By Priya Raman

CoreWeave plans to finance most of the capital spending needed to deliver its $132.5 billion revenue backlog through delayed-draw term loans tied to specific customer contracts. The plan follows record third-quarter revenue of $3.7 billion, up 168% year over year, and changes the funding mix rather than the reliance on debt.

CoreWeave, a neocloud, or AI-focused cloud infrastructure provider, is re-evaluating its financing approach as heavy losses and interest costs absorb cash. Its $132.5 billion revenue backlog represents a pipeline of future customer revenue.

Net losses reached $709 million in the third quarter, up from $110 million in the same period a year earlier, while net interest expense, the cost of debt after interest income, reached $846 million. The quarter's record revenue has not yet translated into profit.

What changes in CoreWeave's financing?

CEO Michael Intrator said on Tuesday's earnings call: “Looking ahead, we will not use the unsecured debt market to fund our business at current yields, nor do we need to. [W]e expect to finance the majority of the remaining capex needed to deliver our backlog through delayed-draw term loans matched to specific customer contracts.”

The unsecured debt market covers borrowings not backed by a specific asset. Capital expenditure, or capex, means spending on data centers and related infrastructure. Delayed-draw term loans are credit facilities committed in advance and drawn when needed; matching them to customer contracts ties funding to identified work.

This is financing engineering, not financial retreat. CoreWeave still expects to borrow substantially, but it will reduce reliance on the unsecured market at current yields. The practical effect is a closer link between each major financing decision and the customer revenue it supports.

When CoreWeave went public last March, Intrator told CNBC's Jim Cramer that debt was both an “engine” and “fuel” for the company. Tuesday's plan refines that proposition by matching most remaining build-out borrowing to specific customer contracts.

Can AI demand support the capital build-out?

More than 1.7 gigawatts (GW), units of electrical power capacity, is online, twice the level at which CoreWeave started in January. The company expects 1.8 GW to be online by year-end. That build-out supplies the operating capacity needed to turn reported AI demand into delivered services.

CoreWeave presents the $132.5 billion backlog as evidence of growing AI demand. It must service that backlog to demonstrate the commercial payoff, and doing so requires considerable capital for data centers and related infrastructure. Many industry projects stalled over the past year, while opposition surfaced in this year's midterm elections.

Where is the demand coming from?

More than 75% of CoreWeave's committed AI spending correlates with enterprise customers, while the remainder is largely tied to frontier labs, the companies developing the most advanced AI models. Named customers include OpenAI, Meta, Jane Street, Canva and Perplexity.

The mix supports the logic of financing against contracted work, although the 75% measure describes spending correlation rather than recognized revenue. The disclosed customer list spans major AI developers and smaller enterprise names, giving the contract-matching strategy multiple commercial counterparties to draw on.

What will investors watch next?

Shares see-sawed in after-hours trading on Tuesday and hovered around $49, alongside record revenue and a widening net loss. Investors now have a clearer financing test: whether customer-linked debt can fund delivery without deepening cash strain.

The strategy hinges on execution: secure capital, bring data centers online, convert the backlog into recognized revenue and cover borrowing costs.

The decisive issue is whether that funding can support the build-out without letting the $709 million quarterly loss and $846 million net interest expense deepen cash burn. Intrator has set out a tighter match between debt and customer work, not an end to borrowing.