Cloud VPN vs Dedicated Interconnect vs Partner Interconnect: Choosing a Hybrid Connection

GCP Study Hub
November 19, 2026

Google Cloud offers several ways to connect an on-premises network to a VPC network, and they differ in how traffic travels, how long they take to set up and what they cost. HA VPN runs encrypted tunnels over the public internet, Dedicated Interconnect is a direct physical link to Google, and Partner Interconnect carries traffic to Google through a supported service provider. Cross-Cloud Interconnect, Direct Peering and Carrier Peering sit nearby but solve different problems. In Professional Cloud Network Engineer scenarios, the stated constraints (timeline, bandwidth, budget, where the company has equipment, and whether traffic may touch the internet) usually rule out every option but one.

HA VPN over the existing internet link

HA VPN builds IPsec tunnels across the internet connection a site already has. If the on-premises firewall already supports IPsec and BGP and has a static public IP address, the tunnels can be built the same day. On the Google Cloud side, that means creating the VPN gateway, a Cloud Router and the tunnels, after which the two sides exchange routes over BGP. Traffic inside the tunnels uses internal IP addresses, so VMs and on-premises servers reach each other privately even though the packets cross the internet.

Cloud VPN is billed as an hourly charge per tunnel plus egress. For short-lived, low-volume connections, that makes it the cheapest of the hybrid options. Its weaknesses follow from the path it uses. Because traffic crosses the public internet, latency is less predictable, and a lossy internet path can cap throughput. Cloud VPN also doesn't negotiate compression with the peer gateway.

Dedicated Interconnect for a direct physical link

Dedicated Interconnect is a direct physical link to Google, delivered as 10 Gbps or 100 Gbps circuits, and it gives the lowest-latency path of the hybrid options. It requires your own equipment in a Google colocation facility and a physical cross-connect between that equipment and Google. Arranging the cross-connect usually takes weeks, and the connection carries a high monthly port charge. Dedicated Interconnect doesn't use pairing keys. Instead, you receive an LOA-CFA and use it to arrange the physical cross-connect at the facility.

Partner Interconnect through a service provider

Partner Interconnect uses a supported service provider to carry traffic between your network and Google. It fits when there is no Google colocation facility nearby or when the company has no equipment in one. Traffic travels over the provider's network instead of the public internet, and VMs stay reachable on their internal RFC 1918 addresses. The costs are a contract with the provider, time for the provider to provision the circuit, and provider fees on top of Google's charges. VLAN attachments are sized to the bandwidth the company actually needs.

The provisioning workflow starts in Google Cloud. You create a VLAN attachment, and Google generates a pairing key for it. You send that key to the provider, which uses it to provision its side of the connection. Once the provider finishes, you activate the attachment and set up BGP through Cloud Router. You can also choose to pre-activate the attachment when you create it.

Options that solve a different problem

  • Cross-Cloud Interconnect links Google Cloud to another public cloud provider such as AWS or Azure. It doesn't connect an on-premises data center.
  • Direct Peering and Carrier Peering reach Google's public IP ranges and services. Direct Peering doesn't give access to internal IP addresses in a VPC network, so it can't stand in for Cloud VPN or Cloud Interconnect when on-premises hosts need to reach VMs privately.

How to choose between them

A few rules settle most hybrid connectivity decisions on the Professional Cloud Network Engineer exam.

  • When the timeline is short, the volume is small and the on-premises device already supports IPsec and BGP with a static public IP address, HA VPN is the fastest and cheapest path.
  • When policy says traffic must not cross the public internet, even when encrypted, every Cloud VPN option is out, however it is configured.
  • When bandwidth needs are modest and the company's carrier is already a Partner Interconnect provider, Partner Interconnect is cheaper and simpler than putting equipment in a colocation facility for full 10 Gbps circuits.
  • When the company has equipment in a Google colocation facility, needs 10 Gbps or 100 Gbps circuits or the lowest latency, and can accept weeks of lead time, Dedicated Interconnect fits.
  • When the other end is another public cloud rather than a data center, Cross-Cloud Interconnect is the relevant product.

Ongoing egress also enters the comparison. Egress from Google Cloud over Cloud Interconnect is billed at a lower rate than internet egress, which helps when most traffic leaves Google Cloud. For a brief, low-volume connection, the hourly tunnel pricing of Cloud VPN remains the lower-cost choice.

Our Professional Cloud Network Engineer practice exams cover choosing a hybrid connection alongside Cloud Router BGP setup and Cloud Interconnect provisioning, with detailed explanations that drill these distinctions.

Get tips and updates from GCP Study Hub

arrow