반도체·하드웨어 · 리포트
Inference now drives two-thirds of AI workloads, which is changing where chips get deployed.
The shift toward inference moves chip demand toward inference-optimized accelerators and changes where GPU clouds need to place capacity.
업계 전문지Slicast · 2026년 11월 21일 18:28 UTC · 미국 · 출처: SiliconANGLE
중요도 50