OVHcloud raises server prices 40 to 87 percent as AI demand drives memory costs up sixfold

OVHcloud’s price increases reflect a broader supply chain shift where AI workloads are consuming an outsized share of high-bandwidth memory production. The company reports memory costs rising sixfold in a year, with forecasts suggesting further increases. This is not a temporary spike but a structural realignment, as suppliers prioritize higher-margin components for AI over standard DDR4 and DDR5. For engineers, this means the cost of running non-AI workloads on cloud infrastructure is rising, particularly for newer, more performant hardware.

The price hikes are uneven, targeting recent server generations while sparing older equipment. Gaming servers see the steepest increase at 87 percent, while other recent servers rise 40 to 59 percent. OVHcloud is also restructuring some charges, such as separating storage and IP costs for Gen3 instances. This suggests the company is attempting to balance cost recovery with customer retention, but the changes may complicate budgeting for teams relying on predictable pricing. The removal of shorter-term saving plans further pressures customers to commit to longer contracts to lock in rates.

The asymmetry between OVHcloud and hyperscalers like AWS highlights the advantage of scale and vertical integration. AWS has raised prices only for reserved GPU instances, leaving the rest of its catalog untouched. This is likely due to long-term procurement contracts and in-house accelerator designs, which insulate it from short-term market fluctuations. OVHcloud, by contrast, buys components on rolling monthly orders without price guarantees, leaving it exposed to volatility. This gap may not last, however, as supply constraints could eventually force even hyperscalers to adjust pricing.

Customer reactions suggest frustration less with the price increases themselves than with their frequency and unpredictability. OVHcloud’s founder has indicated that the current situation may persist until 2028, implying that further adjustments are possible. For engineers, this underscores the need to evaluate infrastructure costs as an ongoing operational concern rather than a fixed line item. Teams using mid-tier providers may need to explore alternatives, such as migrating to hyperscalers or optimizing workloads for older, unaffected hardware.