Fluidstack

Fluidstack is an AI infrastructure company that provides on-demand GPU clusters and, increasingly, builds and operates entire dedicated data centers for frontier AI labs and large enterprises. Founded at Oxford University in 2017, the company runs a dual business: a GPU marketplace offering dedicated, single-tenant clusters with NVIDIA H100, H200, B200, and GB200 hardware, and a larger private-cloud arm that designs, constructs, and operates custom data-center facilities, including power acquisition, for the largest AI compute buildouts. The company's stated goal is delivering gigawatt-scale compute in around 6 months, versus the roughly 18-24 months typical for the industry, achieved through combined hardware and power-infrastructure teams working alongside software engineering. Its client base includes major AI labs and model developers such as Meta, Mistral, Character.AI, Poolside, and Black Forest Labs, and it is reported to be central to Anthropic's large compute buildout. Fluidstack manages more than 100,000 GPUs across its network and raised an $830 million Series A round in January 2026, valuing the company at $7.5 billion. Pricing is not published, which is typical for infrastructure sold through direct enterprise contracts rather than self-serve signup; the company's own site also does not disclose whether a public API exists for provisioning GPU resources directly.