OpenAI seeks alternatives to Nvidia chips

The AI development company OpenAI is dissatisfied with the performance of some of Nvidia's current AI chips for certain tasks and is seeking alternative hardware solutions, according to Reuters. The doubts primarily concern the chips' use for AI inference—the phase where models respond to queries—which has different requirements for memory architecture, latency, and performance than training models. OpenAI is therefore exploring alternatives from other providers, including established chipmakers like AMD and specialized AI hardware firms, to potentially achieve higher performance or better energy efficiency for inference, reflecting a broader trend of major AI developers diversifying their hardware strategies amid rapidly growing computational demands