The numbers come first: Nvidia's biggest customers have been told that servers containing its AI chips will cost more than 15% more in many cases, with the increases set to take effect on systems shipped early next year. The hikes will affect systems built around the flagship Vera Rubin and Grace Blackwell chips, according to people familiar with the process who asked not to be identified.
The increases vary depending on the generation of Nvidia chips and the memory configurations involved. Companies that build servers under contract for large data center operators — including Microsoft, Alphabet's Google and Oracle — have already notified those customers of the forthcoming price changes. Nvidia representatives did not respond to requests for comment.
Memory makers hold the leverage
The driver is DRAM. Nvidia's accelerator processors depend heavily on how much dynamic random access memory they are paired with, and the three companies that dominate global DRAM production — Samsung Electronics, SK Hynix and Micron Technology — have not kept pace with surging demand. That supply gap has handed memory chip manufacturers unprecedented pricing power across the entire AI infrastructure stack.
Nvidia is hardly alone in absorbing the pressure. Apple and Qualcomm have both recently said chip shortages have forced them to raise prices on their own products.
The situation underscores a structural tension in the AI buildout: Nvidia commands a 75% gross margin and charges tens of thousands of dollars per chip, yet even that pricing power has limits when upstream suppliers control a critical input. Taiwan Semiconductor Manufacturing Co. handles Nvidia's chip production, but TSMC cannot solve a DRAM shortage.
Customers are captive — for now
Major buyers including Amazon, Microsoft, Google and Meta are all developing in-house chip programs, but remain dependent on Nvidia purchases for their current data center expansion. Their ability to reduce that dependence will itself hinge on securing adequate memory supply from Samsung, SK Hynix and Micron — the same bottleneck driving today's price hikes.
Nvidia has also raised prices on gaming-oriented PC graphics cards, according to industry news site Tom's Hardware, suggesting the cost pressure is not limited to enterprise AI hardware.
Nvidia is scheduled to report fiscal second-quarter earnings next week. The results will be closely watched by investors who have committed heavily to AI infrastructure on the expectation that it will reshape the broader economy.
CEO Times take: Free markets price scarcity honestly, and the DRAM shortage is doing exactly that — sending a clear signal that capital needs to flow into memory production capacity. The real story here is not Nvidia's pricing power but the concentration risk embedded in a global AI buildout that runs through three memory suppliers and one contract manufacturer. For enterprise buyers, the 15%-plus increase is a tax on the speed of their own AI ambitions. For investors, the question heading into next week's earnings is straightforward: at a 75% gross margin, how much of this cost can Nvidia absorb before the number moves? The market will have an answer before the week is out.
