Labs Shared ~$1B of Training Compute Free on Hugging Face, Up From $40M in 2023
mervenoyann · x · 2026-10-18
Hugging Face co-founder Thomas Wolf's napkin math: Meta, NVIDIA, Alibaba, DeepSeek, Moonshot, Mistral and 30 other labs have shared close to $1B worth of training compute on the Hugging Face Hub—free to download—versus under $40M at the end of 2023. He likens it to Wikipedia being written in volunteer hours, while this AI-age commons is written in GPU-hours and data (a floor: about half from official compute numbers, the rest estimated). HF's ML lead Merlin Nouyann adds that the labeling-cost value of RL environments and datasets should also be estimated, though that's very hard.
Related event: Hugging Face Co-founder Estimates Open-Model Compute at ~$1 Billion(3 posts)→
More from Infra
- Qwen3.8 27B benchmarks on NVIDIA RTX Spark: 19.3 tok/s decode, 100% accuracy — draginol · 2026-10-19
- Gemma 4 12b on NVIDIA RTX Spark laptop: 31.6 tok/s decode, 100% accuracy — draginol · 2026-10-19
- 4x RTX 5090 rig melts a 13A wall socket before the 12VHPWR plug — power-limit your cards — Zo0x78 · 2026-10-19
- NVIDIA shows how to scale model endpoints on Kubernetes with Dynamo in 5 minutes — NVIDIAAI · 2026-10-19
- Mac Studio M5 Ultra beats dual DGX Sparks in local LLM decode speeds — SumitGup · 2026-10-19
- Generating a 5:53 music video on an AMD 7900 XTX: any tips? — wgizmo · 2026-10-19