SpaceX AI Debt Financing Plan Exposes a Bond Market Still Sleepwalking

SpaceX AI debt financing SpaceX AI debt financing

The SpaceX AI debt financing story is being read as a sign of robust investor appetite for the AI infrastructure build-out. The bond market’s pricing, however, tells a more awkward story, one the bulls are working hard not to look at directly.

SpaceX is in talks with CNBC-reported sources confirming that Apollo Capital is said to be playing a leading role in facilitating a $40 billion financing package, structured to fund the purchase of Nvidia GPUs. According to TIKR, the plan calls for roughly $10 billion in bank loans and $30 billion in investment-grade debt, with GPUs expected to serve as collateral. The deal is expected to close in 2027.

That structure deserves a moment’s attention. SpaceX is pledging chips as security for a loan to buy more chips from the company whose sky-high valuation underpins a significant part of the AI equity story. The circularity is not a bug in the narrative, it is the narrative.

SpaceX AI Debt Financing: What the Bond Prices Are Already Saying

This new $40 billion facility follows SpaceX’s bond sale in June, in which the company raised $25 billion spread across several maturities. Demand appeared robust at the time: the book received $90 billion in orders. The $6 billion slice of 10-year senior unsecured notes was priced at a coupon of 5.875%, roughly 120 basis points above the 10-year Treasury yield of around 4.7% at the point of sale. Fitch rated the notes BBB+ and S&P Global rated them BBB, the low end of investment grade, but investment grade nonetheless.

The market has since delivered its verdict. The 10-year Treasury yield now sits at 5.25%, and those SpaceX notes trade at 90 cents on the dollar with a yield of 7.20%, according to TradingView. The spread to Treasuries has widened to nearly 200 basis points, a range consistent with BB-rated bonds, which is the upper tier of junk. SpaceX sold investment-grade paper; the market is now pricing it like high-yield.

The consensus response has been to call this spread widening manageable. And technically, that is not wrong. The ICE BofA BB US High Yield Index Option-Adjusted Spread currently sits at 1.94 percentage points, having briefly touched 2.04 percentage points in early October before pulling back. During the March 2020 shock, the equivalent spread exceeded 8 percentage points. At the peak of the financial crisis, it topped nearly 15 percentage points. Even following the April 2025 “Liberation Day” tariff announcement, the spread briefly cleared 3 percentage points before retreating. By any of those comparisons, today’s levels look calm.

The question worth asking is whether “not panicking” and “correctly priced” are the same thing. The consensus appears to assume they are.

Scale and the Revenue Model Problem

The amount of debt being absorbed is not trivial. According to SIFMA, roughly $12 trillion in US corporate bonds are currently outstanding across investment-grade and high-yield categories. The large AI-related companies alone are directing another $500 billion into that pool this year. Some of that borrowing has already migrated off balance sheet into special purpose vehicles, commercial leases and purchase commitments, a structuring choice that does not reduce the debt, it just relocates it.

Private credit is absorbing a growing share of what the bond market cannot cleanly price. The SpaceX $40 billion facility, structured predominantly through private channels rather than the public bond market, fits that pattern. TIKR reports that Morgan Stanley estimates AI infrastructure will require $1.5 trillion in outside financing by 2028. The bond market, at present spreads, appears to be discounting most of that as routine.

The AI revenue model has not yet demonstrated the capacity to service the debt being raised in its name. Neither consumer nor enterprise spending on AI has reached a level that obviously justifies the planned capital expenditure, let alone the debt layered on top of equity already raised at elevated valuations. The SpaceX IPO raised $86 billion. OpenAI and Anthropic are on the horizon. The Firmus IPO in Australia, which had been seeking to raise about $5.5 billion, was pulled after investors balked. That last data point is the market doing its job. The rest of the picture, B-rated spreads at a historically narrow 3.15 percentage points, BB spreads below 2 percentage points, no AI-related loan or bond in default, is the market not doing it.

Elsewhere in the credit stack, the Paramount Skydance acquisition of Warner Bros. Discovery is being funded by $52 billion in debt sales, including $44 billion in bonds of which $12 billion are junk-rated, leaving the combined entity carrying roughly $80 billion in total debt. AI is not the only source of pressure on the system’s absorptive capacity; it is simply the largest and the fastest-moving one.

The $40 billion SpaceX facility, expected to close in 2027, will be a useful test. GPU collateral has never been stress-tested in a credit downturn. When the deal prices, the spread it clears will say more about where the bond market actually is than any index currently suggests.

Add a comment

Leave a Reply

Your email address will not be published. Required fields are marked *

Keep Up to Date with the Most Important News

By pressing the Subscribe button, you confirm that you have read and are agreeing to our Privacy Policy and Terms of Use