$3.54bn Expected, Interest Bill Above Operating Profit (CoreWeave Earnings Preview, 10 November 2026): What to Expect From Q3 — and Why Power Online and Financing Cost Decide the Read

CoreWeave reports third-quarter 2026 results on Tuesday 10 November after the US close, with the call at 5 p.m. Eastern (22:00 UTC). It guided revenue of $3.45 billion to $3.6 billion and the Zacks consensus sits at $3.54 billion. The number most likely to matter is not revenue but whether power is coming online fast enough to cover an interest bill that is running at more than three times operating profit — five days after the market sold the data-center build-out on fears it will be slowed.

When CoreWeave reports, and what it guided

CoreWeave confirmed on 27 October that it will publish third-quarter results after the market close on Tuesday 10 November and hold its call at 5 p.m. Eastern Time, according to its investor-relations announcement. With US clocks back on standard time since 1 November, that is 22:00 UTC and 06:00 on Wednesday in Hong Kong.

The guidance comes from the company's second-quarter call on 11 August. Consensus figures are the Zacks estimates published on 3 November, carried by Yahoo Finance.

The guided range implies sequential revenue growth of roughly 34% to 40% from the second quarter. CNBC reported the low end of the range as $3.4 billion; the corrected transcript has management saying $3.45 billion, and that is the figure used here.

The business in one paragraph

CoreWeave rents computing capacity built on Nvidia GPUs to AI developers and large technology companies. It signs multi-year contracts, buys the hardware, leases or builds data-center space, and borrows heavily to pay for it. Revenue starts when a contracted cluster is switched on — which needs the building finished, the hardware delivered, and, above all, power connected. Its revenue backlog was $104 billion at the end of June, up 246% year on year, and management said that figure excluded more than $25 billion of new customer commitments signed early in the third quarter. CNBC reported those included an additional $21 billion of spending from Meta and a $6 billion commitment from Jane Street, alongside a multi-year agreement with Anthropic.

So the demand question that dominates chip earnings is largely answered in advance for CoreWeave. What is not answered is the timing and the cost of turning that backlog into revenue.

The real question: power online

The single figure that links the backlog to revenue is active power — the capacity actually energised and earning. CoreWeave ended the second quarter with 1.5 gigawatts active, adding nearly 500 megawatts in the quarter, more than 300 megawatts of it in June alone. It raised its year-end target to more than 1.85 gigawatts, said contracted power had reached 4.2 gigawatts, and reiterated a goal of at least 8 gigawatts by 2030.

Every step in that chain costs money before the last one pays. A large June energisation flows into third-quarter revenue, which is why the Q3 guide is a big step up. The fourth quarter depends on how much more power was switched on between July and December. That makes the quarter-end active-power figure, and any change to the 1.85 gigawatt target, more informative than the revenue line itself.

Why 5 November changes the questions on the call

Five days before the report, the market re-priced the whole data-center chain. On 5 November the Nasdaq fell 5.52%, the semiconductor index more than 12%, and CoreWeave almost 10%, according to Reuters, after the midterm results added to concern that tighter state and local rules could slow construction. We walked through that mechanism in our note on the 5 November sell-off: the bottleneck is the site, not the chip.

CoreWeave's chief executive, Michael Intrator, was asked about exactly this in August, after New York's governor ordered a moratorium on new large-scale data centers in July. On the call he said moratoriums are "going to impact where this infrastructure gets built," not the demand for it, and that "none of those numbers will be impacted by the regulatory pushback as of today."

That was a statement about the guidance in force on 11 August. The 10 November call is the first chance for management to say whether anything has changed since — whether projects have been delayed or relocated, whether the 1.85 gigawatt year-end figure still holds, and how much of the 4.2 gigawatts contracted sits in jurisdictions where rules are tightening. Those answers address the specific doubt the market priced on 5 November, which revenue in the third quarter cannot.

The financing line: interest above operating profit

The second variable is cost. CoreWeave guided third-quarter interest expense of $860 million to $940 million, up from $640 million in the second quarter and $267 million a year earlier. Against that, it guided adjusted operating income of $200 million to $260 million. In other words, the company is currently paying more than three times as much in interest as it earns before interest, and is borrowing more each quarter to fund capex of $11.5 billion to $13.5 billion.

That is a deliberate structure rather than a hidden flaw: the debt is largely secured against hardware tied to specific customer contracts, and management argues that margins rise as clusters move from ramp-up to steady operation. It guided adjusted operating margin to reach the "low teens" in the fourth quarter, from 5% in the second. It also said its weighted average cost of debt had fallen by almost 300 basis points over the year, worth about $1.1 billion a year in interest at the end-June debt load, and that it held more than $6.9 billion of cash and equivalents at quarter-end. CNBC put total debt at about $35 billion.

The sensitivity runs in both directions. If power arrives on schedule, interest is paid on assets that are earning; if it slips, interest accrues on assets that are not. That is why a delay in energisation hurts a leveraged landlord more than it hurts a hyperscaler funding the same build from operating cash flow — the same contrast that showed up in the 5 November price moves, and the same financing theme we covered in SpaceX's $40 billion debt raise for AI hardware.

Three scenarios, and where each one lands

None of these is a forecast. The point is that the same revenue print can sit inside any of the three, so the market's reaction is likely to be driven by the guidance and call commentary rather than the headline.

How one AI-cloud report reaches the indices — and the currencies

CoreWeave is small next to the mega-caps, so its own move contributes little to the Nasdaq-100 or S&P 500 directly. Its influence is as a read on the build-out. It sits between the chipmakers that supply it and the AI developers and hyperscalers that rent from it, and its disclosures on power, permitting and customer timing feed into how investors model Nvidia's next quarter, the capex plans behind Microsoft's data-center leases, and power-equipment orders. Nvidia's own report follows a week later, on 17 November.

The currency channel is indirect and should not be overstated. A sharp risk-off move in technology tends to show up in safe-haven demand — the yen and the Swiss franc — and in the dollar through risk appetite and flows. A single AI-cloud report rarely moves those on its own; it matters to currencies only if it reinforces or reverses a broader shift in risk sentiment like the one on 5 November.

What to watch on the night

• Active power at 30 September, and whether the "more than 1.85 gigawatts" year-end target is held, raised or cut.
• Any reference to delayed or relocated projects, and the share of contracted power in states or localities with new restrictions.
• Backlog, including whether the more than $25 billion of early-Q3 commitments is now counted, and how much of it has started delivery.
• Interest expense and new financing: the rate on any debt raised since August, and whether the margin path to the "low teens" in Q4 is reaffirmed.
• Full-year capex against the $35 billion to $39 billion range — a cut can mean either discipline or delay, and the call will need to say which.

How pip theory approaches these previews — mechanism first, no forecasts — is set out on the about page.

Educational macro context only — not investment advice.