On November 20, Hugging Face co-founder and CTO Julien Chaumond said on X that Hugging Face serves almost 1 exabyte of open models and datasets over the internet per month and had to build its own CDN for AI downloads, sharing a live traffic map. Co-founder Thomas Wolf added that the CDN has run since this summer and handles about 420 GB per second through 80,000 requests per second.
Key Takeaways
- ✓Scale: Hugging Face serves almost 1 exabyte (about 1 billion GB) of open models and datasets per month, as disclosed on X by co-founder and CTO Julien Chaumond
- ✓Throughput: co-founder Thomas Wolf says this is about 420 GB per second through 80,000 requests per second, and the in-house CDN has run since this summer
- ✓Why build one: classic CDNs were designed for many small web files sent to human users across many cities; AI downloads are a few huge files, largely pulled by machines, often GPU clusters in a handful of cloud regions
- ✓Design focus: caches tuned for massive objects that can absorb release-day floods and several giant datasets and trending models being pulled at once
- ✓The team published a live traffic map of where models and datasets flow; all figures are Hugging Face-stated, with no third-party audit attached
Key Decision Metrics at a Glance
Turn your technical choice into a development budget
Compare 40 dev plans & simulate token costs vs $20/mo subscriptions
Project Links & Resources
Direct AccessIn-Depth Technical Analysis
Background
Open models keep getting bigger, often hundreds of GB or even over a TB, and every hot release triggers a flood of downloads. On November 20, Hugging Face co-founder and CTO Julien Chaumond said on X that Hugging Face serves almost 1 exabyte of open models and datasets over the internet per month, a scale at which it had to build its own CDN for AI downloads.
How it works
Co-founder Thomas Wolf explained why on X: classic CDNs were designed to ship many small web files to many people, caching small objects close to users in hundreds of cities. AI downloads are the opposite: a few enormous files, a large share pulled by machines, with many clients being GPU clusters in a handful of cloud regions. A cached web page serves a thousand visitors, while a cached 1,000 GB model serves far fewer clients, each wanting all of it immediately. So the new CDN uses caches adapted to massive objects and is built to absorb release-day floods and several giant datasets and trending models being pulled at once.
Numbers that matter
There are no benchmarks here. Hugging Face's figures: almost 1 exabyte per month, about 420 GB per second through 80,000 requests per second, with the CDN running since this summer. These are company-stated numbers, and the public posts include no third-party audit.
What it means for developers
For regular users, downloading models with huggingface_hub or the hf CLI works the same as before. The team published a live traffic map showing where models and datasets flow. Developers interested in the Hub's transfer architecture can read the Hugging Face blog post Rearchitecting Hugging Face Uploads and Downloads. Teams pulling weights at scale in the cloud should still keep a local mirror or cache in their own region to avoid release-day peaks.
Benchmark side-by-side against alternatives, or calculate monthly token cost vs subscription break-even.
Discussion & Comments
0Sign in to join the discussion
Connect with AI developers to exchange benchmark insights.