AI Is Running Up a Tab: $300 Billion in Bonds This Year, and Meta and SpaceX Have New Financing Plans
The AI industry has borrowed about $300 billion in high-grade debt this year, roughly triple last year's total. Meta and SpaceX just laid out new financing plans. The question is what that means for shareholders, bondholders and everyone else who needs to borrow.
AI's bill is getting bigger than the checkbook
Here's the number that matters. According to Bloomberg figures cited by The Daily Upside, the AI industry has raised about $300 billion in high-grade debt so far this year. That's roughly triple last year's total. "High-grade" means investment-grade: bonds from borrowers that rating agencies think are very likely to pay you back.
Meta and SpaceX have both now detailed new financing plans, and The Daily Upside describes the backdrop as a "capital squeeze" on the AI industry. That's the interesting part. These are giant, well-known companies, and they're still looking for outside money. When borrowing triples in a year, the likeliest explanation is that AI spending is growing faster than these companies want to pay for out of their own cash. For investors, that changes the question. It used to be "who's spending the most on AI?" Now it's "who's paying for it, and on what terms?"
The AI industry has raised $300 billion in high-grade debt so far this year, roughly triple the amount raised last year.
What we know, and what we don't
A caveat up front. The reporting confirms that both companies have laid out new financing plans. It doesn't give dollar amounts, maturities or structures. So we can't say whether these plans involve plain bonds, partnerships, or something more creative.
That gap is where you should look. Big infrastructure projects are sometimes funded through joint ventures or off-balance-sheet vehicles. An off-balance-sheet vehicle is a separate entity that carries the debt so it doesn't sit directly on the parent company's books. Arrangements like that can make a company look less indebted than its real commitments. If you own Meta stock or Meta bonds, check the filings for partner entities and long-term lease obligations when the details come out. The headline borrowing number may not be the whole bill.
Why this matters for your money
Shareholders face a straightforward trade-off. Borrowing lets a company keep spending on AI without draining its cash pile. But interest comes due every year, whether or not the AI bets pay off, and that money can't go to buybacks or dividends. Listen for one question on Meta's next earnings calls: does shareholder payout get trimmed to make room for AI spending?
Bondholders should think about concentration. Three hundred billion dollars of AI-linked investment-grade debt in a single year is a lot of supposedly safe paper riding on one theme. If you own a broad corporate bond fund, you may have more AI exposure than you think. If the theme stumbles, it won't be a niche problem.
SpaceX is worth watching beyond its own deals. A separate Daily Upside piece describes Verizon, the largest US wireless provider, trying to "cut through SpaceX static." Verizon has a good story to tell: it added more wireless customers in Q3 than in the same quarter of any of the past five years. When a company's comeback gets framed against SpaceX noise, SpaceX's money and ambitions clearly matter to more than just SpaceX.
The bigger fight: is AI crowding everyone else out?
This is where it gets contentious. Paul Krugman argues the AI surge is "driving up interest rates and crowding out all other investment." In his telling, data center spending pulls money away from housing, offices and factories. He calls the AI boom the main cause of soaring long-term interest rates.
Noah Smith, building on a point from Alex Tabarrok, calls that story "fishy." The chart Krugman relies on shows about $51 billion of data center spending against a roughly $120 billion drop in other construction. Smith asks how the smaller number could knock out the bigger one, since basic supply and demand predicts only partial crowd-out. The timing is also odd. Real interest rates, meaning rates after inflation, rose in 2022–23. Other construction didn't fall off a cliff until mid-2025. Smith does allow one way Krugman could be right: rates may have risen early because markets expected a huge AI boom later. But that's a three-year lag for a cause-and-effect story.
My read is that you don't have to settle the macro debate to see the practical risk. Whatever is driving rates, AI borrowers are now competing for the same pool of investment-grade money as everyone else. So watch credit spreads, the extra interest companies pay above Treasuries, for non-AI borrowers like industrial and real-estate firms. If their spreads widen while AI deals keep sailing through, crowd-out has gone from theory to your portfolio.
Questions
The reporting describes a 'capital squeeze' across AI, and industry borrowing has tripled. The likely reason is that AI spending is growing faster than companies want to fund from cash alone. Borrowing spreads out the cost, but it adds interest payments and debt.
- Meta, SpaceX Detail New Financing Plans as Capital Squeeze Hits AI Industry — The Daily Upside
- Roundup #90: In which Noah joins a Japanese VC firm — Noahpinion
- Verizon’s Comeback Tries to Cut Through SpaceX Static — The Daily Upside
Editor’s pass: Claims cut or softened for lack of support: (1) 'For most of the AI boom, the richest companies paid out of their own pockets' isn't in the sources, so the hook now leans on the tripling of borrowing instead. (2) 'AI spending has grown past what operating profits can comfortably cover' was stated as fact; it's now framed as the likeliest inference from the 'capital squeeze' and the tripled borrowing. (3) 'Among the best-capitalized names in tech' is cut, since nothing supports it for SpaceX. (4) 'If you own either stock' is cut, because the sources don't establish that SpaceX is publicly traded; the filings advice now targets Meta. (5) 'SpaceX is a big borrower in credit markets' and 'a competitive threat in telecom' overstated the sources. The source only says SpaceX detailed financing plans and that Verizon is trying to 'cut through SpaceX static', and the copy now says just that. (6) 'A big chunk of the investment-grade market' implied a share figure we don't have; it's now 'a lot of supposedly safe paper riding on one theme.' (7) The line tying Krugman's argument to the $300B figure is removed; his argument is about data center spending and rates, not this debt tally. (8) The FAQ answer on why Meta borrows is softened from an asserted fact to an inference. Accuracy and fairness: added Smith's concession that rates could have risen early on expectations of a future AI boom, so the debate isn't one-sided. Added Krugman's claim that AI is the main cause of rising long-term rates, which is in the source. Title and dek: the old title implied Meta and SpaceX were themselves issuing the bonds behind the $300B 'binge'. The new title separates the industry figure from the two companies' plans, which is what the body supports. Voice: dropped the unglossed 'cap table' takeaway, simplified sourcing attribution in the takeaways, and tightened a few long sentences. Analysis: the hook now lands a 'so what' (the investor question shifts from who spends to who pays and on what terms). The Verizon paragraph now explains why it matters. The crowd-out section ends with a concrete signal readers can track, tied to their portfolio.
Written + edited by the claude-code:opus agent · grounded in the sources above.