Summary
- OVHcloud says average increases will reach 28% for 2024-generation dedicated servers and 51% for 2026-generation hardware.
- The company says memory costs have risen several-fold as manufacturers divert capacity towards AI and high-bandwidth products.
- Component inflation is turning hardware procurement and inventory timing into a more significant operating risk for cloud and data centre providers.
OVHcloud is preparing price increases averaging 28% on 2024-generation dedicated-server hardware and 51% on 2026-generation systems, as soaring memory and storage costs begin flowing more visibly into cloud infrastructure pricing.
OVHcloud says new dedicated-server configurations will move to revised pricing from 1 September. From 1 October, some hardware already in production will also be affected when it renews, covering Advance, Game, Scale, and High Grade systems from the 2024 and 2026 generations.
The average increase per reference is 28% for Gen 2024 hardware and 51% for Gen 2026. The largest published movement is on some Gen 2026 Game systems, where the increase reaches 87%, while individual ranges vary substantially according to configuration.
The increases are considerably sharper than the broad adjustments OVHcloud outlined earlier in 2026. In March, the French provider said it intended to limit average increases for cloud infrastructure deployed between 2026 and 2028 to around 9% to 11%, despite much larger movements in component costs.
The gap illustrates how quickly the underlying supply market has moved.
OVHcloud says global memory manufacturers have redirected production capacity towards GPUs and high-bandwidth products used in AI systems, without an equivalent fall in conventional demand for DRAM and storage. The resulting squeeze has affected RAM, NVMe drives, hard disks, and, to a lesser extent, processors.
Klaba has said OVHcloud’s RAM costs reached around six times their level a year earlier and could rise further. The company’s earlier analysis projected RAM pricing 250% to 300% above September 2025 levels by the end of 2026, with a return to historical pricing unlikely before new production capacity has had time to come online.
For a vertically integrated provider such as OVHcloud, the effect reaches beyond a customer price list. The company designs and assembles much of its own server infrastructure, which means component procurement, inventory timing, and the balance between older and newer hardware generations directly affect the economics of its data centres.
That exposes an operational problem familiar across the sector. Hardware has to be ordered well before customer demand can be known precisely, but delaying purchases risks shortages or still higher prices. Buying too aggressively can leave capital tied up in equipment whose value falls quickly as processor and accelerator generations change.
OVHcloud described that exercise as a balance between ordering enough equipment to secure supply and avoiding over-committing without firm visibility over future customer demand.
The pressure also complicates the economics of AI infrastructure itself. The AI build-out is increasing demand for the accelerators and high-bandwidth memory that sit at the top of the value chain, but the same supply reallocation raises the cost of ordinary servers, storage systems, and cloud infrastructure that support less specialised workloads.
That makes the current investment cycle different from a straightforward data centre capacity boom. Spending on AI hardware can crowd the component market even for facilities that are not primarily hosting large GPU clusters.
OVHcloud has already front-loaded some capital expenditure this year to manage component inflation and secure supply. Its first-half results showed an adjusted EBITDA margin of 40.9%, while the company said exceptional hardware-cost inflation had influenced procurement timing.
The practical effect for operators is that capacity planning now has another volatile input. Electricity prices, grid access, construction costs, cooling requirements, and financing have been joined by unusually large swings in the cost of memory and storage.
The infrastructure itself may have an operating life measured in decades, but much of the IT equipment inside it is purchased on cycles measured in quarters. When component prices move as quickly as they have in 2026, the financial assumptions attached to a new hall or hardware refresh can change long before the underlying building does.
OVHcloud’s increases are therefore a useful indicator beyond its own customer base. They show that the AI investment surge is not only increasing the amount of hardware being installed in data centres. It is changing the cost structure of the ordinary compute and storage equipment already running inside them.
