Ollama just landed $65 million to keep developers running AI models on their own machines instead of paying OpenAI or Anthropic by the token.
The company confirmed a $65 million Series B led by Theory Ventures on July 9, according to TechCrunch, pushing its total funding to $88 million. Founder and CEO Jeff Morgan told the outlet the round arrives as Ollama's user base has swelled to nearly 8.9 million developers a month.
That growth number is the real story here. Ollama sits inside 85% of the Fortune 500, according to the same TechCrunch report, yet the company runs on just 14 employees. You don't need a headcount in the hundreds to build infrastructure the biggest companies in the world quietly depend on. That ratio, millions of monthly users served by a team you could fit in a single conference room, tells you how much leverage open source distribution gives a small team right now.
Morgan and Michael Chiang started Ollama in 2023. The two had worked together before, building Docker Desktop after Docker acquired their earlier startup, Kitematic. That background shows in what they built: a command line tool that lets you pull an open weight model like Llama or DeepSeek and have it running locally in minutes, no cloud account or API key required. The project has picked up 176,000 stars and close to 17,000 forks on GitHub, the kind of adoption curve investors notice.
Theory Ventures led this round. Benchmark's Peter Fenton led the prior $15 million Series A. Fenton and Morgan wouldn't discuss revenue or the company's new valuation when TechCrunch asked, which is its own kind of data point. A startup with 8.9 million users and Fortune 500 penetration staying quiet on price says the company either doesn't need the marketing bump from a big number or would rather investors argue about it privately.
Here's the thing that makes Ollama different from a typical dev tool raise: you never buy access to a model. You buy a way around needing one. Every prompt you send through Ollama on your own laptop never touches OpenAI's or Anthropic's servers. For a bank, a hospital system, or any company with data it legally cannot ship to a third party, that's not a nice feature. It's the only way they're allowed to use generative AI at all. That's almost certainly why Ollama shows up in 85% of Fortune 500 companies despite selling nothing to most of them for free.
Ollama does make money, though. The company hosts larger, more demanding models on what it calls its neocloud, and sells access through subscription tiers running from free up to $100 a month. That's a modest business model sitting underneath a much larger free user base, not unlike the shape Docker itself had for years before Morgan and Chiang ever left the company.
The $65 million also lands at a moment when venture money is chasing the infrastructure layer underneath the model wars rather than the models themselves. OpenAI, Anthropic, and Google keep releasing bigger frontier systems, but a growing share of day to day AI development happens closer to the metal, on laptops and internal servers, away from any API meter. Ollama's raise is a bet that this local, private inference market keeps expanding even as the frontier labs spend billions chasing the opposite end of the spectrum.
What Ollama does with $65 million and fourteen employees over the next year will say a lot about whether open source infrastructure companies need to scale headcount to match their user numbers, or whether small teams can keep riding distribution this large. For now, Morgan and Fenton are keeping the number that would answer that question, the valuation, to themselves.
Also read: What Is MCP, the Model Context Protocol Powering AI Agents Now • Salesforce Keeps Promoting Anthropic's Claude Tag Inside Its Own Slack • Bank of America extends OpenAI its first loan after months of saying no
