Valkey vs Redis in 2026: Why AWS, Google, and Oracle All Bet on the Open-Source Fork

In March 2024, Redis Ltd. quietly rewrote the rules of one of the most widely used pieces of infrastructure on the internet — and within weeks, the community forked it out of existence. Two years later, that fork, called Valkey, has surpassed 100 million Docker pulls, ships as the default in-memory cache on Fedora, Ubuntu, Debian, and Arch, and powers managed services at AWS, Google Cloud, and Oracle. Redis, the project that started it all, is now the outlier.

If you run a cache, a session store, a rate limiter, or a message queue anywhere near production, you've probably already touched this fork without realizing it. This article breaks down what actually happened, how Valkey stacks up against Redis on real benchmarks, what a real company saved by migrating, and whether you should make the switch.

What Happened: The License Change That Started It All

Redis had shipped under the permissive BSD-3-Clause license since 2009 — free to use, modify, and resell, no strings attached. That changed in March 2024, when Redis Ltd. moved the project to a dual RSALv2/SSPL license. According to DevClass's retrospective, the Open Source Initiative doesn't consider SSPL an open-source license at all, and the practical effect was blunt: any cloud provider offering a managed Redis service now needed a commercial deal with Redis Ltd.

The response was fast. Within days, engineers from AWS, Google Cloud, Oracle, Ericsson, and Snap began coordinating a fork, and the Linux Foundation accepted it as an official project by late March 2024. According to The New Stack, the fork started from Redis 7.2.4 — the last release under the old BSD license — and the whole sequence, from announcement to first release, took about two months.

The new project's name: Valkey.

What Is Valkey, Exactly?

Valkey is a byte-for-byte-compatible fork of Redis 7.2, developed under Linux Foundation governance rather than a single vendor. It's a drop-in replacement: the same commands, the same wire protocol, the same client libraries. Most applications can switch by changing a connection string and nothing else.

The governance model is the actual point. As Valkey maintainer Kyle Davis put it, "from this point forward, Redis and Valkey are two different pieces of software" — but critically, Valkey can't be quietly relicensed by one company chasing a new revenue stream, because no single company owns it. That structural guarantee is what convinced AWS, Google, and Oracle to build managed services on top of it instead of waiting to see what Redis Ltd. did next.

It also explains why the backer list reads like a who's-who of hyperscale infrastructure: AWS, Google Cloud, Oracle, Ericsson, and Snap all signed on as founding supporters, and each has since shipped production tooling around the project rather than treating it as a stopgap. Two years is not a long time for a piece of core infrastructure to change hands, get a new governance model, and still gain ground on the incumbent it forked from — but that's exactly what happened here.

Valkey vs Redis: Performance, Pricing, and Ecosystem

Two years in, Valkey isn't just "Redis with a different license" — it's outperforming the project it forked from on several independent benchmarks.

According to AWS, its own internal benchmarks show throughput gains of up to 270% and latency improvements up to 70% in specific workloads, and Valkey 8.1 has become the default engine for new ElastiCache and MemoryDB clusters. Google Cloud has separately declared Memorystore for Valkey 9.0 generally available, and Oracle OCI now ships its own managed offering, OCI Cache.

The adoption numbers back up the benchmarks. Valkey now logs roughly 1 million container pulls a week — over 70 million total since the fork — and AWS has migrated millions of existing ElastiCache nodes onto it. Meanwhile, according to DevClass, Redis Ltd. lost the majority of its external contributors within the first year of the license change — a governance cost that's hard to reverse.

Real-World Migration: How Snap Cut Caching Costs 60%

Numbers on a benchmark page are one thing; production traffic is another. Snap Inc. migrated 70% of its Redis clusters to AWS ElastiCache for Valkey in the first half of 2026, while serving roughly 5 billion daily requests. The result: caching infrastructure spend dropped from $2.1 million a year to $840,000 a year — a 60% reduction — without a rewrite of application logic, because the protocol and client libraries didn't change.

