Some of Nvidia's largest customers have been informed that servers equipped with the company's artificial intelligence chips could become more than 15% more expensive, according to Bloomberg on Saturday (22).
The new prices are expected to affect systems shipped in early 2027 and mainly reflect rising memory costs, an increasingly relevant component in the infrastructure required to train and run advanced AI models.
According to people familiar with the process, the impact will not be uniform. The size of the increase will depend on the generation of Nvidia chips used in each server and, mainly, on the memory configuration chosen by the customer.
Among the affected systems are machines based on the Grace Blackwell architectures and the new Vera Rubin generation, which occupies a central position in Nvidia's roadmap for the next cycles of artificial intelligence infrastructure expansion.
The AI bill is starting to get more expensive
The increase draws attention because it occurs in a part of the supply chain that already concentrates some of the biggest investments of the current technology cycle.
Cloud companies and large data center operators are buying thousands of AI accelerators and building specialized infrastructure to meet the growing demand for model training and inference.
In this scenario, even relatively small variations in the individual price of servers can represent significant differences in projects involving hundreds or thousands of machines.
According to the report, manufacturers responsible for assembling servers under contract for large data center operators have already begun notifying customers about the upcoming increases.
Among the companies cited are Microsoft, Google, and Oracle, three of the main groups currently involved in the global expansion of computing capacity aimed at artificial intelligence.
Memory becomes a critical part of the equation
The increase also highlights an important change in the economics of AI systems: the processor is no longer the only component that determines cost.
Modern servers designed for advanced artificial intelligence workloads depend on large volumes of high-speed memory to continuously feed GPUs and accelerators with data.
The more complex the systems, the greater the pressure tends to be on this part of the infrastructure.
With memory prices rising, this pressure is now starting to appear directly in the final price of machines sold to data center operators.
For Nvidia, this means managing higher costs precisely while its customers compete for capacity to rapidly expand their AI clusters.
Next generation will also be affected
The fact that Vera Rubin appears among the affected systems is particularly significant.
The platform represents the next stage in the evolution of Nvidia's accelerators after Blackwell and is expected to reach the market amid an even more intense race for computing capacity.
The price increase, therefore, is not limited to older products or to a one-off catalog adjustment.
It could accompany the next generation of infrastructure from its first large-scale commercial deployments.
For companies planning data center expansions for 2027, this adds a new variable to cost projections even before many of these systems begin to be delivered.
Nvidia remains at the center of AI investment
The timing of the increase also precedes an important event for the market.
Nvidia will release its 2nd-quarter results on August 26, in a presentation expected to be closely watched by investors seeking signs about the continuity of global spending on artificial intelligence infrastructure.
The company's position has made its results a kind of barometer for the entire AI ecosystem.
Its supply chain today connects chip manufacturers, memory suppliers, server companies, cloud providers, and groups responsible for financing the construction of new data centers.
Therefore, an increase of more than 15% in certain configurations is not just a price change for Nvidia.
It shows how the race for computing capacity is starting to pass on cost pressures from other components of the supply chain directly to some of the largest buyers of AI infrastructure in the world.
