Nvidia’s largest customers have been notified that servers containing its AI chips will rise in price by more than 15% in many cases from early next year, driven by memory costs. The company reports quarterly earnings next week.

Nvidia’s biggest customers have been told what the memory shortage is going to cost them. Servers containing its AI chips will rise in price by more than 15% in many cases, on systems shipping from early next year.

The increases reach the newest hardware. Systems built around the Vera Rubin and Grace Blackwell chips are affected, with the size of the rise depending on the chip generation and the memory configuration.

The warnings came through the middlemen. Companies that assemble servers under contract for operators including Microsoft, Google and Oracle have notified their customers, and Nvidia did not respond to requests for comment.

What makes this notable is who is passing on the cost. Nvidia holds a gross margin of about 75% and is the most valuable listed company in the world, and it is still not absorbing this.

The leverage sits with three companies. Samsung, SK hynix and Micron produce most of the world’s DRAM, and while output is rising it has not caught up with demand from AI infrastructure.

Consumers met this first. Apple and Qualcomm have both said component costs are pushing prices up, Nvidia raised its gaming card prices this month and AMD followed within days, while Amazon Web Services has already put GPU prices up 20%.

Europe’s exposure runs through public money as much as private. The EU has committed around €20bn to a set of AI gigafactories, and a French consortium has bid $10bn for one site.

Those bids and budgets were built on last year’s hardware prices. A 15% increase on the servers such a facility exists to buy is a material change to a plan drawn up before the memory market tightened.

Commercial European operators are buying into the same market. Nebius is tripling Nvidia capacity at its Finnish data centre, on economics that have just moved.

The build-out was already getting harder. Project delays, labour shortages, tighter capital markets and local opposition have complicated data centre plans, and this adds cost to all of it.

Nvidia reports quarterly earnings next week, which is when the useful question gets asked. Not whether demand is holding, but who ends up paying for memory, because at the moment the answer appears to be everyone downstream.

Get the TNW newsletter

Get the most important tech news in your inbox each week.