167 points61 comments
Qualcomm launches 150W AI accelerator card for generative models
What is s?
The Cloud AI 100 Ultra is a PCIe card that delivers up to 150 watts of compute for generative AI workloads. It includes 128 GB of memory with a 548 GB/s bandwidth to accelerate inference.
Who it is for
Enterprises and researchers running large language models can offload inference to a single card and reduce cloud costs.