In the early hours of Sunday morning, Beijing time, the price of servers equipped with Nvidia chips will rise. Several of Nvidia's main customers have been told that the price of these servers could increase by more than 15% due to the rising cost of memory chips. The price increase applies to systems shipped early next year, including servers powered by a combination of Vera Rubin and Grace Blackwell chips, depending on the Nvidia chip and memory configuration used.
Companies that manufacture servers for large data center operators such as Microsoft, Google and Oracle have notified customers of the upcoming price hike. Consumers have recently experienced a similar situation, with Nvidia's graphics card price for the consumer market rising sharply, such as the sub-flagship RTX 5080 graphics card from 8,000 yuan to 12,000 yuan in a month.
Advanced AI hashrate servers are expensive, and even a 15% increase can cost hundreds of thousands of dollars or even nearly a million dollars. Blackwell-based NVL72 GB200 racks are priced between $2.8 million and $3.4 million, while Vera Rubin-based NVL72 VR200 systems have reached $5 million to $7 million each.
The news has raised concerns about whether the “storage frenzy” will affect Nvidia's margins. Nvidia plans to release its latest earnings report after the US stock market on Wednesday next week. Earlier financial reports showed that the company achieved revenue of $81.6 billion in the three months ended April 26, with a gross profit margin of about 75%, and the company expects the gross profit margin in the second quarter of this fiscal year to remain between 74.9% and 75%.
Nvidia's chips are critical for training and running AI software, but their performance relies heavily on the accompanying DRAM. The current surge in demand for computing power has led to a significant increase in the price of such chips, which has given memory manufacturers unprecedented influence. Technology companies such as Apple and Qualcomm have also had to raise prices due to chip shortages.
The price of expensive Nvidia servers has further increased, prompting large cloud service providers such as Amazon, Microsoft, Google and Meta to speed up the pace of self-developed chips. Nonetheless, most new large data centers still rely heavily on Nvidia chips. For Nvidia's competitors, seizing the opportunity depends on whether they have access to an adequate supply of chips.
