CoreWeave reported higher revenue and operating income for the third quarter of 2026 as demand for AI computing capacity supported increases in GPU rental prices. The company said pricing continued to rise during the quarter and into the fourth quarter, giving it more flexibility to direct capacity toward short-term deals or longer contracts.
The results also highlight the financial pressure behind CoreWeave’s expansion. The company is spending heavily on data centers and computing equipment, while debt-service costs contributed to a substantially larger net loss year over year. Its ability to convert tight GPU supply and customer commitments into returns will remain central to the economics of its build-out.
Higher prices for installed GPUs
Chief executive and co-founder Michael Intrator said CoreWeave increased prices across its product offerings by about 25% in July, then raised them by a further 10% later in the quarter. He said prices continued to move higher in the fourth quarter, more than offsetting cost increases.
Intrator also reported sequential third-quarter rental price increases of 30% for Nvidia A100 GPUs and 50% for H100 and H200 GPUs. The A100 began shipping in 2020, while the H100 began shipping in 2022. The pricing gains show that demand can support higher rates even for generations of hardware that predate the latest accelerators.
CoreWeave is balancing longer-term commitments against opportunities to earn more from short-term capacity. Intrator said short-term sales were exceeding $40 million per megawatt annually, while pricing for new infrastructure on four- and five-year contracts was moving to $20 million to $25 million per megawatt. He described long-term contracts as a source of visibility for expansion, while shorter arrangements allow the company to respond to strong market pricing. CoreWeave evaluates the trade-off by considering expected revenue, operating expenses and financing costs.
Renewals and customer commitments
Intrator said about 130,000 GPUs would come off contract and become available for new agreements, with another 170,000 GPUs scheduled for renewal in 2028. At current prices, he estimated the 2027 renewals could add $4 billion in annualized revenue, rising to $10 billion from the 2028 renewals. These are potential incremental annualized amounts, not reported revenue already earned.
He also noted that capacity purchased in advance by customers is helping fund capital spending. CoreWeave’s backlog reached $132.5 billion at quarter-end, 2.4 times its level a year earlier. The company’s results show the scale of contracted demand, but backlog should not be confused with revenue recognized in the quarter.
Intrator said a slowdown in data-center construction across the industry could make installed computing capacity scarcer and keep GPU pricing elevated. He presented that as one possible market condition, while saying CoreWeave’s plan was based on contracted demand, investment tied to customer success and the agreements it is already signing.
Growth comes with financing costs
For the quarter ended in September, CoreWeave reported $3.66 billion in sales, up 2.7 times year over year and 42.2% from the second quarter. Operating income was $207 million, compared with $52 million a year earlier. Net loss, however, widened 6.4 times to $709 million, with debt service identified as a significant burden.
CoreWeave ended the quarter with just under $5.6 billion in cash and equivalents. It spent $14.5 billion on capital expenses during the quarter and $30.7 billion year to date on data-center facilities and equipment. The article’s analysis estimated the company’s debt at roughly $48 billion, underscoring the financing exposure involved in its rapid expansion.
Rising prices and a large backlog provide support for CoreWeave’s investment plans, but they do not remove the need to manage capital costs. The company is trying to use customer prepayments and revenue growth to finance more of its expansion, while continuing to serve demand for AI training and inference. The gap between operating income and net loss illustrates why growth in sales alone will not determine whether that strategy delivers sustainable returns.
