Supabase expands into analytics and vector storage

This turns Supabase from a simple app backend into a broader data platform that can keep more of a developer’s workload inside one product. Postgres still handles live app data, but Analytics Buckets move large historical datasets into cheap columnar storage for scans and reporting, while Vector Buckets hold large embedding archives for search and RAG, both managed from the same dashboard and connected back to familiar Postgres workflows.1 2 3 4

• Analytics Buckets are built for data that is too large or too scan heavy for the main app database. Supabase stores it in Parquet on S3 with Apache Iceberg metadata, so teams can archive old events, clickstreams, or logs cheaply, then query that history without stuffing it into regular Postgres tables.1 2 5
• Vector Buckets do the same unbundling for AI data. Instead of keeping millions of embeddings inside pgvector tables, teams can store them in S3 backed vector indexes with built in similarity search, while still querying through Supabase clients or from Postgres through a foreign data wrapper.3 4
• The strategic pattern matches what is happening across the database market. Supabase is pushing into analytics and vector storage at the same time that specialist vendors like ClickHouse and turbopuffer are growing quickly, and while warehouse companies have moved upstream through deals like Databricks buying Neon to get closer to application data creation.5 6

From here, the storage layer is likely to become the main way Supabase expands beyond startup CRUD apps into heavier production workloads. If developers can start with app tables, then add cold analytics storage, vector archives, ETL, and better database infrastructure without leaving the product, Supabase gets closer to owning the full lifecycle of application data.5 7