Cloudflare details how D1 scales to millions of small SQLite databases without putting queries on a central hot path

Cloudflare says D1, its serverless SQLite-based database, has grown 15x in live non-deleted databases over the past year and 65x over two years as of October 8, 2026, with daily active databases up 18x year over year. The post explains D1’s architecture for supporting millions of databases by keeping query traffic at the edge, using Durable Objects per database, Workers KV for routing metadata, and Postgres/Hyperdrive for control-plane operations.

For builders planning many small databases per user, tenant, or preview environment, D1’s design suggests separating data-plane queries from centralized control-plane workflows to avoid adding latency to every request. The trade-off is reliance on caching, queues, and eventual consistency for some metadata and background work; the post does not quantify end-to-end query latency or failure behavior under real customer peaks.