Hugging Face will distribute ten million dollars of GPU computing power to small AI developers. In this way, the company hopes to make it competitive against the big boys.
Hugging Face reaches into its pockets to support smaller AI developers in building AI applications. Developing such applications can be expensive because it requires valuable GPU computing power. Big players like AWS or OpenAI have access to massive AI supercomputers, but smaller players have to rent that computing power. That’s why Hugging Face will provide $10 million in GPU computing power for free.
Beginners and academics
Computing power is for startups, small developers, and academics. Hugging Face, which emerged as a kind of GitHub for AI and machine learning models, aims to counter the centralization of AI. This is of course to the benefit of Hugging Face itself, which has a central role to play in a large and diverse AI ecosystem.
“We are fortunate to be in a position to invest in the community,” Clem DeLange, CEO of Hugging Face, told The Verge. Hugging Face is now more or less self-funded and valued at $4.5 billion after a new $235 million investment round.
Zero GPU
However, the company fears that the high cost of AI development will make diversification in the ecosystem difficult. The main innovations come from players with many resources, who keep their achievements to themselves. Moreover, it is not easy for small players to purchase the necessary computing power from cloud providers, as the cost of entry is high.
Freely available computing power should solve this problem. Hugging Face calls its ZeroGPU sponsorship software. Computing power will be available through Spaces. The pool of computing power is shared between users and applications, so that ZeroGPU is used as optimally as possible. On the server side, the software runs for Nvidia A100 GPUs.
