Every OpenAI product relies on fast, reliable data access—whether logging in, checking Codex settings, or starting a ChatGPT conversation. Each action may require many separate lookups. Slow requests make the product feel slow; failed requests stop it entirely. Habitat is the online storage platform built so OpenAI products can quickly and reliably access needed information.

Habitat now handles more than 70 million requests every second, supporting products used by over 1 billion people each week across almost 40 geographic regions. Two years ago, it started as a simple Python client-side library connected to a single database. Today, it’s a complex distributed system serving more than 500 petabytes of data.

Building at this scale is no easy feat, but what made our situation unique is the unprecedented rate of scaling. We've grown more than 10x year-over-year for the last three years. Habitat evolved: first becoming reliable for mission-critical traffic, then fast enough for global users, and finally operating at massive scale.

This post is the first in a two-part series. We’ll share how Habitat evolved, why we turned it from a library into a service, and how we stretched a Python-based serving stack into a reliable storage platform layer. A future post will detail multi-tenancy reliability, read performance optimization, and scaling our partnership with Azure Cosmos DB.