That's the underappreciated part of this story. Migrations of this scale usually come with months of application-layer rework. Here, the savings came almost entirely from swapping the underlying engine and license, not from re-architecting anything.

Should You Migrate? A Practical Checklist

For most teams already running Redis OSS (not the commercial Redis Enterprise features), migrating to Valkey is close to zero-risk because the wire protocol is identical:

# Point your existing Redis client at a Valkey instance —
# no code changes required for standard commands
docker run -d --name valkey -p 6379:6379 valkey/valkey:8.1

# Existing redis-cli and client libraries work unmodified
redis-cli -h localhost -p 6379 SET migration-test "it just works"
redis-cli -h localhost -p 6379 GET migration-test

Questions worth answering before you cut over:

• Are you using Redis Enterprise-only features (like Redis modules such as RediSearch or RedisJSON under their newer licensing)? Those don't ship with Valkey and need a separate compatibility check.
• Which managed service fits your cloud? AWS (ElastiCache/MemoryDB), Google Cloud (Memorystore), and Oracle (OCI Cache) all now offer Valkey natively, often at a lower hourly rate than their Redis-based equivalents.
• Do you need multi-vendor governance guarantees for compliance or procurement reasons? Valkey's Linux Foundation backing is increasingly a checkbox item in enterprise vendor reviews.

On the data side, Valkey retains both RDB snapshotting and AOF (append-only file) persistence in formats compatible with Redis 7.2, so existing backup and restore tooling generally works without modification. Client libraries are the other reassuring part of this migration story: the popular drivers across Python (redis-py), Node.js (ioredis, node-redis), Go (go-redis), and Java (Jedis, Lettuce) all talk the same RESP protocol Valkey implements, so there's no new SDK to learn or swap in. Teams that have already migrated report the actual engineering work is closer to updating a Terraform module or a docker-compose.yml image tag than anything resembling a database migration.

Where it gets more nuanced is at the cluster-operations layer. Valkey has diverged with its own cluster manager tooling and is shipping features on its own roadmap — including improvements to replication and multi-threading — that don't map one-to-one onto newer Redis Enterprise releases. If your team depends on Redis Enterprise-specific operational tooling (not just the open-source engine), budget time to validate that your runbooks still apply after a cutover.

FAQ

Is Valkey a complete replacement for Redis? For the vast majority of workloads using standard Redis OSS commands, yes — Valkey is protocol-compatible and works with existing Redis client libraries without code changes. Redis Enterprise-only modules are the main exception.

Why did AWS, Google, and Oracle all back Valkey instead of sticking with Redis? Redis Ltd.'s 2024 license change (RSALv2/SSPL) blocked cloud providers from offering fully open managed Redis services without a commercial agreement. Valkey, under Linux Foundation governance, removed that restriction and can't be unilaterally relicensed by a single vendor.

Is Valkey actually faster than Redis? Independent benchmarks from AWS and third-party reviewers show Valkey 8.1+ delivering roughly 8% higher throughput, 20-22% lower P99 latency, and 20% lower memory usage compared to Redis OSS, though results vary by workload.

Which major Linux distributions ship Valkey by default? Fedora, Ubuntu, Debian, and Arch all now ship Valkey as the default in-memory data store, and several have stopped packaging Redis entirely.

Is migrating from Redis to Valkey risky for a production system? For standard use cases, the risk is low because the command set and wire protocol are unchanged. Snap Inc. migrated 70% of its Redis clusters to Valkey while serving 5 billion daily requests without an application rewrite, cutting infrastructure costs by 60% in the process.

Conclusion

The Redis-to-Valkey story is a rare case where a licensing decision, not a technical one, reshaped an entire layer of internet infrastructure in under two years. Valkey isn't winning because it's a clever rebrand — it's winning because Linux Foundation governance removes the single point of failure that made Redis Ltd.'s 2024 decision possible in the first place, and because the performance numbers hold up independently of the politics.

If you're still running Redis OSS in production and haven't looked at Valkey, the switch is probably smaller than you think — and for some teams, it's already paying for itself.