Why the Fortune 500 Is Quietly Ditching Rented AI

A startup with just 14 employees now sits inside 85% of the Fortune 500, and it got there by giving its core product away for free. Ollama’s $65 million Series B, announced on July 9, is the clearest signal yet that the era of simply renting artificial intelligence from a handful of frontier labs is starting to crack. Two of the loudest voices in open-source AI, Ollama and Hugging Face, are now telling the same story: as companies scale, the bill for closed AI APIs pushes them toward models they can download, run, and own.

The $88 Million Signal That Enterprises Want to Own Their AI

Ollama, the open-source tool that lets developers run open-weight AI models on their own machines in minutes, raised a $65 million Series B led by Theory Venture, 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 raised to $88 million. Since launching in 2023, the project has amassed 176,000 stars and nearly 17,000 forks on GitHub, a level of developer adoption that most venture-backed startups never reach.

The pedigree is not accidental. Morgan and co-founder Michael Chiang previously helped build Docker Desktop, landing at Docker after it acquired their earlier startup, Kitematic. Docker made it trivial to move cloud apps between environments by abstracting away messy hardware configuration. Ollama, in effect, did for AI what Docker and Docker Desktop did for cloud: it turned a painful, expert-only setup into a one-command experience.

From researcher toy to enterprise default

“Open models started coming out in 2023 but they were really hard to use,” Morgan said, noting they were built for researchers, not working programmers. “As a result, it was really hard to get them up and running.” Three years later, Ollama is “used by over 8.9 million developers every month, sitting in 85% of the Fortune 500 and growing like crazy,” he said, all with a team of only 14 people. Its cloud tier hosts larger models that are too big to run locally, priced from free to $100 per month, and billed by GPU time rather than token limits, a subtle jab at the metering model of the closed API giants.

Why the Bill, Not the Benchmark, Is Driving the Switch

The strategic case for open models is no longer about raw capability. It is about economics. Hugging Face CEO Clem Delangue, whose platform has become something like a GitHub for AI and is now used by roughly half the Fortune 500, says he has watched the same pattern repeat: companies start out on frontier APIs, but as they scale, the costs push them toward open-source models they can host themselves. Speaking on TechCrunch’s Equity podcast, Delangue framed the open-versus-closed fight as a question of who ends up in control, warning about a future where a handful of big companies control everything.

Fenton, who joined Ollama’s board, is blunt about the math. “It’s not an either/or,” he said of open versus closed AI, arguing there will be plenty of business for both. But every company with high inference expenses, the recurring cost of actually running the models, has what he calls a “vital existential project” pushing it “to open-weight models.” In practice, that means enterprises increasingly reserve premium closed models, like Anthropic’s, for as-needed tasks while routing high-volume workloads to cheaper open alternatives.

The January inflection: when open models learned to do the work

Morgan pinpoints the business turning point to around January, when open models “suddenly became able to do these agentic tasks, like coding.” Once open-weight systems could reliably power coding assistants and multi-step agents, the idea that they could do real work stopped being theoretical. That shift is what turned self-hosting from a cost-saving experiment into a credible production strategy for both deep-pocketed enterprises and fast-growing application-layer startups.

A New Open-Source Supply Chain Is Being Funded Into Existence

Ollama is not an outlier. It is one node in a fast-growing ecosystem of open-source AI projects that are turning into venture-backed companies. There are open-source inference providers like Inferact, maker of vLLM, and RadixArk, maker of SGLang, that squeeze more performance out of open models. There are agent frameworks like the open OpenClaw and its lighter-weight alternatives such as NanoClaw, whose creator turned down a $20 million buyout to raise a $12 million seed instead. And there are tiny teams building their own open models from scratch, like Arcee. Layer on Hugging Face as the distribution hub, and the makings of a full open-weight supply chain, from model to serving to tooling, come into view.

The catch: control, and the “enshittification” worry

The movement has tensions. About a year ago, some developers accused Ollama of “enshittification,” arguing its paid cloud business was pulling focus from the beloved free desktop project. Morgan frames the cloud service as an extension of the mission, helping developers run open models that are simply too big for a laptop, and Fenton insists “nothing has changed for the core product that’s free on the desktop.” The deeper anxiety is the one Delangue raises: even open ecosystems can concentrate power if a few players own the models, the compute, and the distribution. Open weights are a hedge against that outcome, not a guarantee against it.

The Bottom Line for Builders and Buyers

For developers, tools like Ollama mean you can prototype against a capable open model on your own hardware today, with no API key and no per-token meter running in the background. For executives, the signal from $88 million of funding and 85% Fortune 500 penetration is that “own versus rent” is now a real line item, not a philosophical debate. The practical move is not to abandon closed models, which still lead on the hardest frontier tasks, but to audit where your inference spend actually goes and ask whether an open-weight model, run on Ollama or a provider like vLLM, could carry the bulk of it at a fraction of the cost.

Related

• Microsoft Quietly Swapped OpenAI and Anthropic for Its Own Models
• Meta Enters the AI Coding Battle With Muse Spark 1.1
• SpaceXAI Releases Grok 4.5, an “Opus-Class” Model
• Apple Sues OpenAI Over Alleged Trade Secret Theft
• Anthropic’s New Claude Feature Is Quietly Selling You on AI

Sources

1. Julie Bort, “Popular open source AI developer tool Ollama raises $65M, grows to nearly 9M users,” TechCrunch (July 9, 2026)
2. Theresa Loconsolo, “Hugging Face’s CEO on why companies are done renting their AI,” TechCrunch (July 10, 2026)
3. “Why the rise of open-source AI isn’t hurting Anthropic yet,” TechCrunch (July 7, 2026)