The ITF World 2025 conference, hosted by Imec, brought together global leaders in technology to discuss the future of hardware design amidst the fast-evolving landscape of artificial intelligence (AI). Luc Van den Hove, CEO of Imec, emphasized the need for “disruptive innovation over incremental change,” highlighting the urgency of global collaboration to address challenges in computing power, energy consumption, and chip architecture. As AI technologies become increasingly integral to applications ranging from autonomous driving to language processing, the demand for advanced semiconductor research and development has never been greater.
Pushing the Boundaries of Semiconductor Technology
As AI progresses at breakneck speed, traditional chip design cycles struggle to keep pace. Van den Hove noted that by the time an application-specific integrated circuit (ASIC) is ready for production, the desired AI model may already be outdated. He pointed out that while large language models (LLMs) like GPT-4 demand increasing computational resources, they face diminishing returns in performance, creating a paradoxical challenge for hardware development.
To counteract these challenges, Imec is advocating for a concept known as system-technology co-optimization. This methodology aims to integrate chip design with software architecture more tightly, enabling the creation of smarter hardware that can adapt dynamically to evolving AI workloads. “It’s not just about more chips; it’s about bridging the gap between algorithms, architectures, and technology platforms,” Van den Hove stated during the conference.
Innovations in Various Industries
Imec’s efforts span multiple sectors, with a significant focus on healthcare and automotive applications. In healthcare, the institute is advancing organ-on-chip platforms designed for AI-driven drug testing. This microfluidics technology facilitates the development of biologically accurate models of the human blood-brain barrier, which can test drug delivery systems for diseases like Parkinson’s and Alzheimer’s with unmatched realism.
In the automotive space, Imec has revealed updates on its chiplet-based architecture meant to power edge AI in vehicles. This modular approach allows for efficient processing of sensor data from various inputs like lidar, radar, and cameras, all within compact, high-performance electronic control units (ECUs). Van den Hove described these advancements as paving the way for “supercomputers in cars,” showcasing the potential of AI in enhancing vehicular safety and performance.
Quantum Computing Developments
Imec’s work also extends into the realm of quantum computing, where the focus is on scaling quantum hardware. Rather than creating entire systems from scratch, Imec harnesses its foundry capabilities to produce silicon qubits on 300mm wafers—an essential step towards practical quantum processors. According to Van den Hove, leveraging foundry-scale expertise is vital for developing scalable quantum technologies, marking a significant stride towards the realization of practical quantum applications.
Addressing Moore’s Law Limitations
Much of Imec’s roadmap is shaped by the challenges of compute density, power consumption, and memory bandwidth, which have become critical in extending Moore’s Law. The organization is pioneering a “CMOS 2.0” approach that incorporates vertical scaling and 3D stacking of transistors and chiplets. This architecture not only enhances functional customization but also seeks to optimize performance in increasingly compact footprints.
As higher power densities introduce thermal management challenges, Imec is innovating novel thermal management techniques, including direct liquid cooling and architectural designs that minimize heat concentration. These advancements are crucial for creating energy-efficient devices that can sustain high performance.
Memorizing the Future of AI
Imec is also exploring memory technologies crucial for AI applications. By focusing on the energy costs associated with data movement rather than computation, the company is investigating alternatives like ferroelectric and magnetic memory. Vertical Compute, Imec’s spin-off, is developing vertical memory arrays designed to address these challenges, hinting at a future where data transfer becomes significantly more efficient.
Collaborative Ecosystem and Future Outlook
Apple’s presence at the conference brought additional perspectives on the industry’s trajectory. Johny Srouji, Apple’s senior vice president of hardware technologies, spoke about the necessity of customized silicon in addressing AI demands. He noted that Apple designs chips tailored to specific experiences rather than merely creating general-purpose components. “Every microwatt matters,” Srouji emphasized, underlining the importance of energy efficiency in deploying billions of devices capable of performing complex tasks.
Srouji praised Imec’s collaborative model, which he described as a “rare combination of academic depth and industrial pragmatism.” He highlighted how Imec’s pilot lines have been instrumental in enabling Apple to validate advanced packaging and transistor concepts, ultimately supporting the rapid pace of innovation required in today’s tech landscape.
Europe’s Semiconductor Sovereignty
Beyond the technological advancements, the conference underscored Europe’s ambitions to secure its position as a leader in semiconductor development. With a €2.5 billion infrastructure expansion backed by the European Chips Act, the region is committed to enhancing its semiconductor R&D capabilities and establishing new hubs across countries like Belgium, Spain, and Germany.
Conclusion: The Path Ahead
The call to action from ITF World 2025 is clear: Europe must innovate, invest, and collaborate swiftly to thrive amidst the rapid advancements in AI, quantum computing, and semiconductor technology. As Van den Hove concluded, “Let’s supercharge our innovation engine and future-proof our prosperity together.” The collaborative efforts at the conference signify a collective drive towards technological excellence, with the potential to significantly influence the global tech landscape in the years to come.