How OpenAI Scaled Postgres to 800 Million Users (Step-by-Step)
Get the source code for this video here → https://the-dotnet-weekly.kit.com/ef-...
Get the 2026 .NET Developer roadmap here → https://the-dotnet-weekly.ck.page/202...
Want to master Clean Architecture? Go here: https://dub.sh/clean-architecture
Want to master Modular Monoliths? Go here: https://dub.sh/modular-monolith
Scaling PostgreSQL to power 800 million ChatGPT users
https://openai.com/index/scaling-post...
In this video, I break down how OpenAI scaled Postgres to 800 million users — step-by-step — using a visual walkthrough in Excalidraw.
Postgres isn’t the “problem.” The problem is everything around it: connection management, read scaling, replication strategy, caching behavior, and how you protect the database when traffic spikes.
Here’s what you’ll learn:
Why OpenAI kept Postgres (and where the real bottlenecks appear)
The shift to 1 primary + a large read-replica fleet
Why PgBouncer (connection pooling) becomes non-negotiable at scale
How WAL + replication works in practice, and what replication lag breaks
Cascading replication and why it matters when replicas explode in count
Caching done right: TTL strategy + stampede protection (locks)
Geo routing / traffic shaping to keep reads stable and close to users
Rate limiting and guardrails that prevent database meltdowns
If you’re building systems that are starting to feel “too big for one database,” this video will give you a practical mental model for what to scale first, and what to avoid.
Check out my courses:
https://www.milanjovanovic.tech/courses
Read my Blog here:
https://www.milanjovanovic.tech/blog
Join my weekly .NET newsletter:
https://www.milanjovanovic.tech
Chapters