PostgreSQL 18 is coming soon, and it promises some big improvements for speed and efficiency. The latest version adds features like asynchronous I/O and a new version of UUID, which can help databases run faster and handle larger workloads more smoothly. But despite these upgrades, some experts say it still isn’t quite ready for the AI and analytics work that many companies are now focusing on.

Many developers love PostgreSQL because it’s open-source, flexible, and cost-effective. It’s a solid choice for building modern apps. But when it comes to hybrid systems that combine transactional data processing and analytics — known as HTAP — PostgreSQL still has gaps. HTAP systems are important for AI tasks because they let companies analyze live data in real-time, helping AI systems make smarter decisions. Unfortunately, PostgreSQL 18 doesn’t bring specific features aimed at boosting analytical or AI workloads.

PostgreSQL’s Strengths and Limitations for Analytics

Some industry insiders see potential in PostgreSQL’s ongoing improvements. For instance, Tom Kincaid from EDB points out that each new release, including PostgreSQL 18, expands its analytical capabilities. The recent updates support asynchronous data reads and include optimizer enhancements, which help speed up large queries involving many tables — a common need in analytics. Past releases have added features like Foreign Data Wrappers (FDWs), which connect PostgreSQL to external data sources, and Table Access Methods (TAM), allowing users to tailor how they access data for complex queries.

However, other experts, like Alastair Turner from Percona, believe PostgreSQL still isn’t ideal for heavy analytical work or AI. He notes that the absence of dedicated OLAP features means it’s better to use specialized analytics databases alongside PostgreSQL. Sam Lambert from PlanetScale echoes this idea, suggesting that separating transactional and analytical workloads can prevent resource contention and improve overall performance by using dedicated systems for each.

New Features Boost Transaction Speed but Not AI

While PostgreSQL 18 might fall short on AI readiness, it does introduce notable features that can improve transaction processing. A key addition is asynchronous I/O, or AIO, which uses a Linux interface called io_uring. This lets the database handle multiple input and output operations without waiting for each to finish, reducing latency. Lambert explains that this is especially useful for environments with network-attached storage, where quick data access matters a lot.

However, AIO currently focuses mainly on read operations. Experts say work is ongoing to improve bulk writes and checkpointing, which are important for write-heavy workloads like vehicle telemetry, social media updates, or online gaming. Faster indexing and caching are also supported by an upgrade to UUID version 7, which starts with a timestamp. Since recent UUID 7 values are close together, they’re stored on fewer pages in the database, making caching more efficient and improving overall speed.

Other Improvements and What’s Still Missing

PostgreSQL 18 also brings other useful features, such as an improved Explain command. This helps database administrators and developers understand how queries are executed, making it easier to optimize performance and troubleshoot issues. Additionally, the new OAuth support enhances security by allowing PostgreSQL to integrate more seamlessly with enterprise identity management systems. This reduces the risk of errors and improves access control across organizations.

Despite these improvements, the lack of dedicated tools for analytical and AI workloads remains a concern. Experts suggest that for now, companies interested in advanced analytics or agentic AI — systems that can act independently based on insights — should consider combining PostgreSQL with specialized analytics databases. This approach allows each system to focus on what it does best, delivering better overall performance and capabilities.

In summary, PostgreSQL 18 offers important speed boosts for transactional processing. But for organizations looking to harness AI and large-scale analytics, it’s still missing some key features. As the database community continues to evolve, we’ll see whether future releases will close these gaps or if dedicated solutions will remain the best choice for AI-ready systems.

Inspired by

• https://www.infoworld.com/article/4052185/postgresql-18-to-boost-oltp-performance-but-misses-ai-readiness.html

Sources

• gmpg.org
• postgresql.org
• podigee.com
• podigee.com