Ampere Computing and Qualcomm team up to create energy-efficient AI chips, launching a data center server aimed at reducing power consumption and outpacing rivals in the AI chip market.
Ampere Computing announced on Thursday its collaboration with Qualcomm to introduce a new product designed to reduce power consumption when operating artificial intelligence chips. Ampere, founded by former Intel president Renee James, uses Arm Holdings technology to create central processing chips utilized by companies like Oracle, Google, and others. The startup focuses on producing more energy-efficient chips compared to industry leaders Intel and Advanced Micro Devices.
Qualcomm, known for its dominance in the mobile phone chip market, has been attempting to penetrate the AI chip market for data centres since 2019 with its own power-efficient solutions. On Thursday, Ampere and Qualcomm revealed that they had combined their chips into a single data centre server.
Jeff Wittich, Ampere’s chief product officer, described this collaboration as the beginning of their joint efforts, hinting at future advancements beyond building server-level solutions. The Ampere-Qualcomm servers are designed to run AI models efficiently after they are trained, rather than competing directly with AI chip leader Nvidia, whose chips are used for training AI systems with vast amounts of data.
Ampere and Qualcomm also compete indirectly with Nvidia, as AI chips are often integrated into systems that combine multiple chip types. Their partnership aims to prevent rivals from gaining traction with customers, according to Jim McGregor, founder of Tirias Research. He emphasized that for both companies, the goal is to keep competitors out of the data centre market.
Additionally, Ampere announced the next generation of its central processing unit, which will feature 256 processing cores, an increase from the current chip’s 192 cores. The new chip, set for release next year, will be manufactured using Taiwan Semiconductor Manufacturing’s 3-nanometer process.
