Open source AI is booming, and Hugging Face CEO Clem Delangue thinks that shift is only getting started. Speaking on TechCrunch’s Equity podcast, as reported by TechCrunch AI, Delangue laid out why the open versus closed source fight is heating up right now, and why he’s worried about a future where a handful of companies control the whole stack.
What stands out here is his front-row seat. Hugging Face has become something like a GitHub for AI, a place where builders share and download open models and datasets. Roughly half the Fortune 500 now use it. That gives Delangue a pattern he says he sees on repeat.
The cost curve pushes companies open
His core observation is simple. Companies start on frontier APIs from the big closed labs. The APIs are easy, fast, and good enough to launch. Then they scale, usage climbs, and the bills climb with it. At that point, open source models start to look a lot more attractive.
This is the trend worth watching. Closed APIs win the first mile because they lower the barrier to entry. Open models win the marathon because you control the cost, the hosting, and the tuning. As more products move from prototype to real volume, expect that migration to accelerate.
Why the concentration worry matters now
Delangue’s bigger fear isn’t closed models themselves. It’s concentration. He’s worried a few large companies could end up controlling everything that matters in AI, from the models to the infrastructure they run on.
That concern lands harder in the wake of Anthropic’s halted Fable release, the news hook for the Equity conversation. When a major lab pulls or pauses a release, it’s a reminder of how much power sits with a small group of decision makers. Open source is Delangue’s counterweight: transparency and optionality instead of dependence.
He’s also putting money behind the belief. According to TechCrunch AI, Hugging Face is choosing capital efficiency over the usual Silicon Valley fundraising playbook, and turned down a large investment from Nvidia last year. In a market where most AI companies are raising as much as they can, that’s a deliberate signal about staying independent.
The China question
One wrinkle Delangue doesn’t dodge: Chinese labs are producing the majority of open models being downloaded in the U.S. He frames that as a problem worth fixing, not a reason to distrust open source itself.
It’s a sharp distinction. The answer, in his view, isn’t to retreat from open models. It’s for U.S. and European labs to ship more competitive open releases so the ecosystem isn’t leaning on one region. Watch for this to become a bigger policy and competitive talking point over the next year.
Robotics raises the stakes
Here’s the part that projects furthest forward. Delangue argues robotics is an even more urgent case for open, transparent AI than chatbots or coding tools. The reason is intimacy. A robot in your home ends up seeing your family, your routines, your private life.
A chatbot that gets things wrong is annoying. A closed, opaque robot with deep access to your home is a different level of trust. As robotics moves from labs to living rooms, the demand for models you can inspect and audit gets a lot louder.
What to do with this
A few practical takeaways for builders and operators:
- Model your cost curve early. If you’re on a frontier API, price out what open source hosting looks like at 10x your current usage before the bill forces the decision.
- Keep your architecture portable. Avoid locking every workflow to one provider’s API so you can switch when economics or policy shift.
- Watch the open model leaderboards, including who’s publishing them. Capability and provenance both matter for enterprise adoption.
- If you’re building anything with physical or private-life access, treat transparency as a feature, not an afterthought.
Delangue is betting the next phase of AI rewards the open and the auditable. For more on the China angle, the Nvidia decision, and his robotics case, the full conversation is on TechCrunch’s Equity podcast.