French startup ZML is launching a free product to speed up inference across many AI chips

Nvidia’s days of unparalleled market dominance are not over, but challenges and options are emerging from all directions.

ZMLa leading French startup in the field of artificial intelligence He supported By Turing Award winner Yann LeCun, it was released Inference-Performance programs that allow A variety of large open source language models To run on A variety of chips – Including Nvidia’s, AMD’s, Google’s TPU, Apple Metal, and Intel Arc.

with zml/lmdthe newly launched LLM inference server, the company’s ambition is to break down existing silos and make different segments available for AI use cases at the fastest speed available, and sometimes faster, ZML founder Steve Morin told TechCrunch.

With the integration of AI into our work and daily lives, improving reasoning – also known as claims processing – is accelerating Raining model These companies are important, but they often feel incomplete behind the scenes, with software and architectural barriers that lead to vendor lock-in, Morin said.

The promise of peak performance across a variety of chips is a technological breakthrough, but it could also be a market disruptor, amid growing concerns about the costs associated with artificial intelligence.

ZML hopes to provide enterprises and clouds with the option to use a mix of chips, some of which may be less expensive or consume less power. “The idea is to give people back the ability to create their own system and achieve real efficiency gains that allow (AI) to be deployed,” Morin said.

Such software assistance could help new AI chip makers, many of whom happen to be from Europe, Morin noted To Axel, crisp, Kalray, OLX, S.You, Siberl, SpinNNcloudand VSORA. But what matters to him more than their home region is that ZML can work with them on “things that have never been done before anywhere in the world.”

This doesn’t mean Morin is pessimistic about Nvidia. It’s not like thatThis is partly due to the current supply. He told TechCrunch that ZML had a good relationship with the AI ​​chip giant, which it did Getting ready For high inference.

Heuristics has been an area of ​​such intense investment that the trend has been hailed as “Gold rush conclusion“.So ZML has competition like Basset houndwhich was recently valued at $13 billion; sterilefrom the creators of the open source project vLLM; Besides RadixArkthe The trading company behind SGLang.

Both vLLM and SGLang partly compete with LLMD, but Morin’s ambitions for ZML cover a broader range. “We’ve gotten to the point where we’re involved in silicon design,” he said. He also praised ZML’s lean team of 20 people as the reason the Paris-based startup has been able to move quickly, with more releases in the plans.

It also helped that this small team was well funded for its size. Thanks to his proven track record as Vice President of Engineering ZinleWhich was acquired by Snapchat to nine figures in 2017Morin has raised $20 million from venture firms including Harry Stebbings 20VC, >commit, AALVC, Drysdale Ventures, Xavier Niel’s Kima Ventures, Kindred Capital, LocalGlobe, and Puzzle Ventures.

In contrast, ZML’s first public project was one that focused on inference Machine learning framework Released in 2024 and Updated MarchZML/LLMD is not open source. But it is launched as a free product with the aim of recognizing usage. “I prefer to measure and (then generate revenue) where it is most effective without stupidly stunting my growth because I was too greedy from the beginning,” Morin said.

It’s too early to know when ZML/LLMD might become a paid product, and what its adoption will look like. But the startup’s cap table confirms that other founders are taking notice, including Dagger and Docker founder Solomon Hykes, Hugging Face’s Clément Delangue and Julien Chaumond, as well as LeCun, now with AMI Laboratories. This also builds the argument that AI startups are in Europe It is now possible to build from home. “I couldn’t do ZML anywhere but Paris,” Morin said.