OpenAI says Habitat handles more than 70 million requests each second and over 500 petabytes of data. A new Rust service uses less computing power, but reliable limits and simple data paths made the scale possible.
New to this? Read it in simple words
- OpenAI explained Habitat, the storage system behind ChatGPT, Codex and other services. OpenAI says it handles more than 70 million requests each second.
- OpenAI says two engineers rewrote it in Rust, a programming language, using Codex and GPT-5.5. OpenAI’s own tests show six times better processor use and fifteen times better memory use.
- The new Rust service uses less computing power. But reliable limits and simple data paths made this scale possible.
- The test results come from OpenAI, with no full outside test. OpenAI says Habitat is reliable but does not publish the full record.
- Programming language
- A special language that people use to write instructions for computers.
- Processor
- A computer chip that follows instructions and does calculations.
One layer now sits between products and stored data
OpenAI says Habitat handles more than 70 million requests each second. It supports products used by over one billion people each week across almost 40 regions. The system serves more than 500 petabytes of data. One petabyte is one million gigabytes.
Habitat began as a small Python library connected to a database. As more teams used it, safe changes became hard. OpenAI moved the common work into a service that can control routing, access, encryption, caching, and regional data rules in one place.
The company says products use several storage systems behind Habitat. The layer chooses where a request should go and limits the pressure on systems below it. This can stop one busy service from causing a wider failure.
OpenAI reports more than 70 million requests each second and 500 petabytes of data. These are company scale figures.
The Rust rewrite saved resources, but design came first
OpenAI says two engineers rewrote the service in Rust with help from Codex and GPT-5.5. Rust is a programming language built for fast and safer system software. The new version now handles 95 percent of production requests.
Company tests show six times better use of processor power and fifteen times better use of memory than the Python version. These are OpenAI’s own measurements. The post does not publish a full outside test or every delay and failure result.
The team delayed the rewrite while it solved more urgent problems. That choice matters. A new language cannot fix unclear ownership, too many complex requests, or a weak recovery plan. The article shows many smaller controls that came before the rewrite.
Scale claims need reliability evidence too
Traffic and storage numbers show size, not service quality. Readers would also need failure rates, slow-request results, recovery times, and regional incident data. OpenAI says Habitat is reliable, but it does not publish that full record in this article.
Data rules are equally important. Habitat includes controls for access, encryption, isolation between customers, and data location. A central layer can make these rules consistent. It can also become one important system that many products depend on.
The engineering report is useful because it explains hidden work behind an AI product. The main achievement is not simply changing Python to Rust. It is creating clear paths, strict limits, and shared controls before growth turns a small problem into a global one.
Sources
Every fact in this story comes from the sources below. Open them to check our work.
- 1Primary source · September 11, 2026Rapidly scaling online storage to serve over 1 billion ChatGPT users OpenAI
- 2Research · September 11, 2026OpenAI’s Habitat platform turns a Python shortcut into a two-engineer Rust rewrite SXZ
We used OpenAI’s engineering report for the design, scale, and efficiency numbers. We used an outside technical summary to check our reading. All performance figures remain company claims because no independent system test was available.