Ollama raises $88M to make running open AI models as easy as installing software

Ollama, the tool that lets developers download and run open-weight models locally with a single command, has raised $88M in a round led by Benchmark’s Peter Fenton, with Theory Ventures, 8VC, Y Combinator, and a roster of industry angels including Docker founder Solomon Hykes and Cockroach Labs’ Spencer Kimball. Founders Jeff and Michael previously built Kitematic, which Docker acquired in 2015 and turned into Docker Desktop — and they’re pitching Ollama as the same ‘easy and open wins’ bet applied to AI. The company says it now serves 8.9 million developers and is used inside 85% of the Fortune 500.

The product thesis leans on three selling points: ownership (models you can keep and modify rather than rent behind an API), cost (no per-token bills for anything running on your own hardware), and privacy (local inference keeps data on the machine). Ollama is also pushing beyond local execution with a hosted cloud that runs larger open models — GLM, Nemotron, DeepSeek, Kimi, and MiniMax among them — which the company claims is more than doubling in token volume every month.

The framing is deliberately grand, casting open models as AI’s ‘personal computer moment’ and positioning Ollama as the on-ramp. Stripped of the marketing, the significance is real enough: the raise signals that local-and-hybrid open-model tooling is now a venture-backed commercial category, and Ollama is betting the funding on day-one support for new open models, seamless hybrid inference, and a cloud tier that preserves the ownership pitch at scale.