We have become used to the idea that powerful AI runs somewhere far away in a data center. But a team of researchers at Eindhoven University of Technology (TU/e) in the Netherlands believes the next step could bring AI much closer to home: running it locally on our own devices. That could mean less reliance on data centers, greater privacy, and more efficient energy use.
The TU/e researchers have developed a new chip within the European CONVOLVE project that makes it possible to run AI computations more efficiently and closer to the user, for example in smartphones, laptops and smart home devices. The technical term is edge computing.
The rapid growth of AI is driving an increasing demand for computing power, and therefore for energy and data-center capacity. By running some of these computations locally, data does not have to travel back and forth to a data center. This could reduce energy use while making applications faster and more privacy-friendly. At a time of grid congestion and growing concerns about the impact of data centers on energy and water use, the technology arrives at an opportune moment.
One of the things the EU-funded and TU/e-led CONVOLVE team changed, was the chip design methodology. They developed specifically to make energy-efficient chips, optimizing AI algorithms, processor architecture, memory, and circuits together, rather than designing these components separately.
With research partners in Leuven, Delft, Zürich, and many industry partners, each bringing unique expertise to the project, they managed to design and build the chip they all envisioned. The resulting prototype chips were then tested with real AI workloads, the measurements proving that they succeeded in their goal.
The CONVOLVE team presented its results at the European Solid-State Electronics Research Conference (ESSERC) in Spain last week. The conference brought together leading researchers and representatives from the semiconductor and electronics industry to discuss the latest developments in chip technology.
PhD researcher Rick Luiken presented at the ESSERC and is available for interviews. He can explain how the technology works, why moving data between memory and the processor consumes so much energy, and what this development could mean for the way we use AI.
Assistant Professor Manil Dev Gomony led the research and co-supervises Rick Luiken with Associate Professor Sander Stuijk. He is also available for interviews. Please get in touch with the PIO to set up an interview with either one of the researchers.
Full article: https://www.tue.nl/en/news-and-events/news-overview/08-09-2026-bringing-powerful-ai-computing-to-the-edge
Experimental study
Not applicable