Ollama's $65M Series B: Is the Free Tool Still Free?

A $65 million bet that “free” holds

Ollama has raised a $65 million Series B led by Theory Ventures, founder and CEO Jeff Morgan told TechCrunch. The round follows a $15 million Series A led by Benchmark's Peter Fenton, bringing the company's total funding to $88 million. Its GitHub repository has amassed 176,000 stars and nearly 17,000 forks.

That raise puts a specific question in front of every developer, PM, and engineering lead who has quietly standardized on Ollama to pull and run open models locally: does a tool built around running models for free, on your own machine, stay that way once it is carrying institutional money?

What the company will and won't say

Morgan and Fenton declined to discuss Ollama's revenue or its new valuation. What they did share, instead, are usage numbers. According to Morgan, Ollama is now “used by over 8.9 million developers every month, sitting in 85% of the Fortune 500 and growing like crazy,” with a team of only 14 employees.

Fenton draws a comparison of his own: “what Jeff and Michael built with Docker is being used by 10 million-plus developers every day,” he says, calling that kind of ubiquity “extremely rare.” Both figures — Morgan's 8.9 million and Fenton's 10 million-plus — come from the company and its board, not from an independent audit, which matters given neither would confirm hard revenue.

From a research toy to the Fortune 500 default

Ollama's pitch has always been about removing friction. Morgan says open models started appearing in 2023, but “were really hard to use” because they were built for researchers rather than programmers — “it was really hard to get them up and running.” Ollama's answer was to package that complexity behind simple commands.

The product now goes further than a local wrapper: developers can browse and run larger, more complex models that Ollama hosts on its own cloud, through subscription tiers from free up to $100 a month, billed by GPU time rather than token limits.

Why investors moved now: the OpenClaw moment

Morgan points to a specific inflection point — not the funding itself — as the moment Ollama proved it was a business. He says the “proving point” arrived around January, when OpenClaw became hot. That's when, in his account, larger open models “suddenly became able to do these agentic tasks, like coding,” triggering the wave of assistants like OpenClaw and, with it, proof that open models “can get real work done.”

Fenton frames the open-versus-closed debate as a false choice: “It's not an either/or,” he says, arguing there is room for both open and closed models to build large businesses.

Ollama's place in a crowded open-inference stack

The raise lands inside a wider build-out of open-model infrastructure, not a vacuum. Ollama's own framing of that landscape names several adjacent players:

That list is less a threat to Ollama specifically than evidence that investors are now pricing the entire open and local-model tooling category, not a single company, as this cycle's infrastructure layer.

The “enshittification” question

Ollama has already survived one round of community backlash. About a year ago, a wave of blog and social media posts accused the company of letting its cloud business crowd out “its beloved free project,” explicitly citing Ollama as an example of dev-tool “enshittification.”

“Nothing has changed for the core product that's free on the desktop. There's zero change to the premise that this is the place you can discover and run local models,” Fenton says.

Morgan's explanation for the neocloud push is narrower than a pivot: the newest, largest open models are often “too big to run on your own computer,” he says, “so we said, ‘Hey, let's help find the compute for that.’” Whether the desktop product stays untouched is now a promise that outside investors, not just users, have a financial stake in Ollama keeping.

Why the free tier matters past hobbyist projects

The stakes here aren't abstract. Separate reporting on enterprise LLM deployments notes that running models inside a private network “sounds straightforward until teams hit GPU bottlenecks, inconsistent inference performance, and data governance concerns” — problems that get more visible once customer data can't leave internal infrastructure. That's the operating backdrop against which local-inference tooling in general gets evaluated by engineering leadership, not just by individual developers testing on a laptop.

Individual developers carry a version of the same dependency at smaller scale. One developer building a desktop AI app for non-technical users — people who will “never open a terminal in their life” — described routing the embedding step of the app's local retrieval-augmented-generation pipeline through Ollama. When infrastructure that specific sits underneath both enterprise rollouts and solo side projects, a pricing or governance change at the vendor layer stops being a minor inconvenience.

For teams doing vendor evaluation rather than writing code, the practical checklist is short: confirm which workloads run on the free desktop binary versus the metered cloud tiers, get GPU-time cost estimates in writing before scaling usage, and treat the lack of a disclosed valuation as a data point, not an omission to explain away. None of that requires abandoning Ollama — it requires treating it like the vendor it now is.

The call: what to do with this, depending on who you are

If you're running open models on your own laptop or homelab, this raise changes nothing today. Ollama's leadership says the free, local desktop tool stays as-is, and $88 million in venture funding doesn't retire that promise by itself — stick with it for what it was built to do.

If you're standardizing production infrastructure on Ollama's hosted, GPU-billed tiers, treat this as a moment to watch, not ignore — and keep vLLM or SGLang on the shortlist as a hedge. Pricing, valuation, and revenue pressure from two new institutional investors are exactly the parts of the business Morgan and Fenton chose not to discuss. A 14-person company serving what it says are 8.9 million monthly developers has little room left to disappoint either side of that bet.