Neocloud CoreWeave (NASDAQ: CRWV) has introduced new capacity tiers designed to give its clients more control over computing resources when demand and use fluctuate.
With these so-called Flexible Capacity Plans, CoreWeave seeks to move the industry beyond the traditional binary choice of reserved or on-demand capacity. This should be particularly useful for clients running inference (e.g., on-demand AI model use), where volatility in demand traditionally forces engineering teams to purchase excess capacity or risk severe processing delays.
The updated consumption framework introduces a Flex Reservations tier tailored for applications scaling unevenly. Customers secure a guaranteed capacity by paying a reduced holding fee around the clock. Clients then pay full usage rates only when instances actively process data.
CoreWeave also launched Spot Instances to accommodate interruption-tolerant background tasks. This tier services non-critical operations like batch analytics and data backfills. The infrastructure provider sells this excess computing power at a steep discount to maximize resource utilization.
These new offerings will complement CoreWeave’s offerings for Reservations, where clients pay for always-available capacity, and On Demand compute.
“At production scale, infrastructure planning becomes as critical as deployment. CoreWeave is setting the standard for the AI cloud by providing guaranteed capacity when it counts and flexible pricing when demand shifts. We’re bringing the original promise of the cloud – scale and efficiency – back to the AI pioneers pushing the limits of innovation,” CoreWeave EVP Chen Goldberg said.
This strategic billing shift follows a massive multi-year infrastructure agreement signed last Wednesday. AI search firm Perplexity inked a deal to utilize CoreWeave as its primary computing provider. Perplexity will run its operations on specialized Nvidia clusters housed within the cloud network.
The new flexible capacity plans target the high-volume inference requirements of organizations like Perplexity. CoreWeave previouslysecured a $2 billion strategic investment from Nvidia in January to support related data center expansions. The firm also expects to reach 5 gigawatts of computing capacity by 2030.
