GCP Premium vs Standard Network Tier
GCP is the only hyperscaler that offers two distinct network tiers. Premium Tier uses Google's private backbone (cold-potato routing, fewer hops, lower latency). Standard Tier uses public internet ISP transit. Standard is 25 to 50 percent cheaper per GB but trades higher latency outside Google's metro areas.
Per-tier rates by destination (egressing from us-central1)
| Destination | Premium first 1 TB | Standard first 1 TB | Std savings |
|---|---|---|---|
| Worldwide ex specified | $0.12 | $0.085 | 29% |
| China (excl. HK) | $0.23 | $0.110 | 52% |
| Australia | $0.19 | $0.110 | 42% |
| India | $0.12 | $0.085 | 29% |
Standard Tier is set per-resource (per VM, per LB), not per-project. You can mix tiers within one workload. Standard is only available from specific source regions.
Latency trade-off in practice
Premium Tier keeps traffic on Google's private network until the closest possible edge to the destination user. Standard Tier hands off to public internet at the source region, where the user's ISP determines the path. For users in the same country as the source region, the latency difference is typically under 20 ms. For intercontinental traffic, Premium can be 80 to 200 ms faster.
Cost-optimisation pattern: use Standard Tier for non-user-facing egress (replication to off-site backup, ETL to BigQuery from external sources, logs to third-party SIEMs) where latency is irrelevant. Keep Premium Tier for user-facing API responses and media serving.
See also: GCP per-region egress, Cloud CDN pricing.