1.0

Investing in the Research and Innovation topics described in this document will not only translate into a more competitive and sovereign European ECS value chain, but it will also boost European industrial competitiveness across all sectors.

Electronic components and systems, by their inherent nature, are the result of interdisciplinary research and engineering. These require competencies in diverse technology domains, including process technology, equipment, materials and manufacturing, electronics, and embedded software, as well as cross-sectional technologies such as edge computing, artificial intelligence, high-speed connectivity, and cybersecurity.

As a result, ECS research needs to be interdisciplinary to benefit from the multiple available sources of innovation, as well as research-intensive and market-oriented. This will ensure forthcoming ECS innovations will be of strategic value for Europe and boost its industrial competitiveness in all its value chains, and help building the strong industrial base essential for European strategic autonomy.

With the Chips Act, the Key Digital Technologies Joint Undertaking (KDT JU) has transited to the Chips JU, with a mandate extended to capacity building activities and related research and innovation activities of four operational objectives of the Chips for Europe Initiative, as set out in the Chips Act article 4:

• building up advanced design capacities for integrated semiconductor technologies;
• enhancing existing and developing new advanced pilot lines across the Union to enable development and deployment of cutting-edge semiconductor technologies and next-generation semiconductor technologies;
• building advanced technology and engineering capacities for accelerating the innovative development of cutting-edge quantum chips and associated semiconductor technologies;
• establishing a network of competence centres across the Union by enhancing existing or creating new facilities.

All the new initiatives introduced by the Chips Act have been conceived to boost European competitiveness in the ECS market and ECS-based application domains, guiding investments and resources towards strategic areas of semiconductor technology that are most critical for the European industry's competitiveness. This SRIA contributes to the identification of these critical areas and of the associated challenges that will characterise their development in the next ten years. The role of the SRIA is fundamental to identify the key starting points for RD&I and to setup multi-annual synergies with the pilot lines that will focus on more mature prototyping. The SRIA will also potentially provide the elements for the “first-of-a-kind facilities” envisioned by Pillar 2: the synergies between these three steps represent the key strategy to increase European capacity building, to boost competitiveness, and to anticipate, prevent and effectively manage future crises.

The benefits for European strategic autonomy of the development of innovative ECS technologies focused on security, safety, reliability, dependability and privacy will be discussed in the section “strategic advantages for the EU”. European technology-based, secure, safe and reliable ECS, combined with European AI solutions, are critical to securing global leadership and strategic autonomy in key areas such as Information and communications technology (ICT) and to ensure compatibility with EU values.

These innovative technologies will simplify the implementation of the European Strategy for Data¹,², and ensure security, privacy-by-design and strategic autonomy all along the industrial and digital value chains.

Threats to Europe’s strategic autonomy are to be found in the microelectronics value chain, and then downstream in the component user segments of the electronics industry. In this context, the Major Challenges identified by the ECS SRIA will help develop innovations in secure, safe and reliable ECS technologies for creating EU-based/-made solutions in the key European application domains of:

• Aerospace, defence, security.
• Automotive, transportation.
• Machinery, robotics, electrical equipment, energy.
• Communications, computing.
• Energy generation, distribution and grids
• Healthcare and well-being

and other domains.

The Chips Act aims at consolidating and strengthening European strategic autonomy, extending and boosting the KDT with a new specific focus on the upstream of the ECS value chain. Reducing the dependency on non-European suppliers and avoiding future shortages represent the main steps towards strategic autonomy: reaching the independence on non-European suppliers for critical ECS is crucial to ensure European industries and key application domains to have a stable supply of semiconductors, even during future global disruptions or geopolitical tensions. Technological leadership and strategic autonomy are the primary ingredients required to:

• Control ECS manufacturing to produce solutions designed to meet stringent security requirements; an essential aspect for maintaining secure digital infrastructures and key applications (e.g. automotive, defence, energy, etc.).
• Ensure resilient supply of semiconductors produced in Europe to support the resilience of critical infrastructure, including energy, transportation, and healthcare systems, which rely on advanced chips.
• Achieve a greater control over the technology infrastructure that underpins the European digital economy, ensuring that critical data and systems remain under European jurisdiction.
• Enhance Europe's economic competitiveness, leading to the creation of high-tech jobs, stimulating innovation, fostering economic growth, and making Europe less dependent on external economic forces.

