Performance, efficiency, and cost analysis of wafer-scale AI accelerators vs. single-chip GPUs | Publicación