Starting from Shipments Early Next Year
Includes Vera Rubin and Grace Blackwell
Driven by Memory Supply Shortages and Rising Prices
Nvidia, the world's largest artificial intelligence (AI) semiconductor chip company, has signaled price increases for its AI server products.
Bloomberg News reported on August 22 (local time), citing sources, that Nvidia recently notified major clients that it will raise prices for its AI chip systems by more than 15%. The increase will apply starting from shipments early next year. Systems based on its latest AI semiconductor chips, “Vera Rubin” and “Grace Blackwell,” are also included in the price hike.
Nvidia's price adjustment is analyzed as the impact of memory supply shortages. The AI accelerators supplied by Nvidia must be equipped with memory such as High Bandwidth Memory (HBM) to support AI computations.
With memory demand skyrocketing due to the recent AI boom, prices have been continuously rising. Apple and Qualcomm have also previously raised prices for major products citing memory shortages. Bloomberg explained that this is a prime example showing that the bargaining power of memory semiconductor manufacturers has grown powerful.
In particular, Nvidia AI accelerators are sold at tens of thousands of dollars each, and its gross profit margin reaches the 75% level. Nevertheless, analysts say that reflecting memory price increases in server prices demonstrates the intensity of the recent memory supply shortage.
The production capacity of TSMC, which contract-manufactures Nvidia chips, is failing to keep up with surging demand, and DRAM, mostly supplied by Samsung Electronics, SK Hynix, and Micron, is also short in supply compared to demand. Although the three memory makers are expanding production, price increases continue as they fail to catch up with the pace of AI infrastructure investment.
Although Amazon, Microsoft, Google, Meta, and others are expanding in-house AI semiconductor development, their reliance on Nvidia remains high in data center construction, making whether they can secure sufficient memory a variable in expanding their own chips.
Attention is focused on whether the rise in Nvidia AI server prices will become a new obstacle to infrastructure expansion, such as large-scale AI data center construction. Bloomberg explained that with data center construction already facing difficulties due to project delays, labor shortages, tight capital markets, and community backlash against development, this price hike is expected to add complexity.
