CoreWeave rolls out flexible AI cloud pricing after capex selloff rattled investors

March 10, 2026
CoreWeave rolls out flexible AI cloud pricing after capex selloff rattled investors

New York, March 10, 2026, 4:23 PM EDT.

CoreWeave rolled out new flexible capacity plans on Tuesday, adding Flex Reservations and Spot instances as the AI cloud company pushes harder into inference work, the stage when trained models answer live user requests. Shares were last up about 0.7%. 1

The launch lands at a delicate point. CoreWeave told Reuters on Feb. 26 that 2026 capital spending would reach $30 billion to $35 billion, up from $14.9 billion in 2025, as it races to expand AI data centers and chip capacity, and management warned the faster build-out would put short-term pressure on margins. 2

That makes pricing and usage more than a product issue. Inference demand can swing sharply, unlike more predictable training jobs, leaving customers to either over-book capacity or risk delays. CoreWeave is trying to turn that volatility into a steadier business. 1

The Livingston, New Jersey-based company said Flex Reservations lets customers secure a guaranteed ceiling and pay full usage rates only when instances are active. Its Spot offering sells lower-cost, interruptible capacity for tasks such as batch analytics and backfills; Spot is generally available now, while Flex Reservations is in preview for select customers. 1

Chen Goldberg, CoreWeave’s executive vice president of product and engineering, said “infrastructure planning becomes as critical as deployment” once AI services hit production. The company framed the new options as a way to match cost and certainty to real demand rather than a fixed 24/7 reservation model. 1

Steven Dickens, chief executive and principal analyst at HyperFrame Research, said the shift targets the cost of graphics processing units, or GPUs, as CoreWeave moves beyond training frontier models and into enterprise deployments, where the “buying characteristics change.” He added that the company still has to prove the model against Amazon Web Services, Microsoft Azure and Google Cloud as those providers widen their own inference offerings. 3

CoreWeave has said demand is not the problem. Reuters reported its revenue backlog swelled to $66.8 billion at the end of 2025 from $15.1 billion a year earlier, while CFO Nitin Agrawal said most of this year’s capital spending was tied to already signed customer contracts. 2

But the balance between growth and returns remains tight. AJ Bell’s Russ Mould told Reuters after the earnings report that investors understood the rush to expand, but were still worried about the long-term economics and how the company would fund the build-out. CEO Michael Intrator also told Reuters the faster expansion would create “some short-term pressure on the margins.” 4

So Tuesday’s rollout carries more weight than a routine feature launch. If CoreWeave can use finer pricing to keep expensive GPU fleets busier and pull in more enterprise inference work, it may ease some of the pressure created by that spending plan. If not, the same capex questions will come back fast. 1