The economics of Redis’s license shift and the 2026 fallout

Redis Ltd. transitioned Redis from the permissive BSD license to a dual RSALv2 and SSPLv1 model in March 2024. Cloud provider economics drive these shifts. AWS, Azure, and Google Cloud generate billions from managed services without reciprocating code contributions. Redis developers face a 1% conversion problem where only a tiny fraction of users become paying customers. In a scenario with ten million users and 80% cloud deployment, there are eight million potential customers. The vendor captures less than 10% of the total market value their software creates. This discrepancy forces vendors to adopt source-available models to protect revenue. The SSPL requires that anyone offering software as a service must release their entire management infrastructure, including monitoring and orchestration tools, under the SSPL.

The community reacts fast.

This pattern started in 2018 with MongoDB. Cloud providers capture 90% of operational value while contributing zero code. The Open Source Initiative rejected the SSPL because it restricts the field of endeavor for service providers. This rejection means SSPL-licensed software is source-available rather than truly open source. Red Hat and Debian dropped MongoDB after the 2018 change to protect their users from non-OSI licenses. HashiCorp faced a similar issue in 2023 when it switched to the Business Source License.

The Rise of Valkey and Alternatives

The Linux Foundation launched Valkey in April 2024 as a BSD-licensed fork of Redis 7.2.4. AWS, Google Cloud, and Oracle fund the project. Valkey remains the most direct replacement because it maintains the same API and wire protocol. It delivers 37% higher SET throughput and 16% higher GET throughput than Redis 8.0. AWS contributes the Async I/O Threading model to Valkey, which enables nearly 1 million requests per second on 8-vCPU instances. Valkey achieved 1,000+ commits with 150 contributors in its first year. The project reached 19.8K GitHub stars within twelve months of its release.

Valkey uses the BSD license.

KeyDB uses a multi-threaded event loop. This architecture processes requests on all CPU cores. It delivers 2-5x throughput for CPU-bound workloads.

Dragonfly uses a shared-nothing design. Each core owns a specific subset of the keyspace. For users who need extreme performance, Dragonfly provides a shared-new architecture where each core owns its own slice of the keyspace; this setup lets a single process saturate a large machine. Dragonfly uses 80% less memory than Redis for identical datasets. It uses a custom memory allocator optimized for key-value access.

Does this shift ensure vendor sustainability?

You, who manage high-traffic production clusters, must evaluate your memory requirements before choosing Dragonfly.

Decision Criteria in 2026

The industry shifted.

The verdict is clear.

Cloud providers moved to Valkey to avoid licensing fees. AWS ElastiCache and Google Memorystore now support Valkey. AWS ElastiCache users see 20% lower costs when they use Valkey instead of Redis. Companies like MaiCoin achieved 33% lower costs by switching to Valkey for ElastiCache Serverless.

The license change in 2024 hit the community hard because Redis had fifteen years of open source history. Legal departments at many enterprises flagged compliance risks for the SSPL. Under the RSALv2, users cannot commercialize the software or provide it to others as a managed service. Companies face risks when they redistribute binaries or source code incorporating Redis. Any combined work distributing the RSAL/SSPL code must be licensed under the SSPL. This can force a company to re-license its proprietary application using restrictive copyleft terms.

Redis 8 changed the landscape by adding AGPLv3 as a third license option in 2025. This inclusion provides a way to offer OSI-compliant open source software. However, the SSPL and RSALv2 remain available for those seeking source-available options.

Pick Valkey if you need a drop-in replacement for Redis 7.2.

Pick Dragonfly if you want to replace a cluster with a single node.