CoreWeave: The Canary in the AI Coal Mine
If you wanted to design a company that would be the first to break in an AI infrastructure downturn, you would design something that looks a lot like CoreWeave. Not because the company is poorly run — by most accounts, the team is technically excellent — but because the financial structure of the business concentrates almost every risk factor the AI buildout carries into a single entity.
$14 billion in debt. A leverage ratio of 9x EBITDA. Interest coverage of 0.17x, meaning the company’s operating income covers roughly one-sixth of its interest payments. And 62% of its revenue comes from one customer: Microsoft.
To understand why these numbers matter, you need to understand CoreWeave’s business model. The company buys NVIDIA GPUs in bulk — tens of thousands at a time — installs them in data centers, and rents compute capacity back to customers on contracts. It’s a capital-intensive intermediary: borrow money, buy hardware, lease it out, use the lease revenue to service the debt, and hope the margins hold.
When demand for GPU compute is high and rising, this model prints money. CoreWeave can charge premium rates because capacity is scarce, customers are locked into multi-year contracts, and the hardware depreciates slowly enough on paper to maintain attractive unit economics. This is the bull case, and for the past two years, it has been the reality.
The bear case is what happens when any of those assumptions shift. If GPU supply increases faster than demand — which falling spot prices suggest is already beginning — CoreWeave’s pricing power erodes. If Microsoft renegotiates, reduces, or doesn’t renew its contracts, 62% of the revenue base is at risk. If NVIDIA’s next-generation hardware makes current GPUs significantly less competitive, the assets backing $14 billion in debt lose value faster than the depreciation schedule assumes.
The debt structure amplifies each of these risks. At 9x EBITDA leverage, even a modest decline in revenue or margins cascades through the capital structure. The 0.17x interest coverage ratio means there is almost no buffer. In a stable environment, the company can refinance and roll forward. In a tightening credit environment — the kind that widening AI infrastructure credit spreads suggest is developing — refinancing becomes more expensive or unavailable.
CoreWeave’s position in the structured credit market adds a systemic dimension. The company’s debt has been packaged into asset-backed securities and sold to institutional investors. JPMorgan projects that data center securitization could reach $30-40 billion per year by 2026-2027, representing 7-10% of combined ABS/CMBS issuance. When one company with this risk profile represents a meaningful portion of a growing asset class, its distress doesn’t stay contained. It reprices the entire category.
This is why CoreWeave Distress Indicators is one of our four upcoming signals, weighted at 10% of the composite score. We’ll track debt levels, revenue concentration, coverage ratios, and refinancing conditions from SEC filings and public credit research. The signal isn’t a bet against CoreWeave specifically. It’s a recognition that the company’s financial structure makes it the most sensitive barometer for AI infrastructure credit risk.
If CoreWeave’s numbers hold, it’s evidence that the AI infrastructure buildout can support aggressive financial engineering. If they don’t, it’s the earliest warning signal that the credit cycle is turning. Either way, the data will tell us before the headlines do.
The data updates daily. The analysis goes deeper.
Back to the Index