OpenAI rebuilt its storage platform in Rust with two engineers and Codex

OpenAI has started an engineering series on Habitat, the storage platform that sits behind ChatGPT. It began in mid-2024 as a small Python library so product engineers would not have to deal with databases directly, and it has since grown into a distributed system that handles more than 70 million requests every second, stores over 500 petabytes, and serves products used by more than a billion people each week across almost 40 regions. OpenAI says traffic has grown more than 10x year over year for three years, which forced a long run of tactical fixes to keep squeezing capacity out of what already existed.

The part worth reading is the rewrite. In the second quarter of 2026, a team of two engineers working with Codex and GPT-5.5 rebuilt the whole service in Rust. That version now serves 95% of production requests and is about 6x more CPU efficient and 15x more memory efficient than the Python original, with lower average and tail latency, and the Python code is being retired in the coming weeks. You can read the full write-up for the sequencing decisions behind that.

Why it matters

If you run production infrastructure, this is a concrete sign that a very small team using coding agents can now take on a core-system rewrite that used to need many more people. The efficiency numbers come from OpenAI's own account, so test them against your own workload before assuming a Rust rewrite pays off the same way for you.