The Netherlands and Germany will jointly invest in developing new AI chips. The recently established Dutch innovation organization NADI and its German counterpart SPRIND are making 40 million euros available for this purpose. Small research teams are tasked with significantly accelerating chip design using AI.
The program will run for 20 months and focuses on processors for both training and running AI models. The organizers see particular potential for more efficient hardware in the latter area, inference. NADI co-founder Jelle Prins told Reuters that powerful Nvidia chips are not always the most efficient solution for this purpose. He compares it to doing grocery shopping in a truck.
At the same time, AI must accelerate the design process itself. Developing new chips often takes years. According to Jano Costard, who is responsible for the so-called “challenges” at SPRIND, the goal is to accelerate that process by several orders of magnitude.
Dutch chip expertise
According to NADI, the collaboration aligns with the Netherlands’ strong position in the chip industry. A vast ecosystem of technology companies and research institutions has emerged around ASML. Germany, in turn, brings expertise in research and industrial production.
NADI was established this month and has 500 million euros in government funding. The organization is modeled after the U.S. DARPA, which funds high-risk research projects that could lead to technological breakthroughs in the long term.
The Netherlands is not alone in this endeavor. Germany established SPRIND in 2019, while the United Kingdom has had ARIA since 2023. France has had the defense-focused innovation agency AID since 2018.
Less dependent
Behind the German-Dutch collaboration lies the desire to make Europe less technologically dependent on the United States and China. In AI, this is especially relevant in the chip sector, where American companies hold a dominant position.
SPRIND is also investing 125 million euros in a separate program that will ultimately result in three European labs for frontier AI. The collaboration with NADI now adds a hardware component to this effort: not only developing European AI models, but also striving to design the chips on which those models run more efficiently and quickly.