Qualcomm, the American semiconductor corporation that is a strong player in application processors (AP) used in smartphones, is making a comeback in the central processing unit (CPU) market for data centers.
Qualcomm noted on the 19th that it will develop a CPU for data centers that connects with NVIDIA's leading artificial intelligence (AI) chips.
Qualcomm effectively withdrew from the market after testing a CPU based on the British semiconductor design corporation Arm in the 2010s due to expense issues.
Qualcomm explained that the upcoming chips will be able to rapidly exchange data with NVIDIA's graphics processing unit (GPU).
According to Reuters, Cristiano Amon, Qualcomm's Chief Executive Officer (CEO), who attended Asia's largest information technology (IT) exhibition 'Computex,' said, 'By being able to connect our customized processors with NVIDIA's rack-scale architecture, we have expedited the vision for high-performance and high-efficiency computing for data centers.'
While NVIDIA's chips dominate the AI market, these chips operate in a configuration that combines GPUs and CPUs. The latest NVIDIA AI chips, such as GB200 and GB300, are a combination of NVIDIA's own CPU and GPU, Blackwell.
The CPU market has traditionally been divided between American semiconductor corporations Intel and AMD, but big tech companies are also entering the fray over CPU development for data centers.
In addition to NVIDIA, Google, Microsoft (MS), and Amazon have introduced their own CPUs for data centers starting in 2023, and Intel has also launched its own CPU called 'Xeon.'
Since withdrawing from the data center CPU business, Qualcomm has quietly revived the project by assembling a team of Apple chip designers starting in 2021 and is currently in discussions with Meta about this CPU.
Last week, Qualcomm signed a letter of understanding (LOU) with the Saudi Arabian AI corporation Humain to develop a customized CPU for data centers.