Open-source code and model platform Hugging Face has announced it will be sharing $10 million worth of GPU compute power with developers for free.
Speaking to The Verge, CEO Clem Delangue said that the idea behind the initative, dubbed ZeroGPU, is to help AI startups to compete with large companies by ensuring they have access to the same state-of-the-art technologies.
Hugging Face recently raised $235 million in a Series D funding round, giving the company a valuation of $4.5 billion.
The shared GPUs will be made available to multiple users or applications concurrently, meaning access is determined by usage. Instead of needing a dedicated GPU, users can simply utilize the exact amount of GPU power necessary for their project, while the capacity that is not actively in use becomes available for someone else.
Delangue said that this could save smaller organizations a lot of money as traditionally, a company would need to enter into long-term and often expensive commitments with cloud providers to secure compute resources, even if their project never gets off the ground or they don’t need the amount of power they’ve paid out for.
ZeroGPU uses Nvidia A100 GPUs, which are cheaper than the more popular H100s but offer approximately half the computation speed. The project will be accessible via Hugging Face’s app publishing hosting application called Spaces.
“If you end up with a few organizations who are dominating too much, then it’s going to be harder to fight it later on,” Delangue told The Verge.
“AI should not be held in the hands of the few. With this commitment to open-source developers, we’re excited to see what everyone will cook up next in the spirit of collaboration and transparency.”