Open in interactive viewer → charts, metric popovers & call review

CoreWeave Tells the Bond Market: Thanks, We'll Pass

A 168% revenue quarter and a $133 billion backlog — paired with the quietest strategic pivot in AI infrastructure: the borrower walking away from its own funding market.
CRWV · Earnings Call · 2026-11-10

The quarter that wasn't supposed to happen

On paper, CoreWeave's third quarter is a rebuttal. Revenue of $3.7 billion, up 168% year over year and above the high end of guidance. Adjusted operating income of $379 million against a $200–260 million guide — a 10% adjusted operating margin, reached a full quarter ahead of plan, versus 5% in Q2 and 1% in Q1. Revenue backlog of roughly $133 billion, up 138%. Active power past 1.7 gigawatts, doubling since the start of the year. That step-up is visible in the filings: trailing revenue was $2.6 billion in the June quarter before this print. Revenue compounded 112% year over year into that June quarter, and then added more than a billion dollars sequentially again in September. But the GAAP picture underneath is the one the equity market has been trading. Operating margin was still -21.9% on a GAAP basis in the last filed quarter, and capital expenditure hit $6.4 billion in a single quarter, 249% of revenue. Q3 CapEx was $14.5 billion. Interest expense of $846 million exceeded adjusted operating income by more than two to one, and management guided Q4 interest to $1.05–1.15 billion. Free cash flow, less stock comp, was already -$5.9 billion before this quarter's spend. So the margin flex is real, and the cash burn is also real. What management chose to do about the second one is the story.

The borrower exits stage left

Amid what the CEO called "recent volatility in the capital markets," CoreWeave drew a line in a place no growth company likes to draw it: “we will not use the unsecured debt market to fund our business at current yields. Nor do we need to.” — Michael Intrator, CEO · 2026-11-10 That is a reversal dressed as a policy. On this same call, the CFO explained the prior posture plainly — parent-level unsecured debt was tapped because "these instruments priced at levels that optimized our overall economics." That market is now closed by choice, and the company is substituting three things: delayed draw term loans matched to specific contracts, customer prepayments and deposits, and OEM financing. Roughly $9 billion of prepayments and deposits land in the next three months, more than $3 billion of it from re-contracting infrastructure that is already fully paid off.

We expect to launch our next DDTL in the coming weeks. This will be an investment grade facility with multiple rating agencies indicating a high investment grade rating... we expect the facility to price at a spread that implies a 7% to 7.5% yield based on current SOFR while financing more than 90% of the associated CapEx.

Nitin Agrawal, CFO · 2026-11-10
The weighted average cost of debt is 7.9%, down 250 basis points year over year — and the new paper would come in below that. Add CoreWeave Omni, a lower-capital-intensity wrapper that lets customers own their own hardware and run it on CoreWeave's stack (one customer signed), and the parent-level ask for all of 2027 is under $5 billion. The fresh vocabulary is unmistakable: unsecured debt, parent-level funding, asset-level debt and success-based CapEx now dominate a keyword set that a year ago was about access to capital and cost of debt.

Renewals: from “very limited” to $4 billion

Here is where the transcript contradicts its own history. In August, asked about the installed base rolling off contract, the CFO was dismissive: “In terms of the capacity that's coming up for renewal, Samik, is a very limited part of our fleet.” — Nitin Agrawal, Chief Financial Officer (CFO) · 2026-08-11 Three months later: “More than 130 thousand GPUs come off contract in 2027 as do another approximately 170 thousand GPUs in 2028... At current pricing levels, 2027 renewals would generate approaching 4 billion of incremental annualized revenue with no additional CapEx required.” — Michael Intrator, CEO · 2026-11-10 Nearly $10 billion including 2028. Both statements can be true — a small share of a fleet that is doubling can still be 130,000 GPUs — but the framing shift is enormous, and it is the hinge of the bull case: fully amortized assets re-priced into a market where short-duration capacity clears at $40 million per megawatt annually against $20–25 million for four- and five-year paper. The CFO's payback math makes it concrete: six to eighteen months on short-dated deals, one to two years on long-dated ones at current pricing.

The tape is short the vocabulary

None of this has stopped the market from voting. AI data center exposure is the heaviest decliner cluster on the 30-day tape, with more than fifty negative contributors; critical IT load and data center capacity are down more than 30% over 90 days across the same cohort of tickers, and frontier AI labs has bled through every window. CoreWeave itself is down 32.5% in 90 days and 43.6% from its August peak — a drawdown that predates the print and is about the balance sheet, not the demand curve. Management's answer is that the demand side is not the argument: “We are largely sold out through 2027.” — Michael Intrator, CEO · 2026-11-10 And the fallback is unusually explicit — if sector construction slows, installed, paid-off compute becomes the scarce asset. That is a claim only a company that has already amortized its fleet can make.

What's actually being tested

The bear case was never that AI demand would vanish. It is that a company borrowing at 7.9% to buy depreciating silicon, with $846 million of quarterly interest against $379 million of adjusted operating income, is one funding-market hiccup from a very different conversation. This quarter, management pre-empted that by openly refusing the most expensive money available and lining up $9 billion of customer cash instead — a show of pricing power from the demand side of the table. The test is now mechanical: does the pricing environment hold while the DDTL machine keeps pricing at 7–7.5%, and does the ARR exit at $20–21 billion convert into the promised 2027 EBITDA-less-interest doubling? Adjusted net income positive in the first half of 2028 is the marker management planted. Everything between here and there is a financing story wearing a growth company's clothes.