Introduction
On October 2, 2026, starting at 14:00, a peering breakdown between OVH and Orange took thousands of business websites offline across France and parts of Europe. No cyberattack. No hardware failure in a datacenter. Just a peering agreement that collapsed — and with it, a significant slice of the French professional web.
A Routing Incident, Not a Server Outage
The root cause was an anomaly in traffic exchange at the peering level between the two operators. Peering is the bilateral agreement by which two networks exchange traffic directly, bypassing third-party transit providers. When that direct gateway breaks down, data packets fall back on alternative routes — which quickly became saturated, lacking the capacity to absorb normal traffic volumes.
The result: abnormally high latency, cascading timeouts, and sites responding intermittently or not at all. No data was lost. No server went offline. But from the perspective of end users and business customers, the operational impact was indistinguishable from a hard failure.
The Concentration Risk in French Web Infrastructure
The incident takes on added weight when you consider OVHcloud's footprint in France's digital infrastructure. The group hosts a substantial share of French professional web operations: SMBs, public agencies, software publishers, and e-commerce players. With more than 400,000 servers spread across some thirty datacenters worldwide and network capacity exceeding 30 terabits per second, OVHcloud is not just another hosting provider — it is, in practice, a piece of critical infrastructure.
That concentration is precisely what amplifies network incidents of this kind. A BGP anomaly on a peering agreement with a major operator like Orange does not affect a handful of customers: it can knock thousands of services offline simultaneously, with no remediation available to the hosted teams themselves.
What Your Business Continuity Plans Didn't Model
Most disaster recovery plans (DRPs) and business continuity plans (BCPs) map servers, backups, and databases. Very few go as far as modeling network dependencies at the inter-operator interconnection layer.
Yet that is precisely where some of the hardest-to-anticipate — and slowest-to-resolve — failure points live. They are slow to fix because resolution requires coordinated action between two separate organizations. During the October 2 incident, no restoration timeline was communicated in real time, for the simple reason that the fix required a joint reconfiguration between OVH and Orange.
For CIOs and IT teams, the lesson is directly actionable: cloud architecture resilience is not limited to machine redundancy. It requires scrutinizing your provider's multi-homing setup — its ability to interconnect via multiple operators simultaneously — and, depending on business criticality, diversifying hosting footprint or deploying CDNs capable of rerouting traffic when a network path becomes unavailable.
Digital Sovereignty and Reachability Are Not the Same Thing
A hosting provider can hold sovereignty certifications, guarantee domestic data residency, and maintain full control over its physical infrastructure — and still be structurally dependent on peering agreements with third parties to ensure its services are actually reachable. The network layer is routinely absent from dependency audits.
This is not a criticism of OVHcloud, whose role in France's digital ecosystem remains foundational. It is a useful reminder for every IT decision-maker: infrastructure control does not stop at the datacenter walls. The next blind spot in your BCP may well carry an AS number and a BGP routing table.