In a time when advanced AI systems are redefining technology, Patrick Vandenameele, CEO of imec, is calling for a structural shift in semiconductor development. Speaking ahead of the ITF World 2026 conference in Antwerp, he outlined five technology transitions that he believes will transform the semiconductor industry over the next decade.
The Five Technology Shifts
These transitions include system-wide co-optimization in the angstrom era, the growing importance of silicon photonics, the strategic role of memory, the expansion of chiplets into edge computing, and the industrialization of quantum computing. Vandenameele stressed the need for co-optimizing technologies to maximize their potential. "What is underestimated is the gain that you can obtain by co-optimizing," he said. This strategy is becoming essential as the energy, memory bandwidth, and interconnect demands of AI systems surpass the limits of traditional scaling methods.
CMOS 2.0 and the Limits of Conventional Approaches
For years, the semiconductor technologies that support large-scale AI infrastructures were designed for different markets, including smartphones and traditional data centers. Vandenameele pointed out that adapting these existing technologies to meet AI workload demands has reached a saturation point. "We should optimize these technologies for the very high TOPS they need," he remarked, referring to the tera operations per second required by advanced AI systems.
Imec is promoting a new framework called XTCO, which emphasizes cross-technology co-optimization. This model advocates for a more integrated approach that connects compute capabilities with memory, packaging, photonics, interconnects, and AI model architectures. As Vandenameele noted, the complexity of modern AI systems requires deeper collaboration across the industry, including partnerships with hyperscalers and AI architects.
Advancements in Photonics and Memory
Among the shifts Vandenameele highlighted, silicon photonics is anticipated to have the most substantial impact by 2030. He pointed out that as data volumes and workloads grow, the efficiency of data movement becomes as crucial as processing power. "Electrical interconnects are increasingly hitting fundamental limits in bandwidth, energy efficiency, and scalability," he warned. The future is likely to see photonics moving closer to computation, with a shift from co-packaged to in-interposer photonics expected.
Memory technologies also face challenges that could impede AI growth. Vandenameele cautioned that without significant innovations in memory and interconnect solutions, the financial sustainability of AI scaling may be jeopardized. Current methods often depend on expanding existing architectures, leading to rising energy costs. He indicated that multi-agent AI systems could require up to 150 times more compute power for certain tasks compared to modern large language models.
The Quantum Frontier
While photonics may lead the immediate future, Vandenameele expressed surprise at the swift advancements in quantum computing. Initially expecting a longer timeline for silicon-based quantum technologies, he noted that many groups have quickly embraced semiconductor manufacturing principles following imec's breakthroughs in scalable silicon qubits. At ITF 2026, imec plans to reveal what it describes as the world’s first qubit created using ASML’s High-NA EUV lithography, marking a significant milestone.
Vandenameele believes that the success of quantum systems will depend heavily on the capabilities of the semiconductor ecosystem. He concluded, "Quantum will only succeed if it is developed as part of the same semiconductor ecosystem, leveraging advanced semiconductor skills and tools."
As the semiconductor industry navigates these significant shifts, the interaction between AI, photonics, memory, and quantum technologies will create a level of interdependence that requires collaboration across the entire ecosystem. Vandenameele's insights highlight the necessity of adapting to these changes to fully unlock the potential of future AI innovations.
