revenue can finance it right now. Watch whether the compute revenue grows fast enough to justify a spend this size.

Watch whether SpaceX's AI-compute revenue grows into the enormous bill it just signed.

Here's the logic of the bet. SpaceX is committing to roughly $40 billion in lease payments over five years for chips that start arriving in 2027. That rent comes due whether or not the compute business booms. So the whole deal rests on one question: can SpaceX lease out that computing power — to Grok, to enterprise AI, to cloud customers — faster than the payments pile up?

The near-term signals are concrete. Chip deliveries begin in the first half of 2027, and each tranche starts its own rent clock. Watch the AI-compute revenue line in upcoming quarters against that rising obligation.

Why the structure protects SpaceX, up to a point. Because the debt sits in a separate vehicle and the company has strong non-AI cash flow, a slow compute ramp wouldn't sink the whole business the way it might sink a pure-play neocloud. SpaceX has a cushion its rivals lack.

But the risk is real. If AI demand cools or the chips underdeliver, SpaceX is still on the hook for the rent — and a $40 billion commitment is a heavy thing to carry if the revenue doesn't show.

So watch two markers through 2027: the pace of chip deliveries actually deploying, and whether compute revenue climbs fast enough to cover the lease. Those decide whether this was a masterstroke or an overreach.

Let's keep this simple.

Imagine two people walk into a bank, each asking to borrow a fortune.

The first has a dazzling new idea but no income yet — just a promise that it'll pay off big someday. The second runs a dull, steady business that brings in reliable cash every single month.

Who gets the loan? The boring one… every time.

That's exactly what just happened with SpaceX.

For the past year, the exciting story about SpaceX has been AI — renting out computing power, buying Nvidia chips, chasing the same boom as everyone else. That's the dazzling idea.

But the thing that actually unlocked $40 billion in financing wasn't the AI. It was the unglamorous stuff: rockets that launch on contract, and millions of people paying their Starlink bill every month. Steady, boring… dependable cash.

Here's why that matters right now. Lenders have gotten nervous about AI. So many companies borrowed so much on nothing but the promise of future AI profits that the money has dried up for anyone who can't show real income today.

The companies whose only business is AI — the “neoclouds” — are now locked out. They have the dazzling idea and nothing else, so no one will lend to them.

SpaceX walked in with the dazzling idea and the boring cash machine. The boring half is what got it the money. Lenders could look past the uncertain AI bet because rockets and subscriptions guarantee the rent gets paid.

And that's the lesson hiding in a flashy headline. In good times, the market pays up for the exciting story. But when money gets tight, it's the dull, reliable cash flow that saves you — and that lets you keep chasing the exciting story while your rivals can't.

Follow the boring revenue, not the shiny narrative — when credit tightens, the steady cash is the thing that actually keeps a company in the game.