Supabase Storage Overhaul: 14x Faster Object Listing

Supabase has announced a fundamental re-architecture of its storage engine, delivering listing speeds up to **14.8x faster** for applications managing tens of millions of objects.

As AI-native applications generate massive amounts of unstructured data—from agentic reasoning logs to vector embeddings—the traditional method of listing objects via a prefixes table has become a critical bottleneck. Supabase's new engine solves this by implementing a **Hybrid Skip-Scan Algorithm**, which bypasses redundant index entries and treats the S3-compatible metadata layer like a high-performance relational query.

The Technical Breakthrough: Skip-Scan over B-Trees

In standard Postgres-backed storage, listing files in a folder requires a full index scan of all objects sharing a prefix. For a bucket with 60 million rows, this can take seconds. The new Supabase approach utilizes a skip-scan (also known as a loose index scan). Instead of reading every row, the engine jumps to the next unique prefix in the index, drastically reducing the number of disk I/O operations. This results in sub-100ms response times for folder navigation, even in the most bloated "agent-slop" directories.

Multigres: The 2.5x Parser Speedup

Alongside the storage update, Supabase has transitioned to **Multigres**, a new high-performance Postgres parser. Previous versions relied on cgo-based wrappers for the native C parser, which introduced significant overhead during concurrent requests. Multigres is a pure-Rust implementation that is **2.5x faster** at parsing complex queries. For developers building real-time dashboards or high-frequency agentic monitors, this means lower CPU utilization and higher request throughput across the board.

Log Drains for Pro Plans

In a major win for observability, Supabase has also unlocked **Log Drains** for all Pro plan users. Developers can now export their Postgres logs, Auth events, and Edge Function executions directly to external providers like **Datadog**, **New Relic**, and **Sentry**. This integration is essential for security auditing in the age of agentic AI, allowing teams to monitor for prompt injection attempts and anomalous data exfiltration patterns in real-time.

Conclusion: Infrastructure for the Intelligence Age

The Supabase storage overhaul is a clear sign that the "backend-as-a-service" market is maturing to meet the demands of AI-scale workloads. By solving the object listing bottleneck and providing industrial-grade observability, Supabase is positioning itself as the primary choice for startups that need Postgres flexibility without the traditional scaling "taxes." In 2026, the database is no longer just a place to store data—it is the **high-performance substrate** upon which the world's most capable agents are built.