CoreWeave Q3 2026: $3.66B Revenue, $9B in Prepayments and a Shift to Shorter Contracts

CoreWeave reported third-quarter 2026 results on Tuesday, November 10, and the headline numbers sit alongside a financing change. According to TechCrunch, revenue was $3.66 billion, up 168% year over year, with a net loss of $709 million, and the company expects about $9 billion of customer prepayments and deposits over the next three months. This post separates the reported figures from our own arithmetic and looks at what the pricing and contract-length shifts could mean for how AI capacity gets financed. For earlier context, see our Q2 earnings analysis and our look at CoreWeave bond yields.

What CoreWeave reported

TechCrunch's report does not give a separate capital expenditure figure for the quarter, so we do not estimate one.

Our arithmetic on the reported numbers

Several ratios follow directly from the reported figures. They are our calculations, not company statements.

• Interest versus revenue: $846 million of interest on $3.66 billion of revenue is about 23%. Interest expense was larger than the net loss, which means the company's loss was smaller than its interest bill in the quarter.
• Net loss margin: $709 million on $3.66 billion is about 19%.
• Backlog coverage: $132.5 billion divided by an annualized revenue run rate of about $14.6 billion (four times Q3) is roughly nine years of revenue at the current pace. Contract backlog is not the same as recognized revenue, and delivery depends on capacity coming online.
• Prepayments versus backlog: $9 billion is about 7% of the $132.5 billion backlog.
• Implied prior-year revenue: $3.66 billion divided by 2.68 is about $1.37 billion for the year-earlier quarter.

Pricing: the part that explains the contract shift

According to TechCrunch, CoreWeave raised prices about 25% in July and a further 10% later in the quarter, and prices are still rising this quarter. CEO Michael Intrator said the increases "more than offset[s] cost increases." Short-term capacity is reported to sell for more than $40 million per megawatt per year, while new infrastructure on four- and five-year contracts is moving to $20 million to $25 million per megawatt.

On those figures, the short-term price is roughly 1.6 to 2 times the long-contract price (our arithmetic: 40 divided by 25, and 40 divided by 20). That gap is the economic reason to allocate more capacity to short-term deals, and Intrator is quoted as saying: "We are also allocating more capacity to short-term opportunities, where the pricing justifies it."

The financing tension

Shorter contracts raise revenue per megawatt but weaken the usual basis for debt. TechCrunch notes that lenders generally prefer longer contracts because they make cash flow predictable. CoreWeave's response, as reported, has three parts:

1. Prepayments and deposits. About $9 billion expected over three months, of which more than $3 billion is tied to signing new customer contracts for capacity already built, which needs no additional hardware spending.
2. Contract-backed loans. Most remaining buildout is to be financed with loans backed by specific customer contracts. In August the company closed a $2.6 billion facility called DDTL 5.5, described as its first designed for shorter-term customer contracts, which lets it borrow in installments.
3. Avoiding unsecured debt at current yields. Intrator said: "We will not use the unsecured debt market to fund our business at current yields, nor do we need to." TechCrunch cites a report that yields on some CoreWeave bonds exceeded 15% last month.

How this compares across neoclouds

TechCrunch notes that rivals Nebius and IREN have also relied on large prepayments from Microsoft. We covered that arrangement in our piece on Microsoft's oversight of those neocloud contracts. The common thread is that customers with the strongest balance sheets are, in effect, funding capacity before it generates revenue for the supplier. The difference in CoreWeave's case, per the report, is that prepayments now sit alongside a deliberate move toward shorter terms.

Two ways to read the same numbers

Reading one: demand outpacing supply. Rising prices, a $132.5 billion backlog and customers willing to prepay are consistent with scarce capacity, where suppliers can set terms. Intrator's comment that price increases exceed cost increases supports this view.

Reading two: financing dependence. Interest expense above the net loss, bond yields reported above 15% and a pivot away from unsecured debt show that funding costs remain a central constraint. Prepayments reduce the need to borrow, but they also tie future delivery obligations to customers' cash.

Both readings can be true at once. The data in this report cannot say which dominates, and a single quarter is a short window.

What to watch next

• Whether the roughly $9 billion of prepayments arrives within the stated three months.
• Whether short-term pricing holds as new capacity comes online across the industry.
• The mix of short and long contracts in future disclosures, since it affects how lenders value the backlog.
• Interest expense relative to revenue in the next quarter.

What it means for teams buying AI compute

For companies that rent GPU capacity, the reported price gap suggests that term length is now a major lever on cost. Teams that can commit for several years pay less per megawatt on these figures, while teams needing flexibility pay more. Application-layer users face this indirectly through model API pricing, which is one reason multi-model tools such as Metir let people shift workloads between providers as prices change.

Sources:

• TechCrunch: CoreWeave turns to customer prepayments, shorter contracts to fund AI growth (Nov 11, 2026)
• Metir: CoreWeave Q2 2026 earnings
• Metir: CoreWeave bond yields and backlog
• Metir: Microsoft neocloud oversight

Image credits

• Nasdaq MarketSite, Times Square, New York (the venue for CoreWeave's CRWV listing; it does not depict this earnings report). Wikimedia Commons, CC BY 2.0.
• Server hall, illustrative. BalticServers.com, Wikimedia Commons, CC BY-SA 3.0.