European strategic autonomy will also require the sustainability and resilience of the entire ECS value chain since the development of innovative technologies focused on sustainability and the Green Deal will support ambitions to achieve a green, resilient and competitive Europe.

Moreover, the serious effects of climate change that we are experiencing daily and the current geo-political situation, which highlights our dependency on non-European fossil energy providers, further reinforce the need for Europe to accelerate its transition to climate neutrality by 2050.

This challenge must be perceived as an opportunity to create a new environment for boosting innovative aspects of technology and business models through achieving the following:

• Relying extensively on ECS-based technologies and digitalisation as key factors for lowering our global energy footprint at all levels of the economy, and by placing sustainability at the heart of combined digital and green transitions.
• Positioning the European players in hardware as front-runners in sustainability to secure a wider market so they can become world leaders. This will need European companies to consider the circular economy, new market positioning (by turning small market shares into specialisation areas), the environmental impact of global manufacturing, etc.
• Establishing this carbon-neutrality challenge, based on a close link between the digital and green transitions at the core of future funded collaborative research and innovation in ECS. This will help ensure a positive impact for each stage of the value chain, and achieve carbon neutrality right down to the final application/digital service.

Even in a changing geopolitical context, climate change remains an environmental and economic imperative that cannot be ignored.

Advances in ECS technologies are both a major enabler of the Green Deal for all ECS-based application fields, and a direct contributor through their significant impact on power and resource consumption of ECS manufacturing and use.

Furthermore, the strategic autonomy introduced by the Chips Act contributes to economic sustainability and security, through a reduced dependency on non-European suppliers, an increased production capacity, more competitive manufacturing processes and products, a resilient supply chain, circular economy promotion, etc.

ECS must have intelligence and autonomy capabilities to control their complexity more efficiently and more cost-effectively. This will help provide novel advanced functionalities and services, limit human presence to only where it is strictly required, improve the efficiency of vertical applications, etc. Intelligence and autonomy are also required for the role of ECS in the application domains, representing an important factor for the sustainability and resilience of the value chains: an ECS-based system that provides intelligent energy management, relying on technologies such as AI, represents a key building block – for example, for smart home and energy applications. Moreover, it also improves the resilience required to ensure optimal energy consumption in critical conditions and contributes to the sustainability of the value chain associated with vertical applications, since it reduces operational costs and environmental impact, improves the quality of service (QoS), return on investment (ROI), etc., thereby strengthening the global competitiveness of European companies and helping to achieve the objectives of the EU’s Green Deal.

With an innovative and strong semiconductor industry Europe can improve its competitiveness in the global AI landscape, and this will contribute attracting AI talent, companies, and research initiative in Europe. Supporting the key focus areas identified in the SRIA will be crucial to consolidate the European future position in the AI market and contribute to the supply chain resilience, supported by a robust semiconductor industry in AI hardware. Boosting the RD&I activities described in Chapter 2.1 of this SRIA will ensure a solid support for all the vertical domains covered by the Application chapters.

Modern ECS-based systems are too complex to deliver reliably without a coherent end-to-end engineering ecosystem that spans requirements, architecture, design, verification and validation, deployment/commissioning, operations and remote maintenance, and eventually recycling and evolution. Establishing such an ecosystem—grounded in systems-engineering best practice—improves quality, reduces time-to-market, and enables confident scaling from components to complex systems and systems-of-systems.

This objective promotes lifecycle integration through model-based and data-centric methods (e.g., MBSE and an unbroken “digital thread”) so that product information, analyses and decisions remain consistent from concept to field operation—benefiting especially SMEs that struggle to connect tools and processes. It also supports interoperable toolchains and open interfaces that enable co-design across edge–to–cloud computing and emerging modular hardware (e.g., chiplets), and aligns engineering workflows with design-platform concepts that provide shared access to IP, PDKs and design automation in secure environments.

Security, safety, reliability and sustainability are treated by design. Engineering processes should incorporate established cybersecurity and functional-safety lifecycles and requirements so that risks are managed systematically from early architecture through maintenance. Lifecycle-integrated verification, continuous assurance (including software updates and AI-assisted engineering), and traceability strengthen dependability across domains.

Delivering this objective also depends on human capital: Europe needs coordinated education, up-/re-skilling, and hands-on training pathways that match real engineering workflows across the semiconductor and ECS supply chains. (Multiple recent European analyses highlight the skills gap and the need for joined-up action by industry, academia and governments.)

In practice, Objective 5 will:

• impact industrial competitiveness by simplifying life cycle management, and improves the quality of the engineering process, making it more cost-effective and agile.
• simplify and improve the development of trustworthy ECS technologies, products and applications.
• support sustainability and resilience that reduce lifecycle management costs, as well as ensuring the automation and continuity of operations.
• unleash the full potential of intelligent and autonomous ECS, which requires completely new approaches to engineering, design and development methodologies, as well as toolchains and tools.
• improve professional training and education by strengthening and developing new and specific skills.

By consolidating these capabilities, Europe will simplify lifecycle management, raise engineering quality and agility, and reduce cost and risk—thereby amplifying the impact of Objectives 1–4.

Globally, the long-term market trend for electronic components is expected to exceed US $1,000 billion by 2030³. In Europe, the semiconductor ecosystem employs some 250,000 people, with 2.5 million in the overall value chain of equipment, materials, semiconductors components, system integration, applications and services – mostly in jobs requiring a high level of education.

Figure 0.1 Process technology, equipment, materials and manufacturing is at the base of the digital value chain (Source: DECISION Etudes & Conseil) – 2023 market size numbers

The demand for chips is expected to double by 2030, driven by the digitalisation of society and the pervasiveness of Artificial Intelligence.

Given that in the next decade, 85% of overall global growth is projected to take place outside the EU⁴, it is essential that the EU maintains and increases the worldwide competitiveness of its electronics component industrial ecosystem. In turn, this increased competitiveness will stimulate all the ECS value chain and downstream industries that depend on it, including transportation, healthcare, energy, security, and telecommunications, to name a few. It is a key ingredient to maintain manufacturing as the backbone of the European economy, to guide further digitalisation of society and industry, ensure Europe’s sovereignty and resilience against crises situations, facilitate the digital single market, maintain employment and move towards a better, smarter society.

Besides its economical weight per se, the strategic importance for Europe of the ECS ecosystem is further strengthened due to its role as the enabler of the digitalisation of our society.

Digitised services based on smart electronic components and systems (ECS) are becoming increasingly ubiquitous, penetrating every aspect of our lives. They also provide ever greater functionality, more connectivity and more autonomy. The benefits to society are numerous: safer traffic, less carbon emission and pollution, instantaneous access to information, convenience and cost saving in e-health, more efficient factories, more social safety, and many more. In short, ECS-based applications and services are key to ensuring the stable growth and development of the European Union.

On the other hand, our dependency on ECS-based applications is also continuously growing. This reinforces people’s concerns on – among others – the security of privacy and personal possessions (e.g. through cybercrime, cloud services, surveillance cameras), on personal safety (driver-assistance, e-health), and on the unclear impact of transformational technologies (AI, quantum technology). The societal concerns arising from these technologies need to be addressed. To achieve that, it is vitally important that these electronic systems are safe, secure and trustworthy. A human-centred approach is a key aspect of the EU’s policy on technology development, and in line with fundamental European social and ethical values.

In the following pages, we examine in more details the societal benefits which ECS brings to Europe, and where R&I efforts are needed to ensure further gains.

Chiplets are not a package type, they are part of a packaging architecture. They are integrated circuit blocks that are specifically designed to communicate with other chiplets, to form larger, more complex ICs. Thus, in large and complex chip designs, the design is subdivided into functional circuit blocks, often reusable IP blocks, produced in suitable semiconductor material and process. These are called "chiplets", that are manufactured and recombined on high density interconnect, resulting in a shrinking of the PCB size.

Chiplets have some major advantages over SoC (System on Chip):

• Yield for each chiplet IP block is significantly higher +90% than for large SoC, 30-40%.
• A chiplet package becomes cheaper compared to a SoC.
• Time to market for chiplets are significantly shorter compared to SoCs.
• Modularisation allows for IP reuse
• Chiplets can be manufactured using the most adequate process technology in terms of cost, power and performance, while the monolithic SoC approach requires using the same technology for all IP blocks of the circuit

Current usage of chiplets is primarily in the high-performance computing server industry addressing AI applications. World leaders are NVIDIA, AMD, INTEL all supported by TSMC.

For the automotive industry, Tesla and some Japanese brands already use chiplets in their cars.

Chiplet technology has therefore the potential to lead to versatile and customizable modular chips, enabling reduced development timelines and costs. In addition, IP reuse improves design flexibility and efficiency. With all those promises, chiplets are sometimes touted as a revolutionary advancement in semiconductor technology²¹²². It is especially relevant for Europe, as a path to deliver innovative solutions while keeping the dependence relative to the dominant, non-EU advanced CMOS manufacturers to a minimum.

Chiplets manufacturing is addressed in Ch 1.2, while implications of chiplets are addressed in Ch 1.3, 1.4, 2.2 and 2.3.

Open-source hardware is hardware whose design is made publicly available so that anyone can study, modify, and further distribute the design, and/or freely make and sell hardware based on that design. The use of open-source hardware drastically lowers the barrier to design innovative SoCs. Indeed, this allows research centers and companies to focus their R&D effort on innovation, leveraging an ecosystem of pre-validated IP that can be freely assembled, modified, and customized for specific applications, whereas currently the cost of design of a set of IPs in-house is available to only few companies. And the use of alternative 3rd party IPs licensed by companies imposes constraints on innovation due to architecture. Open-source is also a sovereignty tool and avoids licensing IPs from foreign third parties, in a geopolitical context characterised by trade wars and export control restrictions.

While open-source SW approaches are now in wide use, open source HW has really drawn attention beyond the academic circles only in recent years. In China, USA and India, large governmental funded programmes have been created in order to stimulate the use of RISC-V and Open-Source in industry and products. The widespread adoption of this approach in Europe would have many benefits.

Creating an ecosystem to support open-source hardware is essential as it cultivates innovation and aids dissemination. This can be first based on the implementation of a strategic “Governance Initiative’’ to coordinate activities, maintain and promote the repository. Open-source projects only work well when they are attractive and provide innovation, but also when they are useful and trusted, so that at this point they can be self-supported by the community. One needs to make sure that knowledge is shared, accessible, maintained and supported on a long-term basis, as the success of open-source depends not only on the IP blocks but also on the documentation, support and maintenance provided. In particular, a one-stop-shop model with long-term activities and overall support (e.g., advice for licensing, productisation, etc.) can be promoted to help SMEs and start-ups.

The European strategy on open-source hardware should focus on application domains where there is a stronger impact, i.e. automotive, industrial automation, communications, health and aeronautics/defense. As these fields convey specific requirements (safety, security, reliability, power and communication efficiency…), the key differentiator between European core design efforts and existing players will be the attention given to the aforementioned requirements and also the extensibility with custom operations. Above all, a special focus should be placed on safety and security solutions, dealing with key aspects of collaboration, documentation, verification and certification in open-source communities.

A working group gathering European stakeholders was established in 2022. It drafted a technological roadmap starting with processors (RISC-V, beyond RISC-V, ultra-low power and high-end). The chiplet-based approach mentioned above is identified by the roadmap as a unique opportunity to leverage European technologies and foundries to create European HW accelerators interposers that could foster European More-than-More technology developments. Die-to-Die communication remains a missing link to leverage that approach, and its development in open-source is advocated by the working group.

In conclusion, the emergence of a European open-source hardware community should be encouraged and its initiatives be guided towards application domains where there is a strong impact. A technological roadmap has been drafted to build on the strengths of European industry and fill its gaps, while building blocks being developed will also need to be supported by the required tools, software supports, tests and documentation.

Disruptive technologies, where market positions are not established yet, are level-playing fields where Europe has all its chances to gain leadership. Quantum technologies fall in that category and therefore their evolution and maturity are closely monitored by the ECS SRIA expert community. The importance of the technologies is identified in both the Chips Act and in the European Quantum strategy²³. Together with the European Quantum Flagship SRIA, these documents highlight areas of importance and overlap with the ECS SRIA:

• Technologies for dedicated quantum chips, required to fulfil the quantum challenges.
• Classical chips technologies for quantum (enabling technologies) required to support the industrialization and scaling of quantum technology.
• Systems combining these two for applications taking advantage of quantum technology.

Quantum technologies are very diverse in nature, encompassing quantum computing and simulation, quantum communication, and quantum sensing and metrology²⁴. As application fields, they go beyond the scope of the ECS SRIA, but the community should prepare for using quantum technology in the applications that can benefit from quantum chip technologies, e.g. in navigation and timing using quantum sensors.

Quantum chip technologies include six major modalities: Superconducting circuits, semiconducting quantum dots, ion traps, photonics, neutral atoms and nitrogen vacancies in diamonds. All of these technologies have their specific chip technology development needs. The technologies combine traditional microfabrication technologies in Silicon and other substrates with novel primitives that use quantum physics on single electron/atom/ wavefunction level, with RF and optical control and readout. As an example, ion trap quantum systems combine RF electrodes for ion trapping for with optical control of trapped ions.

Quantum technologies use have special material needs, such as the use of isotopically pure Si, and specific processing needs, such as Al/AlOx/Al tunnel junctions with extreme requirements for dielectric tunnel barrier quality and thickness control. In integrated photonics, the needs of the quantum systems typically include extremely low losses and wavelengths in the visible range, from 800-900 nm as low as 300-400 nm, depending on the atoms or ions used.

In applications, quantum computing is projected to have the largest market size and impact to many application fields. Estimates vary, but quantum computing market growth to 28 B€ in year 2035, with full value chain, including application benefits, potential to be over 1 trillion Euros²⁵. However, quantum sensors while being a smaller market, may be considered more mature, with sensors already use in application, such as: superconducting SQUID magnetometers in biomagnetism and brain imaging diagnostic sensors or quantum sensors in GPS free navigation.

In particular, the emerging field of quantum computing poses its own challenges for the ECS community in process technology, equipment, and materials:

• As there are still several candidates for becoming the standard quantum computing technology, a wide range of materials is relevant, together with innovations in process technology.
• New metrology capabilities are required, especially the measurement of electrical properties, such as local carrier mobility.
• To achieve practical applications, reliable fabrication, connection, and read-out of qubits need to be developed.
• The low temperatures, from milliKelvins to 4 K, at which many quantum systems are operated requires the development of cryogenic devices, to interface conventional electronics.

Beyond quantum computing, all quantum technologies related to sensing, communications and computing, including software, present significant challenges today relevant for the ECS SRIA. Given the level of maturity of those technologies, especially when it comes to quantum computing, those challenges are mainly discussed in the Long-Term Vision chapter, but research focus areas derived from the needs of quantum technologies can also be found in other chapters of this document. This reflects that quantum technologies maturity is expected to evolve rapidly in the years to come, and that Europe should leverage its assets, with excellent research centres and a dynamic ecosystem of start-ups, to gain leadership in that promising field.

Similarly to quantum technologies, spintronics (excluding MRAMs which are already at a commercial stage) belongs to the category of disruptive technologies, where market positions are not established yet and Europe has the opportunity to gain leadership. This is an emerging field that holds significant potential to become a key component of future energy efficient chips technologies. The high performance and ultra-low power consumption intrinsic of such technology, make it a very attractive solution to the increasing demands for computing power, notably with the growing market of AI. By combining processing, logic, sensing and storage on a unified platform, spintronic devices can advance technologies such as neuromorphic electronics, unconventional and quantum computing, and energy harvesting, significantly contributing to low-carbon footprint prospects. Europe is well positioned with world-leading institutes in the domain

With the development of the global internet traffic, AI/ML and IoT, the demand for data centers and high-performance computing (HPC) is increasing. An extremely large link capacity for high-speed datacom interconnects between multi-cores or local/distant caches becomes compulsory. However, it is becoming increasingly difficult for the conventional electrical Cu interconnects to meet such an ever-growing capacity requirement since they severely suffer from limited bandwidth, significant power consumption, integration difficulty on racks and material cost and mass.

Silicon Photonic Integrated Circuits (PIC), referred to as Si-photonics, is the technology that realizes photonic components and circuits on Si substrates using materials and fabrication process flows in the well-established microelectronic industry. Replacing electrons with photons to transmit data brings advantages including high-speed operation, low power consumption and high-capacity transmission. Building photonic integrated circuits (PICs) based on industry-standard Si platform promises additional merits, namely lower cost, high-volume manufacturing, large integration density, advanced functionality and high scalability.

Furthermore, Si-photonics, with their CMOS compatibility and small size, weight, area, and power consumption, have the capability of accommodating growing volumes of computer data arriving in real time and at very high rates in large-scale distributed computing systems. The wavelength and spatial multiplexing properties of optics are beneficially translatable to executing critical bandwidth and communications-intensive connections between increasingly parallel computational resources.

Benefiting from an ideal Si/SiO2 interface and a large refractive index contrast, Si is an ideal platform for implementing passive optical devices such as low-loss waveguides, mode converters, and multiplexers/demultiplexers. Although Si is a centrosymmetric crystal and, thus, not an ideal material for optical modulators, high-performance modulators have been demonstrated. Efficient Ge- and SiGe-based photodetectors operating in the telecom bands are also selectively grown on Si-photonics wafers.

However, if Si can efficiently transmit, modulate, and detect light, its indirect band structure precludes efficient light emission. The achievement of on-chip light sources, using Ge alloys for instance, represents a significant challenge in advancing the complete integration of silicon-based PICs. Therefore, III-V material (InP, GaAs) on-chip lasers are currently embedded in PICs using 2.5 and 3D heterogeneous integration. Monolithic integration of III-V materials is an attractive alternative due to its high integration density. However, it introduces a high density of crystalline defects, significantly degrading the laser’s performance. Exploring methods of reducing these defects is paramount to advance monolithic integration toward high-density, large-scale silicon-photonic integration.

Similarly, for applications that require high-speed modulation, the properties of Si-based modulators are not sufficient. Hence, ferroelectric electro-optic materials such as, III-V semiconductors, LiNbO3 and BaTiO3 should be embedded in the PIC using heterogeneous integration (die-to-die, die-to-wafer, or wafer-to-wafer bonding). Also, improving the quality and performance of waveguides, passive and active photonics devices sometimes requires the introduction other dielectric materials such as nitrides.

The most established form of the technology is optimized for operation in the telecommunication wavelength bands near 1310 and 1550nm and has waveguides defined in the top Si device layer of a silicon-on-insulator (SOI) wafer. It is enabling power-efficient transceivers and large-scale PICs that address the demands of data communications. Three-dimensional (3D) sensing, and computation accelerators are other applications for which convincing demonstration exists. In addition to these applications, there is a growing demand to develop silicon-photonic for new product applications that include chip-to-chip electrical/optical interconnect, automotive LiDAR (in particular for beam steering without any moving part), quantum computing, infrared imaging and biomedical and environmental sensing. The last three applications require the development of PICs in the mid-IR region wavelength range (2–20μm) as it contains strong absorption signatures of many molecules, which requires different sets of materials.

While not a totally new topic (it is used for example in several embedded systems for vision, enabling autonomous vehicles), artificial intelligence (AI) really came to the forefront for the public at large in 2023, with tools such as Chat-GPT (released for public use only on November 30th, 2022) and other generative AI (Dall-E, Stable Diffusion) that grabbed the headlines in mainstream media. The ECS Research and Innovation roadmap interacts with the development of AI tools and algorithms in two symmetrical ways: ECS are an enabler for new AI developments, and AI can in turn be a significant enabler for new ECS advances. Those two aspects are extensively discussed in this current edition of the SRIA.

