At its recent ITF World 2025 conference in Antwerp, Belgian research institute Imec delivered a pointed message: as AI accelerates, the world needs new thinking - not just in software, but in the hardware ecosystems underpinning it
www.computerweekly.com, May. 27, 2025 –
Imec CEO Luc Van den hove opened ITF World 2025 by calling for “disruptive innovation over incremental change”. He underscored the need for tighter global collaboration to overcome looming bottlenecks in compute power, energy use and chip architecture. “By supercharging our innovation engine, we can future-proof our prosperity,” said Van den hove.
The message resonated across sectors. Artificial intelligence’s (AI’s) march into everyday applications – from language models to autonomous driving – is raising the stakes for semiconductor R&D. Europe, facing renewed geopolitical pressure to secure its technology stack, is betting on advanced research and industrial pilot lines as strategic assets.
Van den hove highlighted a paradox: AI is moving so fast that traditional chip design cycles can’t keep up. By the time an application-specific integrated circuit reaches production, the target model may already be obsolete. Large language models (LLMs) like GPT-4 are hitting diminishing returns in performance despite rising computational demands.
To address this, Imec is pursuing what it calls system-technology co-optimisation: tighter integration of chip design and software architecture. The goal is to build smarter, more efficient hardware capable of adapting to dynamic AI workloads. “It’s not just about more chips,” said Van den hove. “It’s about bridging the gap between algorithms, architectures and technology platforms.”
This effort spans several fronts. In healthcare, Imec is advancing organ-on-chip platforms for AI-driven drug testing. The company’s microfluidics technology now supports biologically accurate models of the human blood-brain barrier, allowing researchers to test drug delivery for diseases such as Parkinson’s and Alzheimer’s with unprecedented realism and scale.
In automotive, Imec presented updates on its chiplet-based architecture designed to power edge AI in vehicles. The modular approach allows sensor data from lidar, radar and cameras to be processed on compact, high-performance electronic control units (ECUs). “We’re talking about supercomputers in cars,” said Van den hove. Click here to read more