French startup ZML releases free software to speed AI inference across multiple chips

French startup ZML released free software to accelerate AI inference across various chips, aiming to reduce costs and break vendor lock-in. The new ZML/LLMD software supports multiple large language models on diverse hardware like Nvidia, AMD, and Google TPUs, promising peak performance and efficiency. This move could disrupt the AI chip market, offering enterprises flexibility and potentially aiding European chipmakers, though Nvidia remains a dominant player.