Qualcomm is turning parts from cellphone chips into AI chips to rival Nvidia

"Qualcomm is launching a pair of new AI chips in an attempt to challenge Nvidia's dominance in the market. On Monday, Qualcomm announced plans to release its new AI200 chip next year, followed by the AI250 in 2027 - both of which are built on the company's mobile neural processing technology. The new chips are built for deploying AI models, rather than training them."

"As reported by CNBC, Qualcomm's AI-focused processors are based on the company's Hexagon neural processing units that power AI features in its chips for mobile devices and laptops. CNBC reports that Qualcomm's AI processors can work inside a rack with up to 72 chips functioning as a single computer, similar to Nvidia and AMD's GPUs. Qualcomm's AI200 chip features 768GB of RAM, along with performance optimized for AI inference."

"Qualcomm's AI200 chip features 768GB of RAM, along with performance optimized for AI inference. Meanwhile, the AI250 will come with "a generational leap in efficiency" that the company claims will allow for "much lower power consumption." Humain, the AI company that exists under Saudi Arabia's Public Investment Fund, has already announced plans to use the AI200 and AI250 to power computing systems, building on a partnership to develop AI datacenters across Saudi Arabia."

Qualcomm will introduce two new AI chips: the AI200 next year and the AI250 in 2027. Both chips leverage the company’s mobile Hexagon neural processing technology and are designed for AI model deployment (inference) rather than training. The AI200 includes 768GB of RAM and performance optimizations for inference. The AI250 promises a generational efficiency leap enabling much lower power consumption. The processors can be deployed in racks with up to 72 chips functioning as a single computer, similar to GPU clusters. Humain, owned by Saudi Arabia’s Public Investment Fund, plans to use both chips for AI datacenter builds. The move represents a notable shift from Qualcomm’s historical focus on mobile and telecom processors.

Read at The Verge