A physically separate datacenter inside a cloud region, the unit of fault isolation cloud architectures use for high availability.
An availability zone (AZ) is one of multiple physically separate datacenters within a cloud region (typically 3 AZs per region in AWS, GCP, and Azure). AZs share low-latency connectivity (usually under 1ms) but isolated power, cooling, and networking, so a failure in one AZ does not affect the others. Multi-AZ deployments place replicas in different AZs so the loss of any single AZ does not take the service down.
Single-AZ outages are infrequent but they happen multiple times per year across the major clouds. Services that run in only one AZ are betting that none of those will hit them. Multi-AZ is the default cloud reliability primitive, more impactful and cheaper than multi-region for most use cases, because cross-AZ replication has near-zero latency cost while cross-region adds 50-200ms.
See the part of the platform that handles availability zone (az) in production.