[OpenAI Researcher Praises AMD and Cerebras AI Inference Solution, Predicts 5x Improvement in Performance per Watt] According to DeepTech TechFlow, on July 29, OpenAI researcher Jeffrey Wang described the AI inference solution jointly developed by AMD and Cerebras as 'incredible.' The solution integrates AMD Helios racks with Cerebras wafer-scale engines to handle the prompt processing and token generation stages separately, with an estimated increase of 5x in tokens per second per watt. This technical approach optimizes inference performance through heterogeneous computing, reflecting the ongoing focus of leading AI labs on reducing inference costs.