Overview

San Francisco Compute provides short-term access to large-scale GPU clusters for AI training and inference. Clients rent high-end hardware—like H100 GPUs—on a pay-per-use basis with no long-term contracts, enabling bursts of compute when needed. The service is designed for academia, startups, and enterprises that require substantial power for limited periods, with transparent, competitive pricing and strong technical support. The company differentiates itself by offering some of the lowest-priced H100 training clusters in the world, combined with fast, reliable help desk assistance. Its goal is to make high-performance computing affordable and flexible, so customers can scale computing power up or down according to demand.