OpenAI Announces Major Scalability Improvements for Habitat Storage Platform
OpenAI has upgraded its Habitat storage platform to handle the unprecedented growth of over 1 billion ChatGPT users, migrating from Python to Rust for significant performance gains.
OpenAI has announced significant improvements to its Habitat storage platform, designed to manage the rapid growth of its user base, particularly for ChatGPT. The platform now handles over 70 million requests per second, supporting products used by over 1 billion people weekly across almost 40 geographic regions.
Habitat, initially a simple Python client-side library connected to a single database, has evolved into a complex distributed system serving more than 500 petabytes of data. In Q2 2026, OpenAI successfully rewrote the entire service in Rust, with just two engineers, to handle the massive scale and demand. The new Rust service is now managing 95% of production requests, showing 6x CPU efficiency and 15x memory efficiency compared to the Python version, alongside significantly lower average and tail latencies.
OpenAI plans to share more insights and learnings from this transition in a future blog post. The company also highlighted the importance of optimizing the database layer, using Azure Cosmos DB, to support the vast data requirements of its services.
This upgrade marks a critical step in OpenAI's ability to scale its infrastructure to meet the demands of its growing user base and expanding product offerings.
Source: openai
