Photo by Andrey Matveev on Pexels

Qualcomm is officially entering the competitive AI data center market with its AI200 and AI250 chips, designed specifically to accelerate AI inference tasks. The announcement triggered a positive reaction from investors, evidenced by a jump in Qualcomm’s stock price, signaling confidence in the company’s ability to compete in the booming AI infrastructure sector.

Qualcomm’s strategy involves two distinct chips: the AI200, a more immediate offering equipped with 768 GB of LPDDR memory per card, targeting memory-intensive large language model (LLM) workloads. The AI250, planned for release in 2027, will feature a near-memory computing architecture, promising substantially improved memory bandwidth. Qualcomm is focusing on minimizing total cost of ownership, highlighting the 160 kW power consumption per rack and the use of direct liquid cooling.

A key partnership with Saudi AI firm Humain includes a commitment to deploy 200 megawatts of Qualcomm’s AI data center chips, potentially generating $2 billion in revenue and offering vital validation for Qualcomm’s entry into the AI infrastructure landscape.

Qualcomm’s AI software stack provides compatibility with popular machine learning frameworks and features “one-click deployment” of models from Hugging Face. While facing stiff competition from industry leaders like Nvidia and AMD, market analysts suggest that the rapidly expanding AI market can accommodate several successful players.