ZML's free inference server lets enterprises drop Nvidia's hardware stranglehold

French startup ZML released LLMD on July 8, 2026, a free software layer that runs AI models on Nvidia, AMD, Google TPU, Apple Metal, and Intel Arc chips interchangeably. The move addresses enterprise and cloud providers' growing need to diversify hardware suppliers and avoid vendor lock-in. Turing Award winner Yann LeCun endorsed the approach. LLMD competes directly with commercial engines vLLM and SGLang.