PostgreSQL 18 previews AI-native queries and vector search for Australian startups

PostgreSQL 18, due for release in late 2026 or early 2027, will ship AI-native queries and improved vector search, strengthening the database's grip on the Australian startup stack at a time when AI agents and retrieval-augmented generation are becoming central to product development.

What is new in PostgreSQL 18

PostgreSQL 18 is expected to introduce native vector search capabilities that integrate directly with the pgvector extension, which has been a PostgreSQL extension for several years but has not yet been shipped as a core module. Native vector search would allow developers to perform similarity searches on embeddings using standard SQL syntax, without requiring the pgvector extension to be installed separately. This is particularly significant for startups building AI-native applications, as it removes a dependency and simplifies deployment.

The release will also improve parallel query execution, which is relevant for large-scale analytics workloads. Australian startups that run PostgreSQL on AWS, Azure or Google Cloud can expect better performance for concurrent workloads, with lower latency on complex joins and aggregations. The query planner is being updated to better estimate costs for parallel queries, which should improve performance without requiring manual tuning.

PostgreSQL 18 will also introduce incremental sort, which improves performance for queries that sort large result sets. For startups running analytics dashboards or recommendation engines, incremental sort can reduce query times by an order of magnitude for common query patterns.

Why PostgreSQL dominates the Australian startup stack

Australian startups have gravitated towards PostgreSQL for several reasons. First, the ecosystem: most Australian startup engineering teams are already familiar with SQL, and PostgreSQL's SQL dialect is the most widely taught in Australian universities. Second, the extension ecosystem: pgvector, PostGIS, TimescaleDB and other extensions allow startups to add AI, geospatial and time-series capabilities without migrating to a specialist database. Third, the hosting options: Supabase, Neon, AWS RDS, Google Cloud SQL and Azure Database for PostgreSQL all offer managed PostgreSQL with Australian data residency.

PostgreSQL is now the world's most popular database for startups, used by Stripe, Notion, Robinhood and thousands of other companies that started small and scaled to millions of users. The Australian startup ecosystem is following the same pattern: local companies such as Airwallex, Canva and Atlassian all use PostgreSQL at scale.

AI agents and the PostgreSQL integration problem

The rise of AI agents presents a challenge for PostgreSQL. Agents need to query databases, extract information and write data back to databases in real time. The standard approach is to use an ORM such as Django, SQLAlchemy or Prisma, but these tools are designed for human-written code, not for AI-generated queries. AI agents can write SQL that is syntactically correct but semantically wrong, leading to data corruption or security vulnerabilities.

The solution is not a new database but better tooling. Startups are building middleware that validates AI-generated SQL before it is executed, applies role-based access control at the query level, and logs all queries for audit purposes. Postgres.app and Neon are both working on native AI agent integrations that allow agents to query PostgreSQL through a structured API rather than raw SQL strings.

According to the PostgreSQL 18 features preview, the release is expected to ship in late 2026 or early 2027 with native vector search, improved parallel query execution and incremental sort. For more on Australian enterprise technology, see Tech & Ideas.