OpenAI scales Habitat storage for 1 billion ChatGPT users: 12x faster reads after a Rust rewrite

#What happened

On Oct 23, 2026, OpenAI published part two of its engineering blog series on Habitat, the internal storage platform behind ChatGPT, the API, Codex, and internal services.

• Habitat serves more than 1 billion users across ~40 geographic regions and stores more than 500 petabytes. The Sept 11 part-one post put the scale at over 70 million requests per second, 500+ petabytes, and almost 40 regions.
• The underlying layer is Azure Cosmos DB. OpenAI's largest Cosmos DB account stores tens of petabytes and handles more than 20 million queries per second; OpenAI says that, as far as it knows, no other company has deployed Cosmos DB this large.
• More than 90% of Habitat requests are reads. Median read time worldwide dropped from roughly 120 ms to roughly 10 ms this year — a 12-fold improvement — achieved by keeping data close to users when capacity allows, using a cache, and rewriting the service in Rust (it was originally built in Python).
• The post recounts two outages. In late 2024, a power cut in the main Azure region of one of its largest Cosmos DB accounts revealed that moving the biggest accounts to a secondary region would take several hours — something neither OpenAI nor the Cosmos DB team had tried at that size. Around a month later, moving back to the original region caused another serious outage. Today the Habitat and Azure teams monitor every layer together, from OpenAI's services down to Cosmos DB's physical storage.

#Why it matters

For the next billion users, OpenAI is developing its own exabyte-scale storage system for Habitat's bulkiest data, pairing Rockset's RocksDB Cloud core with Raft-based replication; change data capture feeds isolated Rockset instances for more complex queries. The read path is already 12x faster after the Rust rewrite and global caching, and the storage backend itself is now being rebuilt to carry the next order of magnitude.

Verified Oct 23, 2026.