CoreWeave raised prices by 25%, then another 10%, and is asking customers to pay upfront

If your AI product depends on rented GPUs, the familiar calculation of “number of accelerators × operating hours” no longer captures the full budget. The upfront payment, contract length, and price of the next renewal also matter. CoreWeave, which rents out AI computing capacity, is increasingly asking customers to pay in advance and allocating more resources to short, expensive contracts. For founders, the practical implication is clear: infrastructure spending may be required long before a product generates its first revenue.

According to TechCrunch, CoreWeave reported third-quarter revenue of $3.66 billion, up 168% year over year. At the same time, the company posted a $709 million net loss, while quarterly interest expenses reached $846 million. Rapidly growing demand is bringing in revenue, but financing infrastructure is expensive.

Customers pay today and receive capacity under contract

CoreWeave expects to receive around $9 billion in customer advance payments and deposits over the next three months. More than $3 billion of that amount is expected to come from new contracts for capacity that has already been built. This is a particularly attractive source of funding for the company: it does not need to buy additional equipment for those contracts.

Imagine a team that needs a cluster to train a model. With pay-as-you-go billing, spending is spread over time. With a large upfront payment, part of the budget immediately goes to the provider, and the team receives a commitment to supply computing resources on agreed terms. The provider reduces its need to borrow, while the customer ties up its own funds for the duration of the contract.

A second source of funding is loans secured by specific customer contracts. CoreWeave plans to draw funds as needed to build infrastructure for its $132.5 billion order backlog. That figure represents future contractual commitments, rather than money already in the bank.

CEO Michael Intrator said CoreWeave does not intend to raise unsecured debt at current yields. The reason is clear: according to the report, yields on some of the company's bonds exceeded 15% last month. That signals both the high cost of this financing and investor concerns about the company's debt burden.

Short contracts make it easier to raise prices quickly

In July, CoreWeave raised prices by roughly 25%, followed by another 10% later in the quarter. If both increases are applied sequentially to the same rate, the cumulative increase is around 37.5%. However, the report does not provide enough detail to apply that calculation to every customer contract.

The company is also allocating more capacity to short-term deals. According to Intrator, these can generate more than $40 million a year per megawatt of capacity, compared with $20–25 million for new four- and five-year contracts. These figures measure revenue relative to infrastructure capacity; they cannot be directly converted into an hourly price for a specific GPU.

This creates a conflict of interests within the financing model itself. A short contract allows prices to be revised sooner if demand rises. Banks prefer long contracts because they provide more predictable cash flow for loan repayments. In August, CoreWeave secured a $2.6 billion loan specifically designed for shorter customer contracts, with funds available to draw in installments.

What to check before reserving GPUs

For a product team, the starting point is the financial model. Alongside computing costs, consider the upfront payment, the time before usage begins, and the renewal price. Even an attractive rate can prove impractical if prepayment consumes cash needed for development and hiring.

Check the contract for the capacity delivery date, deposit refund terms, consequences of launch delays, and rules governing price changes. Developers should assess the portability of their environment and data in advance: the ability to switch providers affects negotiations over the next contract.

CoreWeave's strategy does not guarantee further price increases. It benefits the company as long as capacity remains in demand and short contracts can be renewed regularly. If demand weakens, that advantage could disappear. The customer's practical task is therefore to model several utilization and rental renewal scenarios before committing a substantial share of the budget upfront.