SpaceX Reportedly Wants to Borrow $40 Billion for Nvidia Chips Despite a $100 Billion Cash Pile

SpaceX (NASDAQ:SPCX) ended June with $100 billion in cash, cash equivalents, and marketable securities, mostly from its $86 billion initial public offering (IPO) the same month. But the company now reportedly wants to borrow another $40 billion.

The rocket and artificial intelligence (AI) company is lining up around $10 billion in bank loans and $30 billion in investment-grade bonds to buy chips from Nvidia, according to a report on Tuesday. Asset manager Apollo Global Management is expected to lead the deal, which should close in 2027.

SpaceX already had about $39.5 billion in debt and finance leases at the end of June. Shares were down about 2% on Wednesday, near $168 as I write.

Image source: Getty Images.

Why borrow with $100 billion in the bank?

The cash pile is huge, yet SpaceX is spending it fast, and the pace has been speeding up. Capital expenditures were $18.4 billion in the second quarter, up from $10.1 billion in the first quarter and $2.8 billion a year before. Of the latest total, $15.8 billion went to AI computing infrastructure. In the first half of 2026, SpaceX's operations brought in around $3.5 billion in cash, but capital spending was $28.5 billion.

Put simply, outside money is funding almost all the build-out.

And spending isn't expected to fall this year. On the August earnings call, CFO Bret Johnsen told analysts to expect capital spending in each of the next two quarters to be very similar to the second quarter's. Two more quarters like that would eat up around $37 billion, or over a third of the cash pile.

But next year looks even bigger. CEO Elon Musk said on the same call that SpaceX's computing capacity by the end of 2027 could be closer to 10 gigawatts than 5, compared to 1.4 gigawatts at the end of June. SpaceX also expects to close its purchase of EchoStar's wireless spectrum in late 2027, a deal that includes up to $8.5 billion to pay off EchoStar debt.

In short, the $100 billion probably can't fund the full plan. And a financing that closes in 2027 would arrive just as those bills come due.

Double the debt

At the end of 2025, SpaceX owed around $23.3 billion in debt and finance leases. By June, that had risen to $39.5 billion. The largest chunk is $25 billion in bonds sold in June, partly to repay a bridge loan, and most of the remaining $13.4 billion of other financing is linked to AI hardware.

Adding $40 billion would take the total to roughly $80 billion, about double. Still, SpaceX is worth about $2.3 trillion in the stock market, and none of the June bonds mature before 2031.

Borrowing has one edge over selling more stock: It doesn't add shares. SpaceX already issued around 391 million new shares to buy AI coding firm Cursor in August.

Debt also fits management's stated priorities.

"Across all three of these investments, we remain focused on capital efficiency, return on investment, and preservation of a strong investment-grade balance sheet," Johnsen said on the call.

Starlink would be paying the interest

I'd say the bigger cost shows up on the income statement. SpaceX's interest expense was $629 million in the second quarter, up 53% year over year and a bit under the first quarter's $664 million.

The reported terms don't give an interest rate. But if the new debt costs about what the June bonds did (a weighted average of 5.855%), $40 billion would add about $2.3 billion of interest a year, or roughly $585 million a quarter. That alone is bigger than SpaceX's entire $541 million net loss in the second quarter.

For now, Starlink is what makes the bill payable. SpaceX's connectivity segment, which includes the satellite internet business, earned $1.7 billion of operating income in the second quarter, but the AI segment lost $1.3 billion. Increasingly, profits from selling internet service might end up paying the interest on debt taken on to buy chips.

I don't think the borrowing is a red flag by itself. At this rate of spending, SpaceX was always going to need outside money, and debt is arguably easier on shareholders than more stock. But it raises the cost of being wrong.

At around 73 times sales (based on second-quarter revenue annualized), the stock already assumes the AI spending turns into profits. Billions of dollars a year of added interest leave less room if it doesn't. Until the AI segment makes enough to pay its own share of that bill, I think the price is too steep.

Daniel Sparks and his clients do not have positions in any of the stocks mentioned. The Motley Fool has positions in and recommends Nvidia. The Motley Fool has a disclosure policy.