MariaDB buys GridGain to cut latency for AI inference workloads

Relational database supplier MariaDB is buying GridGain and its in-memory computing technology to provide higher performance for large AI inference workloads. It aims to provide a sub-millisecond data infrastructure for the agentic era.

GridGain has developed in-memory software technology, with data in memory stored using key-value pairs and distributed across a server cluster. The software was open sourced via the Apache Foundation as Ignite. The company announced GridGain for AI in February last year, which enabled customers to use GridGain as a low-latency data store for real-time AI tasks such as inferencing, accelerating both predictive and generative AI workloads.

MariaDB CEO Rohit de Souza said: "The rise of agentic workloads has placed unprecedented demands on enterprise infrastructure, causing requirements to explode and requiring a level of scale and sub-millisecond latency that traditional systems simply weren't built to handle. By uniting our platform with GridGain's in-memory data grid, we are entering a new weight class. This enables us to provide a high-performance, scalable, open alternative to the rigid lock-in of Oracle and the fragmented complexity of hyperscalers."

MariaDB provides its eponymous relational database in open source community and paid-for enterprise editions and also as a fully managed DBaaS offering, MariaDB Cloud. Its Enterprise Platform 2026 release combined transactional, analytical, and AI (vector) database engines into a single platform supporting retrieval-augmented generation (RAG) pipelines and AI agents with MCP support. MariaDB has thousands of customers and millions of developers, and was acquired by private equity in 2024. 

GridGain was founded in 2007 by CEO Nikita Ivanov and VP Engineering Dmitry Setrakyan, who have both since moved on. Investor Mark Lockareff is the CEO, coming on board in January 2025. The company has raised $52.1 million across seed A, B, and C rounds, the last one for $13.1 million in 2021, and reported good growth numbers in 2023, saying it experienced its 10th consecutive year of strong revenue growth with cash-positive operations. 

The rationale behind the acquisition is that, as enterprises progress up the AI usage ladder, moving from passive chatbots toward agentic AI systems that reason, plan and execute tasks autonomously, their agents require real-time access to massive datasets. MariaDB can provide the huge datasets and GridGain, with its in-memory structure, can provide the speed needed "that eliminates the disk-drive tax on performance."

GridGain CTO Lalit Ahuja said: "Enterprises today cannot afford the latency introduced by siloed data architectures. With MariaDB and GridGain, enterprise customers will get a unified platform that provides them the best of both worlds, performance and scale, without having to give up on durability."

The transaction is subject to customary closing conditions.

Comment 

 Hazelcast and Redis are both GridGain competitors. They may now be having acquisitive glances cast at them by other database suppliers wanting to add processing acceleration to their dataset storage for AI inference. Think MongoDB, SingleStore, and the like.