Fundraise

Daisytuner raises €6M to run AI models faster on any chip

What's the deal? DaisytunerDealroom has a profile for this one. Try Dealroom →, a Darmstadt-based startup automating AI model performance engineering across chips, has raised €6 million in seed funding. Elaia and LEA Ventures co-led the round, with Amadeus APEX Technology FundDealroom has a profile for this one. Try Dealroom → and Seedcamp participating.

What's the endgame? Founded in 2024 by Lukas Trümper and Adrian Schmitz, the company pairs a self-learning compiler with a patented optimisation platform. Its open-source compiler, DOCC, translates code from frameworks such as PyTorch into native code for each target chip — from Nvidia and AMD GPUs to newer accelerators.

The approach replaces manual performance tuning with AI agents that test optimisations on real hardware around the clock, storing every verified result in a performance database for reuse on similar workloads.

What's new? Alongside the raise, Daisytuner launched Daisyflow, a production-ready inference engine that serves language, vision, speech, and recommendation models through familiar APIs. It can run inside a customer's own infrastructure, giving control over data and costs. The company says Daisyflow ran Qwen 2.5 up to 18% faster than the industry-standard vLLM on AMD Instinct GPUs.

Why now? AI companies and hyperscalers are shifting from adding capacity to optimising existing infrastructure. Each new chip or model generation risks breaking earlier optimisations, leaving much hardware running below its potential.

"Compute is the scarcest resource in AI, and most of it currently sits behind a single vendor's software stack," said Sébastien Lefebvre, a partner at Elaia. "Daisytuner lets the same code run at full performance on the silicon you choose, without a rewrite."

By the numbers: The €6 million round sits in the top 5% of all-time seed rounds for enterprise software startups in Germany, by amount. The money will scale Daisyflow across enterprises, grow the engineering team, and broaden compiler support for emerging chip architectures.

The signal: The round fits a broader push to loosen Nvidia's hold on AI computing by making models portable across chips. Comparables include Dutch chipmaker Euclyd's €200 million Series A and inference-compiler startup LuminalDealroom has a profile for this one. Try Dealroom →'s $25 million Series A — signs of investor appetite for both alternative silicon and the software to use it. Daisytuner's bet is software-first, but funding and compatibility claims alone don't prove it can displace established software stacks.

Read more: EU-Startups

Image credit: Generated with Gemini

More top stories