Ollama, the tool that has become the default way developers run open-weight AI models on their own machines, raised a $65 million Series B led by Theory Ventures, with Benchmark, 8VC, Y Combinator and a roster of other investors and angels participating. The round brings Ollama's total funding to $88 million and comes as the platform has scaled to nearly 9 million developers and more than 67,000 integrations -- a genuinely large distribution footprint for a company that, unlike most of this year's AI mega-rounds, isn't selling compute or a frontier model directly.
The round doesn't stand alone. Together AI closed an $800 million Series C at an $8.3 billion valuation, led by Aramco Ventures with Vista Equity Partners, General Catalyst and Nvidia also participating -- capital aimed at making it dramatically cheaper and easier for enterprises to train and run open-source models at scale, the same underlying thesis as Ollama's developer-facing product, just aimed at enterprise infrastructure rather than individual machines. TwelveLabs, meanwhile, raised $100 million in a Series B co-led by NEA and Naver Ventures, with Amazon participating, extending the open-model tooling thesis from text into video understanding.
Combined, the three rounds represent nearly $1 billion flowing into companies whose core bet is that open-weight models -- not just closed frontier labs like OpenAI and Anthropic -- will make up a durable, growing share of how enterprises and developers actually deploy AI. That's a meaningfully different wager than backing a foundation-model lab directly: these companies profit from open-model adoption regardless of which specific open model wins, in the same way infrastructure providers profited from the cloud computing buildout regardless of which specific application won.
For founders building developer tools or infrastructure, the pattern is a signal that "picks and shovels" plays around open-source AI are attracting top-tier venture capital at scale, not just niche or academic interest -- Theory Ventures, Benchmark, NEA and Nvidia's own venture arm all wrote real checks into this thesis within weeks of each other. For enterprise buyers, growing investment in the open-model tooling layer suggests the build-versus-buy calculus around running your own models is about to get meaningfully easier.
The bear case: open-weight model tooling companies remain structurally exposed to whichever labs choose to open-source their best models, and a shift by major labs toward closed-only releases would directly threaten the addressable market these companies are built around. What to watch next: whether Ollama, Together AI and TwelveLabs show revenue growth that matches their funding momentum, and whether any major lab reverses course on open-weight releases in a way that would reshape the competitive landscape these tools depend on.