Scaling PostgreSQL to power 800M ChatGPT users
openai.com
January 22, 2026
12 min read
Summary
PostgreSQL has been a key data system for ChatGPT and OpenAI's API, supporting a user base that has grown to 800 million. The database load has increased by over 10 times in the past year, necessitating advancements in production infrastructure to meet rising demands.
Key Takeaways
- OpenAI's PostgreSQL load has increased by more than 10x over the past year due to the rapid growth of its user base, reaching 800 million users.
- OpenAI successfully scaled PostgreSQL to support millions of queries per second by implementing extensive optimizations and adding nearly 50 read replicas across multiple regions.
- PostgreSQL's multiversion concurrency control (MVCC) creates challenges during high write traffic, leading to issues like write amplification and increased query latency.
- To address write-heavy workloads, OpenAI is migrating these tasks to sharded systems like Azure Cosmos DB and has restricted new tables from being added to the PostgreSQL deployment.
Community Sentiment
MixedPositives
- PostgreSQL's ability to scale effectively for 800M ChatGPT users demonstrates its robustness, making it a reliable choice for large-scale applications.
- The implementation of nearly 50 read replicas while maintaining low replication lag showcases advanced architectural strategies that enhance performance.
- PostgreSQL's flexibility allows companies to grow significantly without immediate architectural changes, which is a major advantage for startups and growing businesses.
Concerns
- The reliance on a single writer for scaling raises concerns about potential bottlenecks, indicating limitations in the architecture under heavy load.
- Questions about replication lag and handling stragglers suggest that there may be challenges in maintaining performance consistency with such a large number of replicas.
Source
openai.com
Published
January 22, 2026
Reading Time
12 minutes
Relevance Score
35/100
🔥🔥🔥🔥🔥
Why It Matters
This page is optimized for focused reading: quick context up top, a clean summary block, and a direct path to the original source when you want the full story.