AI server costs to jump 15% as memory chip prices surge

Major technology companies building data centre servers have begun informing clients that systems using Nvidia's artificial intelligence processors will cost more than 15 per cent extra when units ship early next year, driven by surging memory chip prices.

Contract manufacturers supplying large cloud operators including Microsoft, Alphabet's Google and Oracle have sent out the notifications in recent weeks, according to people with knowledge of the communications. The increases will vary depending on which generation of Nvidia chips customers order and how much memory the systems are configured with.

Flagship chips hit by increases

The price rises will affect systems built around Nvidia's newest processors, including the flagship Vera Rubin and Grace Blackwell chips. Nvidia did not respond to requests for comment.

Memory makers gain pricing power

The move underscores the bargaining strength memory chip manufacturers, Samsung Electronics, SK Hynix and Micron Technology, now wield as AI infrastructure demand accelerates, leaving even the industry's most dominant player unable to absorb rising component costs. Nvidia's accelerator processors form the core of computers that build and operate AI software, but their performance hinges on the amount of dynamic random access memory, or DRAM, paired with them.

Other major technology firms have recently disclosed similar pricing pressure. Apple and Qualcomm both told investors they have been compelled to raise product prices because of chip shortages.

Source: South China Morning Post