Author: Evolving AI - Bewertung: 10x - Views:48
Qualcomm just entered the AI data center war, and this could be one of the biggest challenges yet to Nvidia’s dominance. At its Investor Day in Manhattan, Qualcomm revealed its new Dragonfly data center portfolio, including the C1000 server CPU, AI200, AI250, AI300 accelerators, High Bandwidth Compute memory, networking, and custom silicon. The C1000 is expected to deliver more than 250 cores, PCIe Gen 7 connectivity, and over 2x better performance per watt compared to competing server CPUs. But the bigger story is who is already involved. Meta signed a multi-generation agreement to deploy Qualcomm CPUs in its next-generation server fleet. Microsoft’s Satya Nadella endorsed Qualcomm’s HBC memory architecture for Azure. Saudi AI company Humain committed to 200 megawatts of Qualcomm accelerator racks. Qualcomm also announced a $3.9 billion acquisition of Modular, a software company that could help break Nvidia’s CUDA lock-in by letting AI models run across different chips more easily.
So can Qualcomm really challenge Nvidia, AMD, Intel, and the hyperscalers in AI data centers? Or is it too late to enter the most brutal chip war in tech?
#Qualcomm #Nvidia #AIChips #DataCenter #Dragonfly #CUDA #ArtificialIntelligence
SOCIAL SHARE CARD GENERATOR