Ollama has raised $65 million in a Series B round led by Theory Ventures, with Benchmark, Y Combinator, and 8VC among the participants. The company has now raised $88 million in total since launching in 2023. It makes a free tool that lets developers download and run AI models on their own computers or company servers, cutting out the major AI providers entirely.
The growth numbers are hard to ignore. Usage has doubled since January to 8.9 million developers per month, and the platform adds close to one million new installs every week. All of this has been built by a team of just 14 people.
What actually is Ollama? Think of it this way: most businesses today that want to use AI pay a service like OpenAI or Anthropic every time they run a query, the same way you pay for a phone call by the minute. Ollama is the alternative. You install it, download an AI model, and run it yourself. The AI stays on your own machine or your company's servers. Nobody outside sees your data, and there is no per-use bill.
This matters most for businesses in regulated industries. Ollama does not train on user data, and when running locally, data never leaves the user's own machine. That is why it has found its way into healthcare, finance, and government, where sending client data to an external AI service raises serious legal and compliance questions.
The founders have done this before. Jeff Morgan and Michael Chiang previously built a tool called Kitematic, which Docker acquired in 2015. Their work there became Docker Desktop, a piece of software now used by more than 10 million developers daily. Docker solved a similar problem for cloud software: it made moving applications between different computers and cloud systems easy, hiding all the messy technical setup. Ollama applies the same idea to AI models.
The business model is free at the base level, then paid tiers up to $100 a month for access to larger AI models hosted on Ollama's own cloud servers. Crucially, it charges by computing time rather than by the number of queries, which makes costs more predictable for companies running automated workflows that run long and quietly in the background. Ollama's cloud token volume has more than doubled every month, indicating that the conversion from free to paid is already happening.
The investor thesis behind this round rests on a clear directional call. One board member has publicly stated that open AI models, the kind Ollama runs, will generate the majority of AI usage within 18 to 24 months. The reasoning: companies paying large monthly bills to proprietary AI services have a strong financial reason to shift routine tasks to cheaper open alternatives, while keeping the expensive services for more complex or sensitive work.
For non-technical business operators, the relevance is practical. If your company currently pays per-use fees to an AI service for any task that runs frequently, such as document summarisation, email drafting, or internal Q&A, there is a genuine cost argument to explore running a local model instead. Open AI models are now capable enough for most of that daily, repetitive work. The technical barrier to getting started is lower than it was a year ago, and Ollama is the main reason why.
The honest caveat is that Ollama has not disclosed revenue or its current valuation, and the 85% Fortune 500 figure likely includes plenty of developers who downloaded it out of curiosity rather than running it in production. Converting that broad installation base into steady paying customers remains the real challenge ahead.