OCI Egress Costs vs AWS and Azure: The 10 TB Free Advantage

Every major cloud charges almost nothing to bring data in and meaningful money to send it out. The economics are deliberate: egress pricing is gravity, and gravity keeps workloads home. OCI broke from the pattern by giving every tenancy the first 10 TB of internet egress per month free and pricing the rest at roughly $0.0085 per GB in most commercial regions. AWS and Azure offer about 100 GB free and then charge in the region of nine cents per GB. That is a hundred times more free allowance and roughly a tenth the marginal rate, and for data heavy workloads it is frequently the largest single line in the platform comparison.

This article quantifies the gap with worked scenarios. It is part of our complete guide to OCI pricing and TCO, and pairs with OCI networking costs for the rest of the network bill.

The rates, side by side

Rates shown are list at the time of writing and vary slightly by region and tier. The structural relationships, the hundredfold free allowance gap and the roughly tenfold marginal rate gap, have been stable for years. Always check current price lists for the exact figures.

Three worked scenarios

A content API serving 20 TB a month. On OCI, the first 10 TB are free and the second 10 TB cost about $85. On AWS, after the 100 GB allowance, roughly 19.9 TB bills at tiered rates near nine cents, landing around $1,790 a month. The OCI number is about 5% of the AWS number, roughly $20,000 a year of difference on one workload, before any negotiated discount on either side.

A media platform pushing 200 TB a month. OCI: 190 TB billable at about $0.0085 is roughly $1,650 a month. AWS: volume tiering brings the average rate down, but the bill still lands in the region of $15,000 to $17,000 a month. The gap, around $160,000 a year, is the size of an engineer. At this scale many AWS architectures route through a CDN partly to manage egress economics, which adds its own cost and complexity, while the OCI architecture can often just serve the traffic.

Cross region replication for DR, 5 TB a month. Inter region transfer on OCI bills at low per GB rates that vary by route, commonly a few cents per GB or less. The equivalent AWS inter region flows price around two cents per GB. Here the gap narrows, and for some routes effectively closes, which is worth knowing before you assume OCI wins every transfer category. The detailed map of what is free and what is not lives in the networking costs guide.

Architectural consequences

Cheap egress is not just a smaller bill, it changes what you can sensibly build. Serving large objects directly from object storage becomes reasonable. Multicloud architectures, where OCI runs the database tier and another cloud runs services, stop being penalized by the data gravity tax, which is exactly the pattern behind Oracle Database at Azure and similar designs. Backup and DR copies to external targets stop being a budget conversation. And analytics platforms that export large result sets to downstream consumers can do so without an architecture review. Teams arriving from AWS habitually design around egress, with caching layers and consolidation patterns whose entire purpose is dodging a nine cent rate. On OCI some of that machinery is simply unnecessary, and removing it is itself a saving.

Where the advantage does not apply

Honesty requires the inverse list. The 10 TB allowance covers internet egress from the tenancy, not every transfer type: inter region replication bills at its own rates, and FastConnect, while free of per GB charges, carries a port fee of roughly $150 a month for 1 Gbps and around $930 for 10 Gbps. Egress savings also cannot rescue a workload that is poorly placed for other reasons, if your compute and data belong elsewhere, cheap egress is a consolation prize. And if your traffic profile is small, under the free allowance on any cloud, the entire category is noise and should not influence the platform decision at all.

How egress actually appears on the bill

Operationally, OCI meters outbound data transfer per tenancy per month, applies the 10 TB allowance across the tenancy as a whole rather than per resource, and bills the remainder against outbound data transfer line items by region. Three details matter for forecasting. The allowance resets monthly and does not roll over, so a spiky profile that alternates between 2 TB and 18 TB months pays for the spikes even though the average sits under the allowance. Transfer between availability domains inside a region is free, which makes architectures that chatter across domains safe in a way they are not on clouds that charge for every cross zone gigabyte. And object storage requests are billed separately from transfer, small money but visible at high request rates, worth a line in any serving heavy model.

