Rethinking data egress
Rethinking data egress
Most teams compare cloud providers the same way. They line up compute prices, check storage rates, and pick the best deal. That approach misses a cost that grows quietly in the background: data egress.
Egress fees rarely show up on day one. They appear when your application finds its audience, when your backups start replicating, or when your AI pipeline begins moving terabytes. By then, the architecture is set and the bill is already climbing.
If you’re a builder or on a DevOps team running data-intensive workloads on cloud infrastructure, this article is for you. We’ll explore what egress fees are, which workloads they affect most, and how you can estimate their impact before they shape your roadmap.
What are cloud egress fees?
Cloud egress is data that leaves a provider's infrastructure for another destination. Depending on the provider and service, that movement can include:
- Application traffic delivered to end users
- Cross-region or cross-zone replication
- Object storage retrieval
- Backup and disaster recovery traffic
- Container images, build artifacts, and model artifacts
- Data exchanged with on-premises systems or another cloud
- Dataset movement for AI and machine learning (ML) pipelines
- Workload and data migration
Most hyperscalers charge for this outbound traffic by volume. Network usage then becomes another variable in your monthly bill, and one that is hard to forecast. This shifts the core question. You still need to know what compute costs. You also need to know what it costs once your workload starts moving data.
At OVHcloud, we take a different approach. In U.S. regions, network traffic is included in the price, so your bill doesn’t grow every time your data moves.
Egress shapes your architecture
When moving data costs more, it changes how teams design systems. Engineers start weighing transfer charges alongside performance and resilience. That affects decisions such as:
- Where workloads run
- Where data is stored
- How often environments replicate
- Whether moving to another provider makes economic sense
In other words, pricing starts making architecture decisions for you. A team might replicate less often than it should, or keep data in one region to avoid a fee. Over time, those trade-offs add up to less resilience and less freedom.
So what does that look like in real life? Let’s walk through two examples. We’re using typical hyperscaler list prices for internet egress, which start around $0.09 per GB and drop to about $0.05 per GB once you’re moving huge volumes. Traffic between regions is billed separately, typically around $0.02 per GB.
Blockchain infrastructure1: A team runs RPC nodes that serve blockchain data to wallets and decentralized apps, and those nodes share blocks with peers around the clock. In a normal month, they send about 20 TB to the internet, or roughly $1,750 in egress. Then a major token launch drives traffic up. That month, 100 TB leaves the network and the egress line climbs to about $7,800, even though the team didn't add a single server. Because the spike follows user demand, it’s difficult to budget. The team either absorbs the cost or throttles the service its users depend on.
Disaster recovery for a SaaS app: A SaaS company runs production on public cloud instances and keeps a warm standby in a second region. Between database writes, new files in object storage, and snapshots, about 5 TB of data changes every day, and all of it has to cross regions to keep the standby current. At a typical inter-region rate of $0.02 per GB, that’s about $100 a day, or roughly $3,000 a month, before a single failover test. When finance reviews the bill, the easy cut is to stop replicating the less critical datasets. The savings are real, but so is the risk: if the primary region goes down, that data isn’t there with which to fail over.
The effect is strongest at the exit. When leaving will result in a large transfer bill, staying becomes the default. Your infrastructure choices should reflect what is best for your workload, not what is cheapest to stay put. Some hyperscalers now waive transfer fees for customers who migrate away, but those credits require approval and cover only the move itself, not the egress you pay every month while you stay.
Predictable pricing, built in
At OVHcloud, we design our pricing so data movement stays predictable. With OVHcloud® Public Cloud instances in U.S. regions, public and private network traffic is included in the instance price. You can scale your application, replicate for resilience, and move datasets without watching a per-GB meter.
OVHcloud Bare Metal servers, VPS, and Hosted Private Cloud, including VMware® on OVHcloud and Nutanix® on OVHcloud, also include free ingress and egress in U.S. regions. That gives you room to design for performance and resilience first.
OVHcloud® is a registered trademark of OVH SAS. Nutanix® is a registered trademark of Nutanix, Inc. VMware is a registered trademark of Broadcom Inc.
Key takeaways
Egress fees turn growth into a cost variable. They hit hardest where data moves most: applications, backups, DR, object storage, AI, multi-cloud, and migration. They also influence architecture and make it harder to leave a provider.
The fix starts with visibility. Map how much data you move, where it goes, and how it grows at 5x and 10x. Then choose a provider whose pricing supports that growth instead of taxing it.
1 The examples are illustrative, use tiered list prices, and exclude free tiers and negotiated discounts.
OVHcloud® is a registered trademark of OVH SAS. Nutanix® is a registered trademark of Nutanix, Inc. VMware is a registered trademark of Broadcom Inc.