Managed Redis, Memcached, and Valkey: AWS, GCP, Aiven, Upstash, Redis Cloud (2026)
The 2026 reality: AWS ElastiCache and Google Memorystore default to Valkey for new deployments, so if you created a new cluster recently you may already be on Valkey. Azure is the exception: it ships no managed Valkey and steers new workloads to Azure Managed Redis (Redis Enterprise) instead.
Quick answer: which managed services support Valkey?
GA dates: ElastiCache Oct 2024, Memorystore 18 Apr 2025, Azure Managed Redis 19 May 2025 (Redis Enterprise, not Valkey). Valkey = Linux Foundation BSD fork of Redis 7.2.4. Verified July 2026 vs provider announcements.
| Provider | Service | Defaults to | License-friendly | Notes |
|---|---|---|---|---|
| AWS | ElastiCache → | Valkey | Yes | Default for new deployments since late 2024. Memcached and Redis OSS still available as choices. |
| AWS | MemoryDB → | Valkey | Yes | Persistent, multi-AZ, Valkey-based. Strong durability guarantees. |
| Google Cloud | Memorystore for Valkey → | Valkey | Yes | Default for new instances since 2025. Recommended path for new GCP deployments. |
| Google Cloud | Memorystore for Redis → | Redis OSS | Depends | Legacy track. Kept for existing customers. Still available for Redis 7.x. |
| Google Cloud | Memorystore for Memcached | Memcached | Yes | Separate managed Memcached service. Continues for existing workloads. |
| Microsoft Azure | Azure Managed Redis → | Redis Enterprise | Depends on use | No managed Valkey on Azure. Microsoft licenses Redis Enterprise from Redis Inc, so its services are insulated from the SSPL/AGPL fork. GA May 2025; the path Azure steers new workloads to. |
| Microsoft Azure | Azure Cache for Redis → | Redis OSS / Enterprise | Depends on use | Legacy service, supported but in maintenance mode for new builds. Still no managed Valkey or Memcached on Azure: roll your own on VMs/AKS for those. |
| Aiven | Aiven for Valkey | Valkey | Yes | Primary Aiven caching offering. Cited ~20% cost savings vs equivalent Redis Cloud. |
| Aiven | Aiven for Redis | Redis | Limited | Legacy support. Migration to Aiven for Valkey recommended. |
| DigitalOcean | Managed Caching for Valkey → | Valkey | Yes | Replaced DO Managed Redis on 24 Apr 2025; Redis-engine deployments closed 30 Apr 2025. Valkey 8.0 / Redis 7.2.4 compatible. From $15/mo single node. |
| Upstash | Upstash Redis → | Redis | Depends | Serverless, REST + RESP, pay-per-request. Strong for spiky/low-traffic workloads. |
| Redis Inc | Redis Cloud → | Redis 8.8 | Depends on use | Vector sets, hash field TTL, commercial support contracts. Redis 8.x tri-license. |
| DragonflyDB | DragonflyDB Cloud | DragonflyDB | Permissive | Redis-RESP-compatible but separate codebase. Multi-threaded. Separate eval for new builds. |
AWS ElastiCache reality (2026)
Late 2024, AWS migrated ElastiCache defaults to Valkey for new deployments. This was a direct response to Redis's SSPL license change: SSPLv1 makes the managed-service business model legally fraught for AWS. Valkey's BSD license removes that risk. If you create a new ElastiCache cluster today, Valkey is the default engine. Memcached and Redis OSS remain available as explicit choices. Existing Redis OSS clusters are not automatically migrated. Source: TechCrunch March 2024, AWS ElastiCache docs.
Source →Google Cloud Memorystore reality (2026)
Memorystore for Valkey launched in 2025 and is now the default for new instance creation. Memorystore for Redis remains available for existing customers on older Redis versions. Memorystore for Memcached has been a separate product for years and continues. If you are creating a new GCP in-memory store today, Memorystore for Valkey is the recommended and default path.
Azure reality (2026): no managed Valkey
Azure is the outlier among the big three: it offers no managed Valkey and no managed Memcached. Because Microsoft licenses Redis Enterprise directly from Redis Inc, its first-party services (Azure Cache for Redis and the newer Azure Managed Redis, GA May 2025) are insulated from the SSPL/AGPL fork and Microsoft had no commercial reason to back Valkey the way AWS, Google, and Oracle did. So if you searched for 'Azure Cache for Redis Valkey support', the answer in 2026 is: there is no managed Valkey on Azure. New Azure workloads are steered to Azure Managed Redis (Redis Enterprise); for Valkey or Memcached you self-host on VMs or AKS. Full tier breakdown on the Azure page.
Source →Self-hosted vs managed
For most production workloads, managed wins on total cost: instance cost + operations time + downtime risk. Self-hosted Redis or Valkey on a single VM is straightforward and cheap for development or low-traffic production. Self-hosted Redis Cluster with multi-AZ replication and automatic failover is real engineering work (Sentinel or Cluster mode, six nodes minimum for HA). The platform team that owns self-hosted Redis at scale has a dedicated operational burden. If you have a platform team and the workload justifies it, self-hosted gives you full control. If not, managed is the pragmatic choice. Sticking with self-hosted as your organisation scales accumulates platform engineering debt.
Source →The DragonflyDB question
DragonflyDB is a separate codebase from Redis and Valkey, Redis-RESP-compatible, multi-threaded by design. Claims 25x fewer nodes than Redis at equivalent throughput. Permissive licensing for the community edition. Worth evaluating for new builds where you want multi-threaded performance and cannot run Memcached. Not a drop-in for existing Redis Stack-using shops (no RedisJSON, RediSearch). The managed DragonflyDB Cloud is available for cloud deployments.
Verdict: 2026
On AWS or GCP, Valkey is the 2026 default. On Aiven, Valkey saves money. On Upstash, the question is workload shape, not engine. Redis Cloud is the path if you need Redis 8.0-specific features (vector sets, hash field TTL) or commercial Redis Inc SLAs. Self-hosting is a tech-debt accumulator unless you have a dedicated platform team. The full stack cost lens: techstackcost.com