AI Inference Startup Daisytuner Raises €6M in Seed Funding

AI Inference Startup Daisytuner Raises €6M in Seed Funding

The race to develop faster and more efficient AI infrastructure is opening up new opportunities. With organisations increasingly using AI models in production, the ability to execute them efficiently on various hardware platforms is becoming a huge differentiator. German deep tech startup Daisytuner has raised €6m in seed funding to advance its AI compiler and inference solution.

The funding round, announced on the 16th November 2026 was led by Elaia and LEA Ventures with participation from Amadeus APEX Technology Fund and Seedcamp. The company plans to use the funding to grow its footprint in enterprise, grow its engineering team and improve its software platform.

Founded in 2024 and based in Darmstadt, Germany, Daisytuner recognises that whilst there are many powerful hardware options available, you still have to optimise your software to utilise what the chips can offer. This can be both expensive and complex.

Why Optimising AI Inference is important

The process of inference involves running a trained AI model to produce predictions, answers, images etc. Whilst there is a lot of focus on training AI models due to the intense compute required, inference is becoming increasingly important.

Every time you interact with an AI assistant or recommendation engine or generative application you are using some compute resource. As organisations start to see more and more AI requests, having an efficient infrastructure will impact how much they are spending.

Different processors have different architectures and requirements. Your AI model may not have the same performance on a different processor.

Traditionally this has been done by specialist engineers. Daisytuner want to simplify this by automating much of the performance engineering so more power can be extracted from existing hardware.

How it works

The technology offered by the company includes an open source compiler, which takes code from frameworks such as PyTorch and converts it to native code. This can be optimised for a range of different hardware architectures such as Nvidia and AMD GPUs.

The system then uses AI agents to test the different optimisations on real hardware. The optimisation is then verified and added to a performance database. This means improvements can be reused for similar workloads.

This reduces reliance on engineers and makes it easier to move AI applications between computing platforms.

The company also mentioned their Daisyflow inference platform for production. This allows users to run language, vision, speech and recommendation models. It enables enterprises to run their AI workloads within their own infrastructure.

The platform can use familiar APIs and open source frameworks.

Early Performance Improvements

One of the most impressive performance improvements the company has made is with Alibaba’s Qwen 2.5 model. Using Daisyflow, the model can run up to 18% faster on AMD Instinct GPUs compared to the popular vLLM inference engine.

Related:

Whilst this is an early benchmark from the company it shows the potential gains you can get by optimising your software.

Even small improvements can save organisations some money as they can run thousands/millions of AI requests.

Beyond Nvidia

Despite Nvidia dominating the AI compute world, the industry is starting to diversify. AMD and new semiconductor companies are producing new processors.

However to succeed they need to provide software solutions that will let developers run their applications on the new hardware with minimal changes.

This is where Daisytuner identifies an opportunity. By helping to run AI models efficiently on different processors they can reduce reliance on the hardware. This makes alternative compute platforms more viable.

This could also benefit organisations who want more flexibility with their infrastructure costs, hardware and data.

Founders, Investors & Growth

The company was founded by Lukas Trümper and Adrian Schmitz. They have over a decade of experience researching high performance computing and compilers. Their team contains researchers and engineers from well known institutions such as ETH Zurich, TU Munich, RWTH Aachen, TU Darmstadt and Lawrence Livermore National Laboratory.

Investors believe as AI becomes more adopted having more efficient compute will become increasingly important. The funding will allow Daisytuner to increase their customer base by providing Daisyflow to more enterprises, hire new staff and increase the capabilities of their compiler to support new chip architectures.

What’s next?

The €6m seed round shows an exciting time for the AI industry. We may see new innovations not only in developing more powerful processors but also making existing hardware more efficient.

The company will need to show consistent improvements across different chips, grow its customer base in enterprises and compete with other inference platforms.

If they can achieve their goals they could become an important part of the growing AI infrastructure space. As organisations look for faster, cheaper and more flexible ways to use AI, software that can help maximise the potential of their compute could become just as valuable as the hardware.

Sources

  1. Amadeus Capital – Daisytuner Raises €6M in Seed Funding
  2. Tech.eu – AI Inference Startup Daisytuner Raises €6M Seed

1 thought on “AI Inference Startup Daisytuner Raises €6M in Seed Funding”

  1. Pingback: AI Payroll Startup Valeria Raises €5.5M in Seed Funding

Leave a Comment

Your email address will not be published. Required fields are marked *

Scroll to Top