Skip to main content
Azure Resiliency Map
Service catalog
Networking

NAT Gateway

The StandardV2 SKU is zone-redundant by default; the older Standard SKU is a zonal resource pinned to one Availability Zone (or, if you skip zone selection, a nonzonal placement Azure picks for you — the weakest option). The zone configuration is locked in at creation and can't be changed afterward. Like other networking appliances, it's single-region with no native cross-region failover.

SLA SLA credits below 99.99% uptime (2+ healthy VMs required; SNAT exhaustion excluded)Last verified 2026-08-14
Local

Standard SKU, zonal (single AZ) or nonzonal placement

Standard SKU pinned to one Availability Zone, or deployed with no zone specified at all (nonzonal — Azure silently picks a zone for you). Either way, one zone's failure takes outbound connectivity down for every subnet attached, even for VMs sitting in otherwise-healthy zones.

Replication
None
RPO
N/A — stateless outbound connectivity, not a data store
RTO
Unavailable until the zone recovers, or until you manually reroute to a NAT gateway in another zone
Failover trigger
Manual (customer-triggered)
Relative cost
No extra cost
Protects against: Nothing beyond individual instance failure within the zone

Standard SKU baseline pricing, zonal or not.

Nonzonal is explicitly not recommended by Microsoft — it gives no protection against a zone outage despite looking like a simpler default.

Meets tier
T0
T1
T2
T3
T4

Shows Unknown at every tier — there's no fixed RPO/RTO for this option; it depends entirely on an architecture you'd have to design and build.

Zonal

StandardV2 SKU, zone-redundant (recommended)

Automatically spreads NAT gateway instances across every Availability Zone in the region. Traffic from any healthy zone routes through a surviving instance with no customer action.

Replication
None
RPO
N/A — stateless; active connections through the failed zone drop and clients retry
RTO
New connections route through a healthy zone almost immediately; existing connections in the failed zone must be retried
Failover trigger
Automatic
Relative cost
Low cost premium
Protects against: Datacenter/zone failure

StandardV2 may carry a modest premium over Standard SKU pricing — check current NAT Gateway pricing for exact rates before assuming parity.

Requires a StandardV2 SKU public IP — you can't attach a Standard (v1) public IP to a StandardV2 gateway.

Meets tier
T0
T1
T2
T3
T4
Regional

Independent NAT gateways per region (customer-built)

NAT Gateway has no multi-region capability of any kind. A second region means a fully separate NAT gateway, its own public IPs, and its own subnet attachment — there's nothing to fail over, only something to duplicate.

Replication
N/A
RPO
N/A — connectivity layer
RTO
Entirely dependent on your regional failover architecture
Failover trigger
N/A
Relative cost
High cost premium
Protects against: Only what you explicitly build

A fully separate NAT gateway, public IPs, and subnet attachment per region — full duplication, nothing shared.

Meets tier
T0
T1
T2
T3
T4

Shows Unknown at every tier — there's no fixed RPO/RTO for this option; it depends entirely on an architecture you'd have to design and build.

Gotchas

medium

Zone configuration is permanent

You cannot change a NAT gateway's Availability Zone configuration — zonal vs. zone-redundant, or which zone — after deployment. Fixing an under-protected NAT gateway means creating a new one (with a new, SKU-matched public IP) and re-pointing subnets to it.

high

"No zone specified" is not the safe default

Skipping zone selection on the Standard SKU doesn't give you zone redundancy — it gives you a nonzonal placement in a single zone Azure chooses, with the same all-or-nothing exposure as an explicit zonal deployment. Zone redundancy is opt-in via the StandardV2 SKU, not automatic.

low

The NAT gateway's resiliency doesn't cover the VMs behind it

A zone-redundant NAT gateway keeps outbound connectivity alive, but if the VMs using it aren't themselves spread across zones, a zone failure still takes down the workload — just not because of NAT Gateway. Assess the compute tier's zone posture separately.

Sources