Ollama, the open source tool that lets developers run AI models directly on their own computers, has closed a $65 million Series B led by Theory Ventures. Founder and CEO Jeff Morgan told TechCrunch about the raise, which brings Ollama's total funding to $88 million, following a $15 million Series A led by Benchmark's Peter Fenton.
The numbers behind this company are hard to ignore. Ollama now has 8.9 million developers using it every month, a presence inside 85% of the Fortune 500, and 176,000 stars on GitHub. It is doing all of this with just 14 employees.
What makes this story worth paying attention to is not just the funding. It is what Ollama represents: a wave of open source AI projects that started as beloved community tools and are now turning into real, investor-backed businesses. That shift is happening fast, and Ollama is one of the clearest examples of it.
Ollama launched in 2023, right as open-weight AI models started becoming available to the public. The problem was that those models were built for researchers, not everyday developers. Getting one up and running on your own machine was painful. Ollama fixed that, stripping away the friction and getting models running in minutes. Developers responded. Blog posts, YouTube tutorials, and social media recommendations spread the tool across the community organically.
Morgan and his co-founder Michael Chiang are not new to this kind of problem. The two previously built Kitematic, a startup that was acquired by Docker. They then helped build Docker Desktop, the tool that made containers accessible to developers who did not want to wrestle with complex infrastructure. Docker simplified moving apps between machines and cloud environments by abstracting away hardware configuration. Ollama is doing the same thing for AI models. The parallel is deliberate.
That background is exactly what drew Benchmark's Peter Fenton in during the Series A. “What Jeff and Michael built with Docker is being used by 10 million-plus developers every day,” Fenton said. “The creative powers to create a product that goes to ubiquity for developers is extremely rare.”
Beyond the desktop tool, Ollama also runs a cloud service where developers can access larger models that are too big to run on a standard PC. Subscriptions range from free to $100 per month, and billing is based on GPU time rather than token limits, which is a more transparent pricing model for developers who want predictability.
Morgan says the business case for Ollama really clicked around January, when large open models became capable enough to handle real tasks like coding and agentic workflows. That moment changed the conversation about what open models could actually do for companies spending heavily on AI inference. Inference costs, the price of running AI models at scale, are a major line item for any company building AI products. Open-weight models offer a way to bring those costs down significantly.
Fenton pushes back on the idea that open and closed models are in a zero-sum fight. “It's not an either/or,” he says. Closed models from companies like Anthropic will still get used, but every company with high inference costs has what he calls a “vital existential project” pushing them toward open-weight models for their everyday workloads. That is the tailwind behind Ollama's cloud business.
Ollama is far from alone in this space. A broader ecosystem of open source AI infrastructure companies is forming around similar ideas:
- Inferact, which makes vLLM, an open source inference engine
- RadixArk, which makes SGLang, another inference framework
- Alternatives to major open models, including NanoClaw
- Small startups building their own open models from scratch, like Arcee
Not everyone has been thrilled about Ollama's commercial direction. About a year ago, a wave of criticism appeared across developer blogs and social media, arguing that the cloud business was distracting from the free, open source tool. Some writers invoked the concept of “enshittification,” a term used to describe the gradual decline of developer tools once they start chasing revenue.
Morgan's response is straightforward. The cloud service exists because large open models are often too big to run on a personal computer. Helping developers find compute to run those models is an extension of the same mission, not a betrayal of it. Fenton echoes that: “Nothing has changed for the core product that's free on the desktop. There's zero change to the premise that this is the place you can discover and run local models.”
Revenue figures and the company's current valuation were not shared. But with nearly 9 million monthly users, a foothold in most of the Fortune 500, and fresh capital behind it, Ollama has moved well past the stage where anyone can credibly call it just a hobby project.
