Redis evolving into real-time data platform

Redis is trying to turn a point product into a larger data path platform. Instead of getting paid only when an application reads or writes from Redis, it can now charge earlier in the workflow, when a company copies records out of Postgres or Snowflake, reshapes them into Redis objects, and keeps them fresh for apps, agents, and models. That makes Redis look less like a cache or serving layer, and more like part of the data plumbing itself.

• RDI moves Redis one step upstream into ingestion. It uses change data capture to watch a source system for inserts and updates, then continuously maps those changes into Redis data structures. That removes the usual custom glue code teams build to keep a fast read store in sync with a slower system of record.
  1 redis 2 redis 3 sacra
• The Decodable deal added the missing connector and stream processing layer. Decodable was built around real time pipelines, so the acquisition gave Redis a faster way to ingest data from many systems instead of waiting for customers to push data into Redis themselves.
  2 redis 5 sacra
• Feature Form pushes Redis into ML operations. A feature store is the system that defines model inputs, tracks which version was used, and serves the latest values at prediction time. Redis already had fast online serving, and Feature Form adds the control plane for managing those features end to end.
  4 redis 6 techtarget

The next step is a bundled real time stack where Redis sells ingestion, transformation, feature management, and serving as one system for AI and operational workloads. If that bundle lands, Redis will compete less like a standalone database and more like a lighter weight alternative to stitching together Fivetran, streaming tools, feature stores, and a separate low latency serving layer.