CDNs change the question, not the answer

High volume serving architectures usually front content with a CDN on any cloud, which moves much of the user facing transfer to the CDN provider's pricing rather than the origin cloud's. That narrows the headline gap for cacheable content, but two OCI advantages survive. Origin fill, the traffic from your cloud to the CDN edge, still prices at the origin cloud's egress rates, a tenth the cost from OCI. And the long tail of traffic that is not cacheable, APIs, personalized responses, uploads and downloads of unique objects, never touches the CDN at all and pays origin rates in full. The honest comparison prices the cache miss and origin fill flows on both platforms, and that comparison keeps coming out the same way.

The multicloud case nobody prices

The fastest growing egress scenario in our assessments is not serving users, it is serving other clouds. Architectures that keep the Oracle database estate on OCI while application services run on Azure or AWS move data between platforms continuously, and on most clouds that interconnect traffic is a permanent tax. OCI's posture is unusually friendly here: FastConnect carries no per gigabyte charge in either direction, just the port fee, and the partnership interconnects to Azure run with low latency and no transfer charges on the OCI side. The result is that a split architecture, priced honestly, often carries a smaller data movement bill than the same architecture confined to one of the other clouds, a fact that surprises every finance team that checks it. The wider architectural pattern is covered across our pricing and TCO guide and the database platform articles in database service pricing.

Egress as an exit cost: the repatriation angle

Egress pricing is also exit pricing, and procurement teams increasingly treat it that way. Moving 500 TB out of a cloud at nine cents a gigabyte is a $45,000 line item before anyone touches a workload, which is small against a migration budget but large as a psychological anchor, and regulators have noticed: switching cost scrutiny in the EU and UK has already pushed the hyperscalers into offering free egress for customers leaving entirely, with conditions attached. OCI's structural rates make the question mostly moot in both directions, the same 500 TB costs about $4,200 at list and nothing within the allowance month by month. For architects, the practical consequence is that data placement decisions on OCI are cheaper to reverse, which lowers the cost of being wrong, and the option value of cheap reversibility belongs in platform comparisons even though no calculator has a field for it.

Measuring egress when your monitoring never did

Plenty of estates arrive at this analysis with no transfer measurements at all, because the current platform buried transfer in bundled bills or nobody set up flow logs. Three estimation paths work in practice. Billing archaeology: even opaque bills itemize transfer somewhere, and two hours with the detailed usage export usually produces a defensible monthly figure by category. Application arithmetic: requests per day multiplied by average response size, from load balancer logs, brackets user facing egress within tens of percent. And for replication flows, the change rate and schedule give the answer directly from the database or storage layer. Any of these beats the two common defaults, guessing zero or copying a dollar line across clouds, and once on OCI the question answers itself: the tenancy level transfer metering and cost analysis tags make the egress profile visible from the first month, which is one of the quieter benefits of the move.

A five step egress evaluation

Putting the number in the platform decision

Where does egress belong in the final platform comparison? As a measured line, not a slogan. The 10 TB allowance and the tenfold marginal gap are real, but their weight depends entirely on your traffic profile: a workload pushing 500 GB a month should ignore the category, a workload pushing 50 TB should let it lead the analysis, and most estates sit somewhere between, with one or two flows that dominate the total. The disciplined move is to rank your flows by volume, price the top three on both platforms at current rates, and carry that subtotal into the TCO model as its own line with the measurement attached. Done that way, egress either decides the comparison visibly or steps out of the way, and both outcomes beat the usual one, where it is discovered after the architecture is committed.

The bottom line

Egress is the rare cost category where the clouds differ by an order of magnitude rather than a few percent, and OCI holds the favorable end. For data heavy workloads, the 10 TB free allowance and the sub cent marginal rate can decide the platform comparison before compute pricing is even discussed. Measure your real flows, price them honestly on both sides, and if the result says the move is worth modeling properly, our consulting practice builds the full comparison, network architecture included, on a fixed project fee.