Valkey Community

Valkey on the cloud

Provider-by-provider matrix of managed Valkey offerings as of mid-2026.

Pricing changes constantly and varies by region. Numbers in this page are directional only — always check the provider's pricing page for your region before committing.

Provider matrix

ProviderServiceEngine choiceFree tierTLSMulti-AZ HANotes
AWSElastiCache for ValkeyValkey 7.2 / 8.x / 9.xnoyesyesAbout 33% cheaper than equivalent Redis OSS option; default for new ElastiCache clusters
AWSMemoryDB for ValkeyValkey 7.2+noyesyesMulti-AZ durable, Valkey-compatible
GCPMemorystore for ValkeyValkey 7.2 / 8.xnoyesyesMemorystore for Redis frozen at 7.2; new versions go to the Valkey product
AivenValkey7.2 / 8.1 / 9.1yes (1 GB shared)yesyesFirst major DBaaS to migrate; multi-cloud
OracleOCI Cache with Valkey7.2 / 8.xtrial creditsyesyesOCI-native, integrated with OCI IAM
DigitalOceanManaged Valkey8.x / 9.xnoyesyes (HA tier)Simple pricing, droplet-style
Linode (Akamai)Managed Valkey7.2 / 8.xnoyesoptionalLower-cost option
AzureCache for RedisRedis onlynoyesyesNo native Valkey offering as of mid-2026 — self-host or use Aiven on Azure
阿里云 AlibabaTair (Valkey-compatible)Tair editionnoyesyesTair is Alibaba's enhanced fork; Valkey-compatible mode available
腾讯云 TencentTencentDB for RedismixednoyesyesMigrating engines toward Valkey, check region
华为云 HuaweiDCSRedis / ValkeynoyesyesValkey support added 2025

Notes by provider

AWS

AWS was a founding sponsor of Valkey and pushed hardest on the migration story. ElastiCache for Valkey is positioned as the default engine for new caches; existing Redis OSS clusters can be migrated in-place via the AWS console. Pricing tends to land roughly 30 percent below the equivalent Redis OSS configuration of the same node type.

MemoryDB is the durable variant — multi-AZ replicated transaction log, intended as a primary database, not just a cache. Valkey support arrived in 2024.

Google Cloud

The original "Memorystore for Redis" product is frozen at 7.2 (the last BSD Redis). New Valkey versions ship through "Memorystore for Valkey," which Google promotes as the forward-looking option.

Aiven

Aiven was the first major DBaaS to migrate, completing a roughly 15,000-server move from Redis to Valkey through 2024. The free tier is meaningful (1 GB shared) and good for prototyping. Aiven runs on AWS / GCP / Azure / DigitalOcean / UpCloud, so you can use it as a portable Valkey layer regardless of underlying cloud.

Oracle

OCI Cache offers Valkey alongside Redis. Tight integration with OCI Identity is the differentiator if you are already in OCI.

DigitalOcean / Linode

Simpler pricing for indie / SMB. DigitalOcean's HA tier is straightforward — single click adds a standby replica. Good for side projects and early-stage startups.

Azure

Azure has not shipped a native Valkey product as of June 2026. If you are on Azure and want Valkey, options are: (a) self-host on Azure VMs, (b) Aiven for Valkey running on Azure region, (c) wait — Azure's roadmap mentions Valkey support.

China clouds

Alibaba's Tair is a Valkey-compatible store with extensions (TairString, TairHash with field TTL, persistent memory tier, etc.). Many features that ended up in upstream Valkey were prototyped in Tair. Tencent Cloud and Huawei Cloud both support Valkey, with the specific availability varying by region.

Self-host vs managed: rough decision rule

SituationRecommendation
Single team, < 10 GB, no SLAself-host on a VM is fine
Multi-tenant, multi-AZ, on-call rotationmanaged
Strict data residency or complianceself-host or sovereign-cloud provider
Latency-sensitive (sub-millisecond)self-host close to app, or AWS MemoryDB with same-AZ pinning
Want vector + cache + memory unifiedany managed Valkey with modules enabled

Cost ballpark

For a single ~ 8 GB Valkey node with HA, expect roughly:

ProviderMonthly USD (rough, on-demand)
AWS ElastiCache (cache.m7g.large, multi-AZ)$200-250
GCP Memorystore (Standard 8 GB)$180-230
Aiven (Business-4, 1 vCPU 4 GB)$130-180
DigitalOcean (Managed Valkey HA 8 GB)$130-170
Self-hosted on a $80/mo VM with manual HA$80-100 + your time

Multi-region replication, backups, and egress are usually billed extra. For low-latency apps, the cost of cross-AZ traffic between app and cache often dwarfs the cache itself — co-locate.

On this page