Hugging Face raises $100 million: a hundred thousand models on the shelf
On 9 May 2022 Hugging Face announced a $100 million Series C led by Lux Capital, with Sequoia and Coatue participating alongside existing investors. In the same post the company gave the size of its repository: 100,000 pretrained models and 10,000 datasets, more than 10,000 companies using it, and growth from 30 staff to more than 120 in twelve months.
Why it matters
This is the point at which an open model repository became an economic unit of its own rather than a by-product of a library. The numbers in the post show what was being funded: not a model and not research, but the place where a hundred thousand other people's models sit. At the same time the company was leading BigScience, a community of more than a thousand researchers training a multilingual model the post calls the largest open one then in existence.
The named participants: Lux Capital leading, Sequoia and Coatue with major participation, and existing investors Addition, a_capital, SV Angel, Betaworks, AIX Ventures, Kevin Durant, Rich Kleiman and Olivier Pomel. The record carries medium confidence deliberately. Its only source is the company's own announcement about its own money, and the project's rule asks for a second independent source wherever a claim is quantitative. A securities filing does not help here: the filing list for Hugging Face Inc. holds only two Form D notices, from 2018 and 2019, and none for 2022. What this record does not assert. It gives no post-money valuation: the post contains no line about $2 billion, and that figure is a press number. The record does not assert that this round paid for training BLOOM: the post names BigScience among the things the company leads, not among the things the money will fund.