Semiconductor News & Analysis Feed

Generative AI applications are expanding rapidly, making computing costs a growing bottleneck to commercial AI deployment. The AI accelerator market, long dominated by graphics processing units (GPUs), has increasingly explored specialised architectures in recent years. Tensor Processing Units (TPUs), optimised for the matrix operations used by AI models, have again drawn industry attention.

Nvidia's plan to launch Rubin CPX, an inference-focused graphics processing unit for its Vera Rubin platform, has become increasingly uncertain as supply-chain activity around the product appears to have stalled, according to The Elec.

For the past three years, graphics processing units, or GPUs, have dominated the artificial intelligence boom. But Johnny Shen, chairman of Alchip Technologies, believes the next phase of the market may belong to something more specialized.