Ollama just proved that the Docker playbook works for AI. The company raised $65 million in a Series B led by Theory Ventures, bringing total capital to $88 million. The headcount: 14 people. The reach: 8.9 million developers every month, embedded in 85 percent of the Fortune 500.

Jeff Morgan and Michael Chiang have seen this movie before. They built Kitematic, sold it to Docker, then built Docker Desktop — the tool that made containers feel native on a Mac or Windows laptop. They abstracted away the kernel nastiness, the networking plumbing, the volume mounting misery. Developers clicked an installer and got to work.

Ollama does the same trick for open-weight models. Download. Run. Done. No dependency hell. No CUDA version mismatches. No quantization guesswork. The GitHub stars — 176,000 — tell you developers noticed. The 17,000 forks tell you they're building on top.

The Docker analogy isn't metaphor. It's lineage.

Benchmark's Peter Fenton led the Series A and took a board seat because he recognizes the pattern. "What Jeff and Michael built with Docker is being used by 10 million-plus developers every day. The creative powers to create a product that goes to ubiquity for developers is extremely rare," he told TechCrunch. Rare is the right word. Most developer tools plateau at niche adoption. Docker Desktop didn't. Ollama isn't either.

The business model is straightforward: free local execution, paid cloud GPU time via "neocloud" subscriptions from free to $100 a month. Usage meters by GPU seconds, not tokens. That alignment matters. Token pricing penalizes experimentation. GPU pricing penalizes waste. Developers prefer the latter.

January changed everything

Morgan pinpoints the proving month: January. That's when open models — specifically the emergence of capable coding agents — crossed a usefulness threshold. "Open models suddenly became able to do these agentic tasks, like coding. Obviously, we saw the explosion of the assistants, and this idea that open models can get real work done."

He's referencing the wave of tools like OpenClaw (likely a reference to open-source coding assistants) that turned raw model weights into functional agents. Before that, open models were research artifacts. After that, they became production dependencies. Ollama sat at the distribution layer, ready.

The enterprise math is brutal. Inference costs on closed models scale linearly with usage. At Fortune 500 volume, that's a budget line item that demands a CFO's attention. Open models shift the cost structure: you pay for compute, not per-token tolls to a model provider. The spreadsheet writes itself.

Not a zero-sum game

Fenton pushes back on the binary narrative. "I still think that this is the part that most of the debate gets wrong. It's not an either/or. There will be plenty of business for both." He's right. Closed models still own the frontier — frontier reasoning, frontier context windows, frontier reliability. But the middle 80 percent of enterprise workloads? That's where open models eat margin.

Ollama's position is structural. They don't train models. They don't fine-tune them. They distribute and execute them. That's a platform moat, not a model moat. Platform moats compound. Model moats erode — every open release narrows the gap.

The valuation question nobody will answer

Morgan and Fenton declined to discuss revenue or the new valuation. That silence is its own signal. At 14 people with 8.9M MAU, the revenue per employee would be staggering if conversion follows typical dev-tool funnels. The Series B price likely reflects option value: Ollama becomes the default control plane for open-model inference, local and cloud.

Competitors exist. LM Studio, Jan, text-generation-webui, any number of Electron wrappers around llama.cpp. None have the Docker pedigree. None have the Fortune 500 penetration. None have the distribution gravity that turns a GitHub star count into a de facto standard.

What comes next

The $65M buys time to harden the neocloud, to build the enterprise features that Fortune 500 procurement demands — SSO, audit logs, air-gapped deployment, support SLAs. It buys the luxury of saying no to acquihire offers from the hyperscalers. It buys the chance to become the Docker Hub of open models: the place you push, the place you pull, the place you trust.

Morgan and Chiang have earned that trust once before. The market is betting they'll earn it again. The developers already have.