General Compute

Meet General Compute

General Compute supplies dedicated computing capacity for AI inference across GPUs and specialized accelerators. Customers specify their models, traffic patterns, latency targets, and deployment requirements, then arrange capacity under a contract. The service covers hardware sourcing, colocation, power and cooling requirements, system bring-up, and ongoing operations.

Its approach separates parallel prompt processing from token generation where suitable, using different hardware for those stages. Customers can work with bare-metal systems or a managed endpoint, depending on the arrangement. The advertised hardware portfolio spans several accelerator vendors, but availability differs by system and deployment schedule; the announced Cerebras service is planned for early 2027. Pricing and capacity are quoted for the particular term, volume, and site.