Ollama has raised a $65M Series B led by Theory Venture, bringing total funding to $88M for the open-source tool that lets developers run open-weight AI models locally in minutes. The round follows a $15M Series A led by Benchmark's Peter Fenton and lands as Ollama reports 8.9M monthly developers and a footprint inside 85% of the Fortune 500. The company runs on 14 employees.
Ollama launched in 2023 to solve a specific pain: open models were shipping fast but were built for researchers, not programmers. Founder and CEO Jeff Morgan and co-founder Michael Chiang wrapped the setup, quantization, and serving pieces into a single command-line tool. The project has since collected 176,000 stars and nearly 17,000 forks on GitHub, numbers that put it in the top tier of developer tools by community engagement.
Morgan and Chiang came to the problem from Docker Desktop, which they helped build after Docker acquired their previous startup Kitematic. The parallel is deliberate: Docker abstracted away hardware configuration for cloud apps, and Ollama does the same for local model inference. Morgan describes Ollama as doing for AI what Docker Desktop did for cloud.
“Open models started coming out in 2023 but they were really hard to use.”— Jeff Morgan, Ollama founder and CEO
Key facts
- 01Ollama raised a $65M Series B led by Theory Venture, bringing total funding to $88M after a prior $15M Series A led by Benchmark.
- 02The open-model runner is used by 8.9M developers every month and sits inside 85% of the Fortune 500.
- 03The GitHub repo has amassed 176,000 stars and nearly 17,000 forks since launch in 2023.
- 04Ollama runs with just 14 employees and offers cloud subscription tiers from free to $100/month, billed by GPU time rather than tokens.
- 05Founders Jeff Morgan and Michael Chiang previously helped build Docker Desktop after Docker acquired their startup Kitematic.
The revenue model runs alongside the free desktop product. Ollama hosts larger models on its own neocloud, sold via subscription tiers ranging from free to $100/month, and meters usage by GPU time rather than token count — an unusual pricing choice in a market where per-token billing is the default. Morgan and Fenton declined to share revenue or valuation figures.
Fenton, who led the Series A and sits on the board, points to Docker Desktop's 10M+ daily developer base as the reason Benchmark moved early. Builders who reach that kind of ubiquity in developer tooling are rare, and the Ollama trajectory now tracks a similar arc three years into the project's life.
“What Jeff and Michael built with Docker is being used by 10 million-plus developers every day. The creative powers to create a product that goes to ubiquity for developers is extremely rare.”— Peter Fenton, Benchmark partner
The commercial proof point, Morgan says, arrived earlier this year when open coding assistants like OpenClaw and NanoClaw took off. Larger open models became genuinely useful for agentic work — coding, tool use, multi-step tasks — which pulled enterprises toward serving them in-house rather than paying per-token rates to closed providers. That shift makes Ollama's cloud business a direct beneficiary.
Fenton pushes back on the binary framing of the open-versus-closed debate. He argues both markets will grow, but every company with meaningful inference spend now has a serious internal project to migrate workloads to open-weight models. Anthropic and other closed labs remain the choice for frontier tasks; open models are increasingly the default for volume.
Ollama sits inside a wider wave of AI-native open-source projects turning into venture-backed companies. That list includes Inferact, the maker of vLLM; RadixArk, the maker of SGLang; the coding assistant OpenClaw and its alternative NanoClaw; and Arcee, which is training its own open models from scratch. Each is solving a different slice of the open-model stack, and Ollama has positioned itself at the developer-facing top of that funnel.
Not every user has welcomed the cloud pivot. About a year ago, a wave of blog posts and social media threads accused Ollama of enshittifying the free project by directing attention toward its paid neocloud service. Morgan frames the cloud as a natural extension of the mission, noting that the largest state-of-the-art open models are too big to run on a laptop, and finding compute for them is part of the same job.
The bigger read here is about where value accrues in the open-model stack. Open weights are a commodity — Meta, Alibaba, Mistral, and DeepSeek all ship them for free — but the tooling that makes those weights actually runnable at scale is not. Ollama's 8.9M-developer distribution advantage is exactly the kind of moat that turns a free tool into a durable business, and the $88M in total capital gives it runway to build the neocloud into a serious inference layer before the hyperscalers commoditize that too.
Working on something we should cover, or seeing a story we missed? Send leads, documents, or feedback to hello@aichatdaily.com. For sensitive tips, see our secure tips page for Signal and PGP options.
Spotted an error? Email hello@aichatdaily.com with the URL and the issue, or read our full corrections policy.
