Your load rises in the evening, falls at night, spikes on sale weekends. Two French clouds make the short list: OVHcloud Public Cloud and Scaleway. Both show competitive hourly prices on paper. Yet at month end the gap can hit thirty or forty percent — without anyone deliberately oversizing.
The difference rarely sits on the “instance” line. It sits in forgotten egress, accumulated snapshots, floating IPs left active, and orphan resources after an abandoned Kubernetes test.
Two similar billing models — but not identical
Both clouds bill instances, volumes, IP addresses, outbound traffic. OVHcloud adds vRack, anti-DDoS included on many ranges, and a broader enterprise path. Scaleway pushes a mature product API, competitive ARM instances, and a tightly integrated Kubernetes ecosystem (Kapsule).
| Line item | OVHcloud | Scaleway |
|---|---|---|
| Instance/hour | Competitive | Competitive |
| Egress | Quotas then billing | Same — read grid |
| Snapshots | Billed | Billed |
| Anti-DDoS | Often included | Plan-dependent |
| Ecosystem | vRack, possible HDS | SecNumCloud, Kapsule |
A CX or DEV1 catalogue price does not predict the bill for a shop with product video, or a platform that ships nightly backups to an external bucket.
Where the bill drifts with variable load
Egress — the most underestimated line. Off-site backup, streaming, CDN misconfigured to pull everything from origin: every outbound gigabyte counts. Both providers bill beyond included quotas.
Snapshots and orphan volumes — a finished test sometimes leaves disk + snapshot + IP. “Stopped” instance ≠ free: attached storage keeps billing.
Kubernetes and load balancers — nodes, LB, persistent volumes, inter-zone egress. Kapsule and MKS simplify deployment; they do not remove multiplied billing lines.
Spot or preemptible instances — good for batch, risky for stateful workloads without a recovery strategy.
Simulate three real months — not a quiet Monday
Export a test invoice or use the calculator with a peak × 3 scenario: worst Friday traffic, retained snapshots, video egress included. Compare line by line over ninety days, not the first provisioning hour.
Cross-check our egress fees enquiry and Scaleway vs Clever Cloud PaaS if your load mixes containers and managed apps.
The peak: the variable cloud bill is read in the worst month
Decide and move forward without blind spots
Start by inventorying current resources: instances, volumes, snapshots, IPs, buckets. Model three months with traffic peaks, realistic egress and snapshot retention. Compare Kapsule versus MKS if Kubernetes is in scope. Read both providers' egress and storage grids. Validate SecNumCloud scope if a public contract requires it.
See egress fees enquiry, Scaleway vs Clever Cloud, the directory and compare tool.
Frequently asked questions
OVHcloud or Scaleway for variable Kubernetes?
Compare Kapsule (Scaleway) and OVH MKS/Managed Kubernetes: price per node, load balancer, persistent volumes and egress. Scaleway is often competitive on API and integration. OVHcloud wins if you need vRack and native anti-DDoS in the same ecosystem.
Where does egress hurt?
Both bill outbound traffic beyond quotas. Video, off-site backup, misconfigured CDN or chatty APIs explode the bill. Model gigabytes per month over a quarter, not a quiet week.
Stopped instances: what do you pay?
Disk, IP and snapshots persist. Scaleway and OVHcloud bill attached storage — stopping an instance does not make infrastructure free. Delete or archive what you no longer use.
SecNumCloud on both?
Qualification possible on certain offers — verify exact scope if a public contract requires it. See SecNumCloud vs European cloud to frame the need.
Variable cloud: the right choice is the one whose worst month you simulated — not the quiet Monday.