In February 2020, the European Commission published its initial proposal – White Paper on Artificial Intelligence: a European approach to excellence and trust – and opened a stakeholder consultation. In June, SEMI Europe, based on member feedback, made key policy recommendations in response to the Commission’s AI White Paper.
Until recently, AI developments have focused on cloud and computations have been performed, by and large, remotely in data centers. Edge AI – computing data at the point of generation thanks to IoT sensors placed in smart objects with some data processing on board – is now challenging the centralized model. AI applications are increasingly called on to recognize patterns instantly, such as people or cars in fully autonomous transportation systems. Such critical applications cannot afford cloud-to-device roundtrip and rely on wireless communications as they are based on terabytes of data originating from HD cameras, radars, lidars and other high-speed sensors. In these cases, data needs to be computed directly where it is collected, making the edge an increasingly crucial pillar of future IoT systems and AI applications.
While the cloud will still complement the edge by pooling (too) big data and training AI inference algorithms that run on the device, more AI applications will target the edge. By 2024, sales of Edge AI chips – chips or parts of chips that perform or accelerate machine learning tasks on-device, rather than in a remote data center – are to exceed 1.5 billion units. This represents annual unit sales growth of at least 20 percent, more than double the longer-term forecast of 9 percent compound annual growth rate (CAGR) for the overall semiconductor industry.
Running AI algorithms on the device, compared to in the cloud, offers critical advantages such as lower latency, higher reliability, increased security and privacy and more efficient use of network bandwidth. By processing large amounts of data locally, Edge AI reduces the risk of personal or business data interception or misuse. AI at the edge eliminates the round-trip journey to the cloud to deliver real-time responsiveness. Low-power microchips allow devices with small batteries to perform AI computations on the device and lowers energy consumption. Embedded Edge AI chips decrease the cost of data storage and bandwidth, as only relevant data is analyzed in real time. Edge AI promises to make significant contributions to EU public policy goals including Europe’s security, environmental sustainability and competitiveness.
Against this background, SEMI Europe makes the following policy recommendations pertaining to Europe’s new AI policy:
Focus on Europe’s strengths in Horizon Europe and Key Digital Technologies Joint Undertaking and invest in microelectronics R&D that brings AI to the edge
SEMI Europe fully supports Europe’s new AI strategy aimed at capitalizing on Europe’s strengths in industrial and professional markets. The forecast CAGR for professional/embedded systems, including automotive, medical, industrial, robotics, aerospace and security-related electronics, is 50 percent over the next five years.
Professional/embedded electronics is a key sