The world is entering an era where data, connectivity and computational power define economic growth and technological sovereignty. At the heart of this transformation lies the semiconductor industry - a sector that underpins nearly every modern device from smartphones to satellites and from renewable energy systems to quantum computers.
Is the end of Moore’s law approaching and what should we do about it?
As traditional CMOS technologies approach their physical limits, Moore’s law can no longer be supported by only scaling down the transistor. There are both the space and need for innovations beyond the transistor design in, e.g. materials, designs, packaging as well as new computing paradigms transforming the R&D&I process. Emerging concepts aim to push beyond CMOS, enabling new device types and scaling paths for next generation semiconductors.
The need for new computational paradigms has never been greater. There are several candidates and rivals for providing the future computing solution: neuromorphic, photonic and quantum computing to name a few. These novel computing approaches offer routes to more efficient computing architectures and step-changes in performance.
Of these, quantum computing combined with AI - especially hybrid quantum classical approaches - are emerging as powerful near future tools to tackle some of the most complex materials and device challenges in the field. These are highly relevant for the semiconductor industry. However, as AI models grow in size and capability, their energy consumption is becoming unsustainable, making sustainable AI is a definite development need.
Hybrid quantum–classical algorithms unlock new possibilities
Many industrially relevant problems, from molecular simulations and material discovery to complex optimisation tasks, are computationally intensive and scale poorly on classical hardware. Hybrid approaches combine classical high-performance computing (HPC) with quantum processors, leveraging the strengths of both.
Use cases include Quantum Machine Learning (QML) for classification tasks, Quantum Approximate Optimisation Algorithm (QAOA) for combinatorial problems, as well as Variational Quantum Eigensolver (VQE) for molecular and materials simulations, to name a few.
These early applications reveal how hybrid pipelines can accelerate exploration of high dimensional materials spaces and deliver best-of-both-world for complex simulations. These are very relevant for the semiconductor industry and for the innovations beyond Moore’s law.
Semiconductor innovation starts with materials
The semiconductor industry’s most pressing challenges are fundamentally materials driven.
Today’s manufacturing processes require better control, new doping methods and a deeper understanding of phenomena such as superconductivity and defect formation.
A game changer: Quantum AI for materials modelling
Quantum AI offers significant advantages in materials modelling:
- faster exploration, exploitation and optimisation across chemical, structural, and synthesis design spaces,
- hybrid models that improve simulation accuracy and offer orders of magnitude speedups,
- quantum-entangled feature spaces that capture features difficult to address classically, such as strongly correlated behaviours, and
- better data efficiency through high-dimensional quantum kernels.
These capabilities are particularly important for materials where classical methods such as DFT fail or become computationally impractical.
Applications across semiconductor materials and processes
At VTT, we have discovered three use cases especially for quantum AI in materials modelling.
Modelling ALD and other process chemistries, in this area, we are working to improve accuracy beyond DFT methods, enabling to reveal new reaction pathways and material properties in processes such as atomic layer deposition (ALD). This opens the door for improved predictions in defect formation, nucleation, transport effects, dissolution and site-specific interactions, as well as greatly accelerating the development of ALD solutions.
In Compound Semiconductor Modelling, we are looking into possibilities on how quantum algorithms can improve predictions of band gaps, localised defect states and traps, electron-phonon interactions, phase transitions and interface properties. QML) has already shown superior accuracy in predicting GaN Ohmic contact resistance using small datasets an important milestone for semiconductor manufacturing.
Research into superconducting materials: Superconductors involve strong electron-electron interactions, making them notoriously difficult for classical simulations. Quantum algorithms allow the use of strongly correlated models, such as the Hubbard model, to study the mechanisms behind superconductivity. This is essential for advancing quantum technology hardware, where the performance of materials limits qubit quality. The Fermi-Hubbard model remains one of the greatest unsolved challenges of condensed matter physics. Despite decades of research, classical simulations are very limited and cannot reveal all aspects of its superconducting phases. Recent quantum experiments indicate that the first signs of key correlated phenomena, such as antiferromagnetism and Mott transitions, are already visible on today’s hardware. We are now investigating scalable algorithms to search for superconducting behaviour under realistic noise conditions.
Unlocking new possibilities
The semiconductor industry is reaching a new stage when following Moore’s law becomes more challenging. Novel approaches are needed.
Quantum computing is no longer a technology of the distant future -it can already be applied for modelling materials relevant for the semiconductor industry. When combined with AI and HPC, quantum tools offer unprecedented opportunities to solve problems that have previously been out of reach.
