Nvidia AI Servers Jump More Than 15% in Price From Early Next Year

Nvidia‘s biggest customers have been told that servers built around its artificial intelligence chips will cost more than 15% extra in many cases from early next year.

What to Know:

Prices for systems using Nvidia’s AI chips will rise above 15% in many cases, on hardware shipping early next year.

The increases cover the Vera Rubin and Grace Blackwell generations, with the exact figure set by chip generation and memory configuration.

Nvidia reports quarterly results on Aug. 26, and analysts are watching the gross margin line rather than revenue.

Nvidia Server Prices Rise On Memory

Server makers that assemble machines under contract for Microsoft, Google and Oracle have already notified their customers, according to people familiar with the process. The increases apply to systems shipped early next year.

The newest hardware is not exempt. Machines carrying the flagship Vera Rubin and Grace Blackwell chips are affected, and the size of each rise depends on the chip generation and the memory configuration.

Nvidia did not immediately respond to a request for comment. Memory sits at the center of the increase, because the real-world performance of an AI accelerator rests heavily on the DRAM shipped alongside it.

Also Read: Nvidia AVO Achieves Breakthrough 100% on ARC-AGI-3

Morgan Stanley Flags Nvidia Margin Risk

The timing matters. Nvidia reports second-quarter results on Aug. 26, and Wall Street is watching what the company says about costs.

Management guided to roughly $91 billion in revenue for the quarter, plus or minus 2%, with a GAAP gross margin of 74.9%. Morgan Stanley projects revenue of $91.1 billion and earnings of $2.07 a share.

The bank has warned that higher costs for DRAM, wafers, advanced packaging and substrates could hold the gross margin in the mid-70% range or push it lower. Shares have fallen the day after each of the past four reports.

Not every analyst sees damage. Citi believes Nvidia has locked in high-bandwidth memory supply for 2026 and 2027, which would blunt part of the pressure over the next several quarters.

Three suppliers hold the leverage. Samsung, SK hynix and Micron account for most of the world’s DRAM output, and they have gained unusual bargaining power as demand from AI infrastructure runs ahead of supply.

DRAM Prices Reshape AI Costs

The squeeze reaches well beyond one product line. J.P. Morgan Global Research estimates that DRAM prices will have climbed more than 400% between the start of 2024 and the end of 2026.

This has been building all year. Conventional DRAM contract prices rose 90% to 95% quarter over quarter in the first quarter of 2026, according to TrendForce, and another 58% to 63% in the second. The research firm expects the pace to have slowed to between 13% and 18% in the third quarter, a deceleration that still leaves prices at record highs.