OpenAI Scales Single Primary Postgresql to Millions of Queries per Second for ChatGPT

OpenAI scaled a single PostgreSQL primary database to handle millions of queries per second for ChatGPT. Using Azure infrastructure with ~50 read replicas, optimized query patterns, and sharded systems for write-intensive workloads, OpenAI maintains low latency across hundreds of millions of users while managing write pressure effectively.