Cerebras launches disaggregated inference with AMD Instinct GPUs
Cerebras has launched a disaggregated inference solution that pairs AMD Instinct GPUs for prompt processing with Cerebras Wafer Scale Engines for output generation. The company reports that this production setup delivers five times more tokens per megawatt compared to single-system configurations. The solution is now available on the Cerebras Inference Cloud and for infrastructure partners through reference architectures and software stacks. Open article →