Ollama, founded by two former Docker employees, closes a $65 million round and approaches 9 million developers

Ollama has closed a Series B round of $65 million, led by Theory Ventures. This was confirmed by founder and CEO Jeff Morgan, who specified that the operation follows the previous Series A of $15 million led by Peter Fenton of Benchmark. The total amount raised by the company now stands at $88 million.

Morgan founded Ollama together with Michael Chiang in 2023. The two are not new to the field: they had previously built Kitematic, a startup that was later acquired by Docker and became the basis of Docker Desktop, now used by over 10 million developers. A resume that convinced Fenton to lead the previous round and join the company board.

The project allows users to download and run open-weight models directly on their computers in just a few minutes, without the technical complications that have hindered adoption since 2023, when these models were more geared towards researchers than developers. The repository on GitHub currently counts 176,000 stars and almost 17,000 forks. Ollama also offers a multi-tier cloud service, from a free plan up to $100 per month, with billing based on GPU compute time rather than on consumed tokens.

According to Morgan, the real test for the business came in January 2026, when the explosion of OpenClaw demonstrated that larger open-weight models could handle complex agentic tasks, starting from code writing. Since then, the company claims that an increasing number of enterprises with high inference costs have begun to shift some of their workloads towards open models, reserving proprietary ones for more targeted use.

However, Fenton dismisses the interpretation of a direct clash between the two business models: "It's not a choice between open and closed," he said regarding the debate, emphasizing that there is room for both approaches in the market. In the official announcement, the company claims to already be the largest developer network in the open ecosystem, with over 67,000 integrations created by the community and agreements with leading model labs and hardware manufacturers.

Not all historical users have welcomed the commercial turn. About a year ago, criticisms circulated that the cloud service would divert resources from the original open-source project, a case often cited as an example of "enshittification" of development tools. Morgan responds that the cloud is simply an extension of the original mission: to help developers find the computing power necessary for larger models, which are too heavy to run on a regular computer.