Hugging Face crosses $100M annual run-rate, CEO says revenue wasn't the priority

Hugging Face CEO Clement Delangue said on X that the company crossed a $100 million annual run-rate, framing it modestly — acknowledging many AI companies are capturing far more revenue, but expressing pride in the milestone and emphasizing that maximizing short-term revenue has never been the priority. He highlighted that Hugging Face stores and serves hundreds of petabytes of models as core to its open-ecosystem mission.

The $100M run-rate is a notable marker for the de facto hub of open-source AI: it shows the open model can sustain a real business (via paid compute, enterprise hub, and inference services) even if it trails the frontier labs' commercial scale. The single-command vLLM server launch on HF Jobs the company shipped this week is exactly the kind of developer-convenience product that converts the free hub's gravity into paid usage.

Strategically, Hugging Face sits at the center of the open-weights momentum that the week's policy debates keep amplifying — as export controls and closed-API gatekeeping make developers nervous about frontier dependency, the open ecosystem and self-hostable models (Mistral OCR 4, DeepSeek variants) become more attractive. Hugging Face's NVIDIA NeMo integration this week further entrenches it as the training and serving hub.

Caveats: a run-rate is not profit, and 'revenue wasn't the priority' is the kind of framing that reads differently to investors than to the community. The concrete fact is the $100M figure. Watch how Hugging Face monetizes inference and enterprise without alienating its open base, and whether the open-vs-closed tailwind accelerates its growth.