AI Exposes the Database World in an Awkward Position
Percona CEO Peter Farkas believes that AI has surprised the database sector and that agentic workloads might ultimately require a novel type of database – once someone determines the actual requirements of those workloads. During a conversation with GadgetLad at the Percona Live event in Amsterdam, Farkas shared several candid insights regarding the current state of databases.
Years of Database Knowledge…with AI Intruding
Percona has been in the game for twenty years, backing open-source database technologies such as MySQL, MongoDB, and PostgreSQL. These databases come with a wealth of experience, making them as reliable as your mum’s cherished cookie jar. However, AI services are barging in unexpectedly, presenting workloads these dependable databases weren’t designed to manage, and Farkas believes the present selections aren’t perfect.
AI-Centric Iterative Workloads – What a Commotion!
Farkas highlights these “iteration-centric workloads” where an AI agent might make around 150 attempts before anyone or another system selects the best outcome. “A human wouldn’t run through 150 options just for someone or another AI to pick the winner,” he pondered. Currently, companies must choose technology capable of handling these workloads but shouldn’t risk everything assuming it will be the right match forever.
The Significant Database Quandary with AI
“Every application requires a database,” Farkas emphasized, though from the Percona CEO, that’s as surprising as encountering rain in Newcastle. If the current trendy term is blockchain, NFT, or any other fad, at the core of it all, there’s still a database, and AI is included.
Blindly Moving Toward the Future
As AI advances at a breakneck speed, databases that once evolved at a glacial pace are now confronted with swift transformations. This fast movement leaves these databases rattling about like spilled drinks in a pub brawl. Farkas chuckled when asked when a shiny new database might emerge to adeptly manage AI. “I lack a crystal ball,” he jested. The AI landscape needs to mature a bit before anyone can construct a database around its distinctly defined rules.
Percona’s Agile New Directions
Farkas assumed leadership at Percona in October 2025, following his stint at FerretDB. It appears he acquired a few battle scars from legal battles on the journey. With Percona transitioning to a database-agnostic, multi-technology model, Farkas has realized that various tech communities have their own unique rhythms.
Gleaning from Past Errors
Different communities possess distinct characteristics. For instance, promoting “Percona TDE for Postgres” felt like trying to convince Eskimos to buy ice. They were not interested. Farkas learned they needed to discard some of their MySQL habits when engaging with other communities.
The AI Lock-In Puzzle
Could Percona dive into the AI ocean? “To be honest, AI is a vendor lock-in scenario,” stated Farkas, directing criticisms at services like Gemini and ChatGPT. He believes there isn’t yet a mature, enterprise-ready approach to running open-weight agents, but Percona is exploring what they might contribute.
AI or Not?
“Percona is not opposed to AI,” Farkas affirmed. However, they aren’t in the habit of hastily attaching the AI label to databases like PostgreSQL or MongoDB. They prefer authentic transformation over mere rebranding antics.
Conclusion
No New Database? You Must Be Joking!
So, there you have it, everyone! Percona’s Peter Farkas is out there claiming that AI workloads are shaking things up even more than a night out on the Toon. But don’t expect a new AI database hero to appear just yet. It’s all about patience, much like anticipating a Geordie summer!
