The ideal database for AI agents doesn't exist yet, says Percona CEO

Percona CEO Peter Farkas expressed his belief that the ideal database for AI agents does not yet exist. The rapid growth of AI services has introduced workloads that existing databases were not designed to handle, making the current options inadequate. Farkas cited "iteration-focused workloads" as an example, where an AI agent may make numerous attempts before a human or another AI system selects the best result.

He emphasized that organizations must select technology capable of handling these new workloads without assuming that today's choice will remain the best fit. Farkas did not favor any particular database, stating that "everything needs a database," including AI applications. He acknowledged that while AI will likely use existing databases for now, the best solution for AI or agentic workloads does not exist at this point due to the technology's infancy.

Farkas highlighted the rapid pace of change in the AI market, contrasting it with the traditional slower maturation of database software. He suggested that the AI market must mature before its requirements become clear enough for developers to create a clean-slate database tailored to AI needs. Organizations are still deciding how to incorporate AI into their technology stacks, including whether to build models themselves or use cloud services, and whether to use large language models or smaller models with lighter footprints.

Farkas, who became CEO of Percona in October 2025, shared insights from his experience with FerretDB, an open source project that converts MongoDB queries into SQL for PostgreSQL or SQLite. He admitted that his tenure at FerretDB left him with "PTSD from courtrooms" due to a patent infringement lawsuit filed by MongoDB. Despite the legal challenges, Farkas remains open to extending Percona's open source approach to AI, but he rejects the idea of rebranding existing databases as AI databases.