CoreWeave's Bonds Are Being Priced Like Distressed Debt While Its Backlog Says Otherwise

In this storyCoreWeave

CoreWeave shares closed down about 15.5 per cent on 14 October after yields on its bonds climbed close to 15 per cent, on investor concern about the debt being raised to fund artificial intelligence infrastructure.

The 9.625 per cent notes due 2032 yielded roughly 14.8 per cent and briefly touched 15 per cent, while the 9.75 per cent notes due 2031 yielded around 14.9 per cent. Bond yields move inversely to prices, so the increase reflects creditors demanding a higher return to hold the paper. Both now trade more than five percentage points above their coupons.

The selloff spread across companies seen as dependent on external financing. Oracle fell close to 9 per cent, Nebius around 10 per cent and Dell roughly 10 to 11 per cent, contributing to a 1.4 per cent decline in the Nasdaq Composite.

Attention was drawn to the yields by investor Chamath Palihapitiya in a post on X on the evening of 13 October. Elon Musk responded to bubble warnings with a one-word reply.

CoreWeave carries a B+ credit rating from S&P, which is below investment grade. Its debt stood at $14.6 billion a year earlier and has more than tripled since, while sales rose 115.3 per cent over the year to June 2026. The company has raised an estimated $30 billion to $35 billion in combined debt and equity during 2026. Its interest expense already exceeds adjusted operating income.

Chief executive Mike Intrator called the market reaction irrational, saying the company's backlog more than covers its debt obligations. The company is continuing with commitments including a 240 MW buildout in India.

A bond yielding five points above its own coupon is not a sentiment indicator. It is a specific statement about repayment.

When CoreWeave issued the 2032 notes, buyers accepted 9.625 per cent as fair compensation. Anyone buying that same paper now demands close to 15 per cent, which means the existing holders are sitting on losses and the company's next dollar of borrowing costs materially more than its last one. For a business whose interest expense already exceeds adjusted operating income, that is not an abstraction. It is the arithmetic of the next refinancing.

Intrator's defence is that the backlog covers the obligations, and on the face of it he has a case. Revenue grew 115.3 per cent in the year to June 2026, which is not the growth profile of a company in difficulty. The contracts are real and largely with creditworthy counterparties.

But backlog and debt service are not the same instrument. Backlog is revenue that arrives over years, contingent on delivery, on data centres being energised on schedule and on customers honouring terms. Debt service is cash due on fixed dates. The gap between those two timetables is precisely what a 15 per cent yield prices, and it is why debt has more than tripled while revenue roughly doubled.

The uncomfortable detail is who noticed. This repricing was not triggered by an earnings miss or a downgrade. It began with a venture investor pointing at a screen on X on a Tuesday evening. Markets that move on that are markets already looking for a reason.

The contagion pattern is the part worth watching, because it was selective.

Oracle, Nebius and Dell fell hard. The Nasdaq fell 1.4 per cent. That gap means the market was not selling technology, it was selling companies that borrow to buy chips and lease them out. Investors have drawn a line around balance-sheet-funded AI capacity and marked it down as a category, separately from the companies generating the demand.

That separation is new and it is awkward for India. CoreWeave is midway through a 240 MW buildout here, and Indian data centre capacity has been financed on similar logic: long contracts, depreciating hardware, debt against both. If the cost of that debt is repricing globally, projects announced at one financing cost get built at another, or do not get built.

It also lands against a domestic backdrop where AI funding is already concentrated. Of the $438 million that went into Indian AI companies in the third quarter, roughly 90 per cent came from four deals, and 63 per cent of surveyed investors said valuations were too high. A credit market that has started charging 15 per cent for AI infrastructure debt does not make that concentration looser.

Infosys chief executive Salil Parekh warned last week that the data centre backlash building elsewhere would reach India. The financing version of it appears to have arrived first.

For daily, sharp analysis of the biggest moves in the Indian business and startup ecosystem, follow StartupFox.