Hot French startup ZML releases free product to speed inference across lots of AI chips

French AI startup ZML, endorsed by Turing Award winner Yann LeCun, has released a free product called ZML/LLMD, an LLM inference server designed to speed up inference across a variety of AI chips, including those from Nvidia, AMD, Google, Apple, and Intel. Founder Steeve Morin told TechCrunch that the goal is to break vendor lock-in by allowing different chips to run open-source large language models at maximum speed. This could be a market disruptor as AI costs rise, offering enterprises and clouds the option to use a mix of potentially cheaper or more energy-efficient chips. Morin noted that the software could help novel AI chipmakers, many from Europe, but emphasized that ZML has a good relationship with Nvidia. The inference space is highly competitive, with rivals like Baseten, Inferact, and RadixArk, but Morin says ZML's ambitions are broader, including co-designing silicon. The Paris-based startup has a lean team of 20 people and has raised $20 million from venture firms including 20VC and Xavier Niel's Kima Ventures, allowing it to move fast with more releases planned.