AMD and Cerebras Join Forces to Develop Ultra-Low-Latency AI Inference Platform

AMD and Cerebras Systems are creating a computing platform that combines AMD Instinct accelerators and Cerebras WSE chips based on SRAM memory for low-latency inference of AI agents. The solution, available in Cerebras Cloud this year, generates five times more tokens per watt and requires fewer chips than similar systems from competitors.

Cerebras Systems NVIDIA