# OVHcloud Raises Prices as AI Memory Demand Reprices Non-AI Infrastructure > Source: > Published: 2026-08-23 08:43:00+00:00 OVHcloud will raise prices across most of its catalog this autumn, and the cause sits outside the products being repriced. Memory and storage costs have climbed steeply since mid-2025, and founder Octave Klaba traces that to the AI buildout pulling fabrication capacity away from the components every ordinary server is built from. Klaba set out the mechanism in a [post on X](https://x.com/olesovhcom/status/2086753118881038591). Three global RAM suppliers have reconfigured their factories toward high-bandwidth memory, which serves GPUs and carries better margins, at the expense of standard DDR4 and DDR5. He posts in French, and the quotes here come from platform translations. The numbers come from OVHcloud's own purchasing. Indexed to June 2025, memory reached 604 by June 2026, SSDs 323, and hard drives 148. Klaba's forecast goes further: This month, in June 2026, we're paying 6x the price for RAM that we paid in June 2025. We already know it will be x9 in September 2026, while forecasts for early 2027 are at x12!? *[Components inflation, indexed to June 2025 (Source: OVHcloud)]* He adds that NVMe drives are at 7x and hard drives at 3.5x, and reports hearing that CPUs, motherboards, and network cards will rise 15 to 20 percent. The increases land unevenly. Renters of 2026-edition gaming servers face the steepest rise at 87 percent, while other recent servers go up 40 to 59 percent, all from September. Customers on 2024-era equipment see smaller increases at renewal, which Klaba puts at three to six times lower than a new order. Older ranges including Kimsufi, Rise, and earlier Advance and Scale generations are untouched, as they were in an April increase. Some charges are also being restructured rather than raised. From October 1, storage and IP addresses become separate line items on Gen3 instances, at €0.000146 per GB per hour and €0.0027 per hour. OVHcloud is dropping its one-month, six-month, and 24-month saving plans, keeping the 12- and 36-month options that lock pricing for their duration. Klaba is direct about how long this lasts: The situation is exceptional and will last until 2028. AI demand is insane at every level: datacenters, GPUs, token-as-a-service, agentic AI. Everyone wants it, everyone is using it. That's what's impacting every business that's not "AI". He is equally direct that absorbing the costs was not an option: The main concern we want to avoid is running out of parts and not being able to deliver the Cloud to you. What lifts this above a vendor pricing note is the asymmetry it exposes. Klaba [described](https://x.com/olesovhcom/status/2086753118881038591) OVHcloud's purchasing position as ordering: month by month, over 12 months, with no guarantee of the purchase price and without knowing what will be the real demand from our customers Amazon buys in the same market and has moved almost nothing. It raised EC2 Capacity Blocks for ML, a reserved GPU product, by roughly 20 percent in July after about 15 percent in January, and left the rest of the catalog alone, including instances running Trainium, the accelerator it designs itself. The change appeared on the pricing page with no announcement, and the company said afterward only that Capacity Blocks prices are updated periodically based on supply and demand. Procurement scale and vertical integration explain the gap. Hyperscalers contract for memory years ahead in volumes that earn priority allocation, and several design their own accelerators. A provider buying merchant components on rolling monthly orders has neither cushion. That gap may prove temporal rather than permanent. [The Register](https://www.theregister.com/off-prem/2026/08/11/ovh-cloud-warns-of-87-price-hikes-to-help-it-cover-rampocalypse-costs/) spoke to the founder of a mid-sized managed service provider who was unsurprised by the OVHcloud news and expects Azure and AWS to announce similar increases before long. Klaba hedges the same way, qualifying his own pricing claims with "if our competitors don't increase their prices." Customer reaction on [Reddit](https://www.reddit.com/r/OVHcloud/comments/1vhwh5n/price_increase/) was limited but pointed, and centered less on the increase than on its frequency. Commenter bammitscamm summarized the schedule and added: Not happy about this, especially after just raising the prices in April. Others treated the pressure as market-wide rather than particular to OVHcloud, noting that RAM prices and availability have gone crazy and that the problem reaches beyond memory into SSDs. One commenter described leaving entirely for self-hosted hardware. If those reactions are representative, the immediate customer question isn't whether the increase is justified, but how often it recurs. Two rises inside six months turn capacity planning into a rolling negotiation, and the surviving saving plans read as OVHcloud saying so. Andrei Nutas, who is serializing a book on digital sovereignty, argues the companies driving memory demand are also the ones insulated from the resulting prices, because they bought ahead of the move. He [frames](https://andreinutas.substack.com/p/eu-sovereign-cloud-providers-pay) it as a structural dependency: OVHcloud's invoices go out from Roubaix in euros, but the numbers on them are decided in fabs and purchasing departments Europe has no part in. Klaba maintains OVHcloud is still the cheapest option for bare metal and public cloud, while putting a number on what the rise costs that position: where before we could be 3x cheaper, we will be 2x cheaper (if our competitors don't increase their prices) That erosion matters for public sector migration. Procurement frameworks and certification schemes across Europe increasingly favor sovereign providers on jurisdictional grounds, with cheap infrastructure doing quiet work in the underlying business cases. As [InfoQ reported](https://www.infoq.com/news/2026/07/airbus-scaleway-sovereign-cloud/), Airbus made protection from non-European extraterritorial law a scored criterion in its cloud tender, alongside technical capability and operational excellence. Nothing in that scorecard accounts for a supplier whose input costs are set by the buying decisions of the providers it competes with. Timing compounds the problem. Elevated component prices through 2028 sit beyond the horizon of most European migration programs now running. Memory markets are cyclical and have crashed as sharply as they have spiked, so the shock will pass. The budgets written during it will not.