OCI Networking Costs: FastConnect, Load Balancers, and VCNs

Teams arriving on OCI from AWS or Azure bring a learned reflex: treat every network component as a meter. That reflex is expensive to ignore on those platforms, where NAT gateway processing fees, availability zone transfer charges, and per attachment transit costs quietly build a five figure monthly line in busy estates. On OCI the reflex mostly does not apply. The virtual cloud network and nearly everything attached to it carries no charge, traffic inside a region is free, and the components that do bill, chiefly load balancers and FastConnect ports, bill flat rather than by the gigabyte processed. This article maps the free from the billed, puts list numbers on the billed parts, and shows how to design so the network line stays where it belongs, near zero.

This article is part of our complete guide to OCI pricing and TCO. Internet egress, the one network line that can grow large, has its own treatment in OCI egress costs vs AWS and Azure, and the wider billing model is in OCI pricing explained.

What is free and what bills

The cleanest way to plan is a simple inventory. Free on OCI: the VCN itself, subnets, route tables, security lists and network security groups, internet gateways, NAT gateways, service gateways, dynamic routing gateways for most attachment types, site to site VPN, DNS within the VCN, and all traffic between resources in the same region, including across availability domains. Billed: load balancers, FastConnect ports, cross region traffic, internet egress past the free 10 TB, and premium DNS traffic at the edges.

Load balancers: flat and shapeable

OCI's flexible load balancer bills a small hourly base plus a rate on the bandwidth range you configure, with a minimum and maximum you set. A modest 10 Mbps floor costs a few tens of dollars a month, and even a load balancer shaped for hundreds of Mbps stays in the low hundreds. There is no per processed gigabyte meter and no capacity unit arithmetic. The cost discipline is simply hygiene: load balancers are the component most often left behind by retired applications, still billing their hourly base in front of nothing. The always free tier also includes a small 10 Mbps load balancer, enough for genuine dev and test fronting, covered in OCI free tier limits.

FastConnect: a flat port, not a meter

FastConnect, OCI's private circuit into your tenancy, bills as a flat port fee by capacity, with 1 Gbps and 10 Gbps the common sizes, and critically no per gigabyte charge for traffic over the circuit in either direction. The port fee lands in the hundreds of dollars per month band at list depending on capacity and partner model, and the partner or carrier delivering the physical circuit charges their own fee on top, which is often the larger number and the one teams forget to model. The flat pricing changes design behavior: on OCI you do not ration traffic over the private circuit, because replication, backups, and bulk movement over FastConnect cost nothing marginal. Organizations running hybrid estates with steady traffic between a data center and OCI find the circuit pays for itself quickly against metered alternatives.

Cross region traffic: the one meter that matters

Traffic between OCI regions bills per gigabyte at rates similar to internet egress. For most estates this is a small line, but two patterns make it material: cross region disaster recovery replication for large databases or object stores, and chatty multi region application designs. Both deserve deliberate sizing in the cost model rather than discovery on the first full bill. Replication for DR is usually worth every cent, but the volume should be estimated from change rates, and compression and scheduling decisions should be made with the meter in mind. This is one of the lines that calculator exercises consistently miss, a theme we expand in hidden costs on OCI.

DRGs, peering, and hub and spoke at scale

Larger estates organize networking around a dynamic routing gateway, with spoke VCNs attached to a hub carrying shared services, inspection, and the private circuits. On OCI this architecture is essentially free to operate: DRG attachments for VCNs carry no charge, local peering inside a region carries no charge, and the traffic flowing through the hub is intra region and therefore unmetered. The contrast with AWS Transit Gateway, which bills per attachment hour and per gigabyte processed, is stark enough that architectures rationed to a handful of attachments elsewhere can be designed with a clean attachment per workload on OCI without a cost conversation. Remote peering between regions does meter the traffic at cross region rates, so multi region hub designs concentrate their attention there, but the single region estate, which is most estates, runs its entire internal topology for nothing.

The edge lines: DNS, WAF, and certificates

The remaining network adjacent services are small but worth naming so nothing surprises. Public DNS bills per million queries at rates that round to tens of dollars for most estates. The web application firewall bills per protected request in a similar band, material only for very high traffic properties. Certificates through the native service are free, as are private DNS zones inside the VCN. Email delivery and health checks carry their own minor meters. The honest summary is that the entire edge layer for a typical estate costs less per month than a single load balancer, and the only line that ever grows teeth is WAF on a property serving billions of requests, which is exactly the kind of workload that should have modeled it deliberately in the first place.

A worked example: a hybrid three tier estate

Take a common shape: a production three tier application stack on OCI, a 10 Gbps FastConnect circuit back to a data center, cross region DR replication for a 5 TB database with a modest change rate, and normal internet traffic inside the free allowance. The monthly network bill breaks down to the FastConnect port fee plus the carrier circuit, a few load balancers shaped to measured traffic, and a cross region replication line in the low hundreds of dollars driven by the database change volume. Total: comfortably under two thousand dollars, of which the physical circuit is more than half. The same architecture on a metered platform, with NAT processing, cross AZ replication, transit attachments, and nine cent egress, routinely runs five to ten times that. This is the gap that compounds quietly in every TCO comparison.

A network cost design framework

Where the savings get spent: designing for the economics

Unmetered internals change what good architecture looks like, and teams that notice this early build better systems for the same money. Free cross availability domain traffic means synchronous replication and properly distributed clusters carry no cost penalty, so high availability becomes a pure engineering decision rather than a budget negotiation. Free NAT and service gateways mean private subnets with no internet exposure are the default posture, not a premium one, which quietly improves the security baseline. Unmetered FastConnect invites moving backup and bulk replication off the internet path entirely. The pattern repeats: designs that would be flagged as expensive on a metered platform are simply correct on OCI, and a landing zone review that imports metered era compromises, single availability domain deployments to avoid transfer fees, public subnets to dodge NAT costs, is leaving both resilience and security on the table at no saving whatsoever.

What this means against AWS and Azure

For a typical three tier estate, the OCI network bill is commonly 80 to 95 percent lower than the equivalent AWS or Azure network line, not because any single rate is dramatically cheaper but because most of the meters simply do not exist. A NAT gateway processing 50 TB a month costs over two thousand dollars on AWS and nothing on OCI. Cross AZ database replication that bills per gigabyte elsewhere is free inside an OCI region. These differences rarely headline a platform decision, but in TCO terms they compound monthly forever, which is why they belong in any serious comparison model, the kind we build in a defensible OCI TCO model.

Bringing it together

OCI networking pricing rewards teams that know the short list of what bills: load balancer shapes, FastConnect ports, cross region meters, and egress past 10 TB. Everything else is free, and designs that exploit that freedom, generous use of gateways, unmetered private circuits, free intra region replication, are both cheaper and simpler than their imported equivalents. A landing zone designed for OCI's economics rather than translated from another cloud's is one of the quiet wins of doing the foundation properly, which is the heart of our implementation practice. If your network line looks larger than it should, or your architecture was lifted from AWS patterns, an OCI assessment will show exactly where the meters are running.