Valkey explained: why Amazon’s Redis fork matters for cloud costs

The licensing split created Valkey

Redis Ltd changed the license for Redis in March 2024. This shift moved Redis from the permissive BSD license to the SSPL and RSALv2 models. The Linux Foundation created Valkey in response as a fork of Redis 7.2.4. AWS, Google Cloud, Oracle, Ericsson, and Snap Inc. support this project. Madelyn Olson, a former Redis core maintainer, leads the project. It uses the BSD-3 license. This license allows commercial use and modification without restrictions. Redis and Valkey now have different teams and different roadmaps. Valkey 9 demonstrated high throughput on a 2,000-node cluster, handling over one billion requests per second.

Economics of the Valkey engine

Valkey reduces cloud expenses. For ElastiCache Serverless, Valkey is 33% cheaper than Redis OSS. This reduction applies to both ECPU and storage costs. In the US East (N. Virginia) region, Valkey Serverless storage costs $0.084 per GB-hour plus $0.0023 per million ECPUs, whereas Redis OSS Serverless storage costs $0.125 per GB-hour plus $0.0034 per million ECPUs for a similar workload.

Valkey is cheaper.

Companies switch to Valkey to cut bills. The minimum storage floor for Redis OSS Serverless is 1 GB. This requirement leads to a monthly cost of about $91. Valkey Serverless has a 100 MB minimum. This floor costs roughly $6 per month.

Snap Inc. migrated 70% of its Redis clusters to ElastiCache Valkey in the first half of 2026. This move reduced annual costs from $2.1 million to $840,000. They still serve 5 billion daily requests. The company saved 60 percent. Costs drop fast.

For node-based clusters, Valkey is 20% cheaper than Redis OSS. A cache.r7g.xlarge node in US East (N. Virginia) costs $0.350 per hour with Valkey. The same node costs $0.437 per hour with Redis OSS. When configuring a cluster, ElastiCache recommends reserving 25% of the node memory for non-data use. If you have 50,000 requests per second and 10 GB of data, Valkey Serverless costs $1.25 per hour. Redis OSS Serverless costs $1.86 per hour for the same workload.

Feature divergence and performance

Valkey and Redis remain 90% compatible. Valkey 8.1 runs 8% more operations per second than Redis OSS. It also cuts P99 latency by 22% and uses 20% less memory. Momento benchmarks in June 2026 showed Valkey 8.1.1 reaching 999,800 SET requests per second with 0.8 ms P99 latency on AWS c8g.2xlarge instances. This outperformed Redis 8.0. DragonflyDB also confirmed these figures in August 2026.

Will Valkey eventually match the integrated search capabilities of Redis?

It scales well.

Redis 8 integrates JSON, time series, and vector search into its core. Valkey provides these capabilities via separate modules like valkey-json and valkey-search. Valkey-search lacks several features, including autocomplete, synonyms, and relevance ranking. If your workflow requires these specific search tools, Valkey fails. You should use Valkey for standard caching tasks.