Ollama, a widely recognized open-source AI platform, has successfully secured $65 million in its Series B funding, with Theory Ventures taking the lead, as reported by founder and CEO Jeff Morgan in a conversation with TechCrunch. This latest funding round builds upon Ollama's previous Series A investment of $15 million, led by Benchmark’s Peter Fenton, bringing the total funding raised to $88 million.
Launched in 2023, Ollama empowers developers to run open-weight AI models directly on their personal computers, enabling swift setup that takes only minutes. The tool has garnered significant acclaim within developer communities, evidenced by its impressive GitHub stats—176,000 stars and nearly 17,000 forks—alongside numerous positive mentions across training platforms, videos, blogs, and social media.
Ollama also offers a way for developers to access larger, more intricate models hosted on its neocloud. Users can choose from several subscription tiers, ranging from free to $100 per month, with pricing based on GPU usage rather than token limits.
Morgan and co-founder Michael Chiang bring valuable experience to the table, having previously developed Docker Desktop. Their journey at Docker began after the company acquired their earlier startup, Kitematic. Docker is known for creating container technology that seamlessly transitions applications between various environments, eliminating hardware configuration challenges. In a similar vein, Ollama aims to simplify AI model utilization for developers, much like Docker did for cloud services.
Morgan highlighted the initial challenges developers faced with the introduction of open models in 2023, which were primarily tailored for researchers rather than programmers. As a result, getting these models operational proved difficult. Fast forward to today, Ollama boasts over 8.9 million monthly users, including 85% of the Fortune 500, all supported by a lean team of just 14 employees.
Peter Fenton's interest in Ollama stemmed from the success he noted with Docker, where the platform has more than 10 million developers engaging with it daily. He emphasized the rarity of the creative ability to develop a product that reaches ubiquity in the developer community.
While Morgan and Fenton opted not to disclose Ollama's revenue or current valuation, Morgan pointed out a pivotal moment for the company occurred earlier this year when OpenClaw gained traction. This marked a turning point for larger open models, which began demonstrating capabilities that enabled them to perform complex tasks such as coding. The rise of these models, especially impactful assistants like OpenClaw, has sparked dialogue regarding the potential shift toward more open models among enterprises and rapidly growing AI startups, as they aim to reduce costs associated with using closed models like Anthropic.
Fenton believes that the ongoing conversation surrounding open versus closed AI models often misses the mark. He posits that both types will continue to coexist, though organizations facing substantial inference costs are compelled to explore open-weight models more aggressively.
The trend of businesses migrating to open models supports Ollama's cloud offering. Furthermore, Ollama is part of a broader movement where AI is fostering a wave of new open-source projects that attract venture capital investment. This includes companies like Inferact, known for vLLM, and RadixArk, which developed SGLang, along with alternatives to OpenClaw.
Not all users of Ollama have embraced the company’s shift towards monetization, with some expressing concerns that the focus on its cloud business detracts from the original open-source project. This feedback has sparked discussions around the concept of “Enshittification” within developer tools.
However, Morgan contends that the cloud service represents a natural progression of Ollama's mission to assist developers in discovering and utilizing AI models. He emphasized that many of the advanced, large open models are typically impractical for individual computers to handle. Thus, Ollama aims to provide the necessary computing resources to facilitate their use.
Fenton reiterated that the core offering—a free desktop product intended for discovering and running local models—has remained unchanged.
