Qualcomm mentioned the AI200 and AI250 chips have been optimized for AI inference — the method of operating already-trained AI fashions to generate responses — quite than the coaching section that requires extra computing energy and is dominated by Nvidia’s highly effective GPUs.
Shares in Qualcomm have been up as a lot as 20 % in Monday buying and selling on Wall Road.
The San Diego-based firm is greatest identified for smartphone and private laptop processors and is venturing into new territory with the discharge.
Qualcomm mentioned the AI200, slated for business launch in 2026, affords 768 gigabytes of reminiscence per card and makes use of direct liquid cooling to handle warmth in densely packed server racks that every eat as much as 160 kilowatts.
The corporate mentioned the AI250, anticipated in 2027, could be ten occasions simpler in reminiscence bandwidth than present merchandise in the marketplace whereas consuming much less energy.Nvidia, the poster baby for the generative AI increase, at present dominates the AI chip market, pushed by its H100 and newer H200 GPU accelerators utilized in coaching and operating AI fashions in huge purpose-built information facilities.AMD, the second-largest participant, has been gaining floor, positioning itself as a challenger to Nvidia’s dominance.
Each firms have not too long ago entered into preparations with OpenAI, the creator of ChatGPT, because the trade seeks to ramp up the huge infrastructure wanted to assist the AI frenzy.
This has raised some fears that Wall Road might be within the throes of an AI inventory bubble paying homage to the web increase and crash skilled within the late Nineties and 2000.
Nonetheless, firms are largely rewarded for his or her connection to the AI bonanza. AMD’s shares rose 35 % on the day of its announcement with OpenAI, the world’s most useful personal firm, price $500 billion.
