Qualcomm has announced plans to develop custom data center CPUs that use Nvidia technology to connect to Nvidia’s AI chips. This marks the company’s return to the data center CPU market, where it was previously active but scaled back its operations due to cost constraints and legal challenges.
The US chip manufacturer has unveiled plans to regain a position in the growing market for data center CPUs. These processors will be designed to work optimally with Nvidia GPUs, which currently dominate the AI market but must be paired with CPUs.
Nvidia has also made inroads into the CPU market by developing its own “Grace” processor based on Arm technology. Intel and AMD have traditionally dominated the CPU market, but the rise of AI workloads is creating new opportunities for other players.
Renewed focus on data centers
In the 2010s, Qualcomm began developing Arm-based CPUs for data centers, which it tested with Meta Platforms (then Facebook). These efforts were scaled back due to cost constraints and legal challenges the company faced at the time.
After acquiring a team of former Apple chip designers in 2021, Qualcomm has revived its efforts in the data center market. The company is again in talks with Meta about a data center CPU and confirmed last week that it has reached a letter of intent with Saudi AI company Humain to develop a customized CPU.
The newly announced chips will use Nvidia technology that enables fast communication with Nvidia’s GPUs. This should ensure that the processors can perform optimally in AI applications, where fast data transfer between different components is essential.
“With the ability to connect our custom processors to Nvidia’s rack-scale architecture, we are advancing a shared vision of high-performance, energy-efficient computing for the data center,” Qualcomm CEO Cristiano Amon said Monday.