CoreWeave funds its AI buildout with customer prepayments and contract-backed loans

A new financing playbook for the AI boom

CoreWeave, the cloud provider that rents out computing capacity for AI workloads, is overhauling how it pays for its expansion. According to TechCrunch, the company's third-quarter results, reported Tuesday, laid out a strategy built on three pillars: getting customers to pay up front, borrowing against individual contracts, and signing shorter deals that command higher prices.

CEO Michael Intrator told investors the company has no intention of using the unsecured debt market to fund the business at current yields, and does not believe it needs to.

The scale explains why the financing question matters. TechCrunch reports that CoreWeave brought in $3.66 billion in quarterly revenue, up 168% year over year, while posting a $709 million net loss and $846 million in interest expense for the same period. The company also carries a $132.5 billion backlog of committed business, and much of the infrastructure needed to serve it has yet to be built.

Customers as lenders

A central piece of the new approach is prepayments. CoreWeave expects roughly $9 billion in customer prepayments and deposits over the next three months, TechCrunch reports, including more than $3 billion tied to new contracts for capacity that has already been built and therefore requires no additional hardware spending.

This is not unique to CoreWeave. Competitors Nebius and IREN have similarly leaned on large Microsoft prepayments to pay for computing capacity, which TechCrunch reads as a broader industry shift toward making big tech customers finance the buildouts they depend on.

The tradeoff is pricing. Money accepted today locks CoreWeave into delivering services at rates that could look unattractive if demand for compute keeps climbing.

Shorter contracts, higher rates

CoreWeave is simultaneously betting that it can charge more by shortening contract lengths. Intrator said the company is allocating more capacity to short-term opportunities where the pricing justifies it, selling at more than $40 million per megawatt annually. New infrastructure on four- and five-year contracts, by contrast, is being priced at $20 million to $25 million per megawatt.

The company has already raised prices across its offerings by about 25% in July and a further 10% later in the quarter, with additional increases this quarter that, according to Intrator, more than offset rising costs.

To fund this mix, CoreWeave closed a $2.6 billion loan in August, its first designed specifically for shorter-term customer contracts. The facility, known as DDTL 5.5, lets the company draw funds in installments as needed. TechCrunch points out the tension here: lenders tend to prefer longer contracts, even at lower prices, because predictable cash flow is what they weight most heavily when assessing repayment.

Bondholders stay skeptical

Equity markets reacted positively to the earnings, but bond investors remain wary. Yields on some CoreWeave bonds rose above 15% last month, TechCrunch reports, a signal of concern about the debt load being taken on to build data centers before those investments start generating returns. TechCrunch says it has contacted CoreWeave for further details on the loan facility.

Why it matters

CoreWeave's pivot shows how the AI infrastructure buildout is actually being financed: less through conventional corporate borrowing, more through a web of customer cash, contract-secured loans and pricing bets. Done well, this lowers borrowing costs and extracts more revenue from infrastructure already in place. But it concentrates risk. Prepayments lock in today's prices, short-term deals depend on renewals, and lenders discount exactly the kind of contracts CoreWeave now favors. The whole structure rests on demand and pricing power remaining strong, which is still unproven. If compute prices fall or customers balk at renewal, contracts signed at premium rates today could become tomorrow's liabilities, making CoreWeave's balance sheet a live test of whether AI demand is as durable as the investment behind it.