What matters in AI.

Subscribe

Daisytuner gets €6 million to move AI code off Nvidia chips

The start-up tells that its compiler makes the same code operate on Nvidia, AMD and new chips, but it gave one test result only.

Daisytuner, a start-up in Darmstadt, Germany, got €6 million in seed funding. The founder made the company in 2024. Elaia and LEA Ventures led the investment. Its open source compiler changes AI code, and the code then operates on many types of chips. The company tells that Qwen 2.5, a language model, gives 18% more speed than vLLM on AMD Instinct chips.

How it works

The compiler gets code from an AI framework and makes native code for each chip. AI agents test changes on the hardware. The compiler checks each change and keeps the good changes in a database for use at a subsequent time.

The numbers

The 18% result is from the first tests, and the company gave it. It gave no other test data. Tenstorrent and Q.ANT buy the software. The team has 12 persons.

What is not known

The company is near its first enterprise contracts. The founder tells that Modular is the primary competitor. In June 2026, Qualcomm agreed to buy Modular for $3.9 billion. It is not known if 12 persons can supply the same software.

Insight: Software, not chips, is the fight over Nvidia

Qualcomm agreed to buy Modular for $3.9 billion, and Daisytuner at this time gets funding for almost the same work. In October 2026, Nvidia made open source the platform that gives GPU resources to AI work. This reads as a fight to control the software layer between models and chips, because that layer chooses which chip a buyer can use. Buyers get an exit from one vendor if these tools operate.

Sources

  1. Elaia and LEA Ventures back Daisytuner’s €6M seed to move AI code off Nvidiatechfundingnews.com
AI MATTER · NEWS · AI MATTER · NEWS ·16 NOV2026

Posted

Tags

Learn the terms in this story

More in Deals

All Deals news