- OVHcloud says vertical integration is its edge
- The company stretches server lifespans and avoids bandwidth fees
- Sovereignty is central to OVHcloud’s cloud strategy, with a separate U.S. structure aimed at protecting European customer data
OVHcloud is betting that its unusual level of vertical integration can help it stand out in a market increasingly defined by AI power constraints, cloud sovereignty concerns and customer frustration with unpredictable costs.
Fierce recently toured OVHcloud’s Vint Hill data center in Virginia with U.S. General Manager Jeffrey Gregor.
Walking into the facility was a bit like entering a mom-and-pop cafe. There’s a grill, an espresso machine and a deli case. The building is older, but while it might lack the polish of a brand-new Starbucks, it still has all the equipment to serve up whatever meal order you can dream up.
In other words, OVHcloud’s facility is not another cookie cutter hyperscale data center. And that’s exactly the point.
France-headquartered OVHcloud operates more than 40 data centers across four continents. Even at this scale, Gregor told Fierce the company prides itself on its “build it here” mindset. And by “it” he means everything. OVHcloud builds its own racks, servers, dry coolers and even its liquid cooling systems itself, with AMD and Intel among its chip partners.
That level of control is both a cost lever and also how OVHcloud ensures its continuity for customers, Gregor said.
“We want to be in control because we know that having to rely on third-party providers or software vendors that may go out of business because they're overcome by AI or something” would be a liability, he explained. “So, we try and be in control of everything that we can to make sure that we can deliver our deliver for our customers.”
The OVHcloud trifecta
Being in control allows OVHcloud to offer a wide range of cost optionality. Gregor said the company refreshes its server range every 12 to 18 months, but it does not simply retire older equipment as each new generation arrives. Instead, OVHcloud repurposes and remarkets servers for different customer needs, stretching some machines to roughly eight, nine or even 10 years of service life.
That matters because not every workload needs the latest and greatest chip, he said. Some customers want high core counts and maximum performance, but others are running applications where older or more modest configurations are a better fit. Gregor pointed to one customer that specifically wanted low-core-count hosts because its software licensing model charged by the core, making newer, denser servers a more expensive option rather than an upgrade.
Bandwidth is another place where OVHcloud is trying to separate itself from larger cloud rivals. Gregor described the company as a pure-play hybrid cloud provider spanning bare metal, public cloud and private cloud, but said one of its biggest selling points is simpler: OVHcloud does not charge for bandwidth. Customers can choose the interface speed they want without worrying that heavy data movement will trigger surprise overage charges.
That can be a meaningful difference for businesses moving data between production, backup and disaster recovery environments. Gregor said one California customer runs production workloads in Vint Hill and backups in OVHcloud’s Hillsboro, Oregon, facility, with data moving across environments in ways that would be hard to justify if every transfer carried an additional cost.
“Where they get this surprise, especially with the hyperscalers, is when they get this massive bill,” he said. “There is no way that they can forecast for how much network traffic they’re going to have and when they get charged for the bandwidth, it blows up their whole budget.”
Sovereignty is the third leg of OVHcloud’s control pitch. Gregor said OVHcloud U.S. was set up as a separate company with its own governance, board and backend systems, creating what he described as a hard separation between U.S. and European customer data. The goal, he said, was to protect European customers from exposure to U.S. laws such as the Cloud Act and Patriot Act.
That sovereignty message may become more important as customers begin asking harder questions about AI infrastructure.
At least among its customer base, Gregor said U.S. demand for GPUs and AI services is still developing, but customers are increasingly focused on where their data lives, what models are running on it and whether that data could be used to train someone else’s model.
For OVHcloud, he argued, the answer again comes back to control: control over the hardware, control over the network and control over where customer data stays.
Read more about OVHcloud and data centers here:
OVHcloud’s data center strategy: Reuse, retrofit and reduce backlash
Can the US curb Chinese transceivers without slowing AI data centers?
AI data centers could unlock more compute without more power