• ECS as enabler of AI: Chapters 1.1 and 2.1, as well as the Long-Term Vision Chapter, stress the importance of moving the processing of AI algorithms locally on a hardware device (in deep-edge devices, edge devices or on-premise computing resources, depending on the application) close to where the data is generated (e.g. by a sensor), a trend known as “embedded AI” or “edge AI”. Benefits include reducing the energy consumption of the data infrastructure by transmitting only relevant data or pre-treated information, improving data protection, increasing security and resilience due to a reduced reliance on telecommunication links, reducing latency, and decreasing memory footprint. Many of the focus research areas identified in Major Challenge 1 (Advanced computing, in-memory, neuromorphic, photonic and quantum computing concepts) of Chapter 1.1 support this move towards AI at the edge. Likewise, the four Major Challenges (increasing energy efficiency, managing system complexity, increasing device lifespan and ensuring European sustainability of embedded intelligence), identified by Chapter 2.1, all cover research topics which will enable the development of a strong European embedded AI ecosystem. Energy consumption of AI solutions is an overarching issue due to the expected widespread increase of their usage, and it is essential to explore new concepts and architectures (bio-inspired and other ones) to respond to that concern, which could turn out to be a major roadblock if not addressed properly.
• AI as enabler of ECS: this aspect is discussed in Chapter 2.3, where several examples of AI use are listed to tackle Major Challenge 3 (managing complexity), such as AI-based methods for the architecture exploration and optimisation, to achieve a global optimum, AI-based guidance in the V&V process, and automatic generation of test cases with AI support. Moreover, in its entirely new section devoted to Machine Learning and Artificial Intelligence, the Long-Term Vision Chapter states that AI/ML methods are increasingly adopted in design space exploration at several stages of circuit design, as well as in design testing and verification. The potential of these AI-based methods is not limited to circuit design and extend to large-scale ECS products encompassing HW and SW, as well as multi-physical, distributed systems. They are also expected to provide guidance to engineers in multi-risk (safety, security, privacy, and other trustworthiness risks) optimisation problems. Gathering / generating the appropriate data to train the models is identified as one of the most difficult issues to be addressed.

A third aspect is that AI adds complexity to the systems it empowers. This in turn raises additional issues at the design phase, as addressed by Chapter 2.3. There, Major Challenge 1 (Enabling cost- and effort-efficient Design and Validation Frameworks for High Quality ECS) stresses the need to develop new design and verification and validation methods for AI-based ECS, including ECS evolving during lifetime.

Finally, a purely technical approach to develop trustworthy and explainable AI, or having AI in the loop with verification of its results might not be enough to address the growing public concerns about the potential dangers of AI (pushing many governments across the world to sign the Bletchley Declaration on AI safety). In the years to come, cooperation between multidisciplinary partners with backgrounds in AI, ECS and social sciences, as well as access to “foundation models” by European academia and industry will be required to achieve and ensure the acceptance of efficient and responsible AI-based ECS, reflecting European values.

The energy world is undergoing a radical transformation, dominated by three factors: installation of renewable energy sources, transition to carbon neutral mobility and massive growing demand for AI datacenters.  Promoted by EU and national roadmaps, the globally installed capacity of renewable generation has doubled within the past 10 years. Europe alone expanded its renewable generation capacity by 10% in 2023²⁶. This increase is dominated by wind and solar energy being characterised by strongly intermittent, distributed generation. The share of wind and solar of Europe's electricity generation reached a record high of 26.6% (718 TWh) in 2023, almost twice the global share of 13.4%.

Highly efficient, reliable, affordable power electronics and power system control are the core elements from the ECS side contributing to the transition. Lower costs are key to safeguard the tremendous societal change, since current high energy prices are one essential factor leading to instabilities and are a thread towards the change to a carbon neutral society.

Having on one side the energy transition challenge, electricity generation from renewables, heating towards heat pumps, mobility towards electric mobility the next challenge is evident – the energy consumption of the AI data centers. According to a study by McKinsey²⁷, global demand for data center capacity could rise from the demand of 60 GW in 2023 at an annual rate of between 19 and 22 percent until 2030 to reach an annual demand of 171 to 219 GW.

The need is given with the upcoming AI capabilities – the challenge is to reduce the energy demand by algorithms and more and more efficient power supply – closer and closer to processor units, more and more efficient. This huge market, driven by the three factors energy system transition, mobility and data centers, is attractive, hence globally research and expansion of fab capacities for power electronics is booming. To keep the technology competence, to serve the demand from European based manufacturing research and investment in Europe is mandatory in the fields of the core technology serving the electricity demand of the future.

Device architecture, the excellence in achieving extremely thin wafer thicknesses, and for the wide bandgap technologies the move to larger wafer diameters are core technology capabilities in Europe which should be further strengthened to keep a significant manufacturing and knowledge share in Europe. Vision is to serve the energy transition with highest efficient, reliable, sustainable manufactured power electronics and be capable to have the system competence to design beyond state-of-the-art power systems, for the renewables, electric mobility and for the powering of the AI data centers.