What is PostgreSQL 18 and what changed

Quick Answer

PostgreSQL 18, released in September 2025, is the latest major version of the open-source relational database, and its headline change is a new asynchronous I/O subsystem that reduces read latency across cloud and on-premises workloads. It also introduces skip scan index optimizations, virtual generated columns, improved logical replication, and OAuth 2.0 authentication, making it one of the most operationally impactful releases in years.

Introduction

For teams running Postgres in production, version 18 is not a routine bump. The asynchronous I/O layer alone changes how the planner interacts with storage, and the ripple effects touch everything from analytical queries to replica lag. Combined with skip scan support on B-tree indexes, virtual generated columns, and native OAuth 2.0, this release compresses several years of infrastructure wishlist items into a single upgrade. The question for engineering leads in 2026 is no longer whether PostgreSQL 18 is worth attention, but which of its changes will actually move the needle for their stack.

Key Takeaways:

• PostgreSQL 18 introduces an asynchronous I/O subsystem that materially improves read performance on cloud-backed storage.

• Skip scan indexing, virtual generated columns, and improved EXPLAIN output reduce query tuning overhead for engineering teams.

• Native OAuth 2.0 authentication and stronger logical replication make Postgres a more serious fit for multi-region and zero-trust architectures.

Performance and query engine improvements

The performance story in PostgreSQL 18 is built around one architectural shift and a cluster of planner refinements. Together they reduce wall-clock query time on workloads that were previously bottlenecked by storage latency or redundant plan work, which is where most production Postgres pain actually lives.

Asynchronous I/O and skip scan indexing

The biggest change is the new asynchronous I/O subsystem, which lets Postgres issue multiple read requests in parallel instead of waiting on each one serially. On cloud block storage, where per-request latency dominates, this delivers meaningful throughput gains for sequential scans, bitmap heap scans, and vacuum. Skip scan, a long-requested B-tree optimization, lets the planner use multi-column indexes even when the leading column is not in the WHERE clause. AWS engineers have published detailed benchmarks on PostgreSQL 18 performance enhancements showing double-digit improvements on analytical queries. Key wins include:

• Async I/O: Parallelized read requests reduce latency on cloud storage and large table scans.

• Skip scan: Multi-column B-tree indexes stay useful even when leading columns are omitted from filters.

• Self-join removal: The planner eliminates redundant joins on unique keys, cutting execution cost.

• EXPLAIN improvements: More granular buffer and I/O timing data makes tuning less guesswork.

How PostgreSQL 18 compares to prior versions

Version-to-version, the jump from 16 to 18 is larger than most recent releases in operational impact. Postgres 16 focused on logical replication throughput and parallel query refinements, while 17 delivered incremental vacuum and memory management gains. PostgreSQL 18 combines a foundational I/O rewrite with planner features that directly shorten query plans, which is a different class of improvement.

The practical takeaway: teams still on 15 or 16 will see the largest gains, but even 17 users get enough from async I/O and skip scan to justify a planned upgrade. TechBriefed readers evaluating backend framework choices alongside database upgrades should treat Postgres 18 as a serious lever for reducing infrastructure spend.

Developer experience, security, and replication

The second wave of changes in PostgreSQL 18 targets the day-to-day work of application engineers, security teams, and operators running distributed clusters. These are the features that shape what backend code looks like and how safely it scales.

Virtual generated columns, OAuth 2.0, and logical replication

Virtual generated columns compute values at query time without storing them, which reduces write amplification and disk footprint for derived data. Native OAuth 2.0 support means Postgres can now integrate directly with identity providers like Okta, Auth0, or AWS IAM without a middleware shim, which matters for teams building zero-trust architectures or tightening API security practices across services. Logical replication also gets a real operational upgrade: replication slots and subscriptions now survive major version upgrades, removing one of the most painful gotchas in multi-region Postgres deployments. TechBriefed's coverage of PostgreSQL 18 features goes deeper on the migration mechanics, but the operational headline is clear: fewer manual steps, less downtime, and stronger identity boundaries.

Practical implications for engineering teams

For teams running Postgres behind a service mesh or breaking a monolith apart, version 18 lowers several coordination costs. Preserved replication slots simplify multi-region failover, which is a recurring source of pain in microservices architecture. Virtual generated columns reduce the need for materialized views in some analytical patterns, and the improved EXPLAIN output makes it easier to justify or reject index changes in code review. For teams still consolidating their microservices design patterns, a stable Postgres 18 foundation removes one variable from an already complex stack.

Conclusion

PostgreSQL 18 is the kind of release that changes the shape of infrastructure decisions rather than just adding features. Async I/O and skip scan address performance ceilings that teams have worked around for years, while OAuth 2.0 and preserved replication slots close operational gaps that used to require custom tooling. For engineering leads planning upgrades in 2026, the calculus is straightforward: audit workloads sensitive to read latency, review authentication surfaces, and prototype the upgrade path on a staging replica. Independent analysis from outlets like TechBriefed and the official release notes should be part of any serious evaluation before committing to a production migration.

Want sharper analysis on the tools shaping backend engineering? Subscribe to TechBriefed for daily briefings that cut through the noise on databases, frameworks, and infrastructure.

Frequently Asked Questions (FAQs)

When was PostgreSQL 18 released?

PostgreSQL 18 was officially released in September 2025 and is now the current stable major version, as confirmed in the official release announcement.

Is it worth upgrading from PostgreSQL 16 or 17 to 18?

Yes, teams on 16 or earlier will see the largest gains from async I/O and skip scan, while 17 users still benefit enough operationally to justify a planned upgrade within the year.

What are the breaking changes in PostgreSQL 18?

The most notable behavioral shifts involve configuration parameters for the new I/O subsystem and slight changes to EXPLAIN output formatting, both documented in the official migration guide.

Does PostgreSQL 18 support OAuth authentication natively?

Yes, PostgreSQL 18 introduces native OAuth 2.0 authentication, allowing direct integration with identity providers without requiring external proxy layers.

How does async I/O improve query performance?

Async I/O lets Postgres issue multiple parallel read requests instead of waiting on each serially, which reduces wall-clock latency on cloud storage and large sequential scans.

Are logical replication slots preserved during major version upgrades?

Yes, PostgreSQL 18 preserves replication slots and subscriptions across major version upgrades, removing a major operational pain point in multi-region deployments.

Does PostgreSQL 18 change how virtual generated columns work?

PostgreSQL 18 introduces virtual generated columns that compute values at query time without storing them, reducing write amplification and disk usage for derived data.

About the Author

Alex Mercer is a Senior Tech Writer covering databases, developer tools, and backend infrastructure. Alex writes with a data-driven, conversational voice, translating complex engineering shifts into decisions that technical leaders can act on. Their reporting focuses on the operational impact of tooling changes across modern software teams.