Rapidly scaling online storage to serve over 1 billion ChatGPT users
22M requests per second. Here's how OpenAI's Habitat storage platform evolved to serve a billion ChatGPT users without melting.
Why it matters
Infrastructure deep-dive: OpenAI is publishing its storage architecture for serving ChatGPT at scale—a rare window into the compute and backend buildout required to run the world's largest AI application.
The key facts
10 to knowHabitat platform serves 1 billion ChatGPT users
22M requests per second throughput
Evolved from Python library to globally distributed system
OpenAI engineering blog (part one of series)
Published September 2026
Habitat evolved from Python library to globally distributed storage platform
Serves 1 billion ChatGPT users
Handles 22M requests per second
Published as engineering deep-dive (part one of series)
Infrastructure/systems story, not model or product feature
Go to the source
OpenAI Blogopenai.com
Publisher excerpt: Learn how OpenAI evolved Habitat from a Python library into a globally distributed storage platform serving 1 billion ChatGPT users and 22M requests per second.