Overview
- The companies finalized a multiyear cloud agreement on Tuesday, Sept. 15, 2026, that gives Perplexity access to Nvidia GB300 accelerators for model training and inference.
- Under the deal Crusoe will host and perform lifecycle management for the GPUs while Perplexity will make its Enterprise Max product available to more than 1,800 Crusoe employees.
- Perplexity sought GB300 chips specifically because they are optimized for large-model work and the contract provides an alternative source of Nvidia capacity beyond AWS and Azure.
- Neither firm disclosed financial terms, leaving the contract’s dollar value unclear even after Crusoe announced a roughly $13 billion cloud deal with Jane Street earlier in September.
- Crusoe’s expansion into rented AI compute raises questions about execution risk given past project setbacks, so observers should watch customer retention, hardware delivery, and how much GB300 capacity Crusoe actually fields.