HuggingFace cofounder: open models on the Hub hold ~$1B in training compute, up from $40M
Thom_Wolf · x · 2026-10-18
HuggingFace cofounder Thom Wolf shares "napkin math": open models from Meta, NVIDIA, Alibaba, DeepSeek, Moonshot, Mistral and 30 other labs on the Hub represent close to $1B worth of training compute, free to download — up from under $40M at the end of 2023.
He analogizes to Wikipedia being written in volunteer hours: "This AI age commons is written in GPU-hours and data." He notes it's a floor — about half from official compute numbers, the rest educated guess, excluding 900k community fine-tunes. An explorable Space is linked.
Related event: Hugging Face Co-founder Estimates Open-Model Compute at ~$1 Billion(3 posts)→
More from Venture
- Founder maps 3 money-making software areas in the AI era, from RL robots to agent maintenance — MarcJSchmidt · 2026-10-19
- Disney agrees to hire most of Character.AI's staff after bankruptcy — TechCrunch AI · 2026-10-19
- How to Build Your Own Algorithmic Trading Analyzer with Claude Code in 5 Steps — mdancho84 · 2026-10-19
- 41 software brands, 5 ways to charge for AI—and only 4 bill per outcome — Small_Luck8177 · 2026-10-19
- Team launches India's first equity research MCP for Claude covering 5,600+ NSE/BSE stocks — viewmind · 2026-10-19
- Nscale discloses going-concern warning as Microsoft pulls out of UK data center deal — SumitGup · 2026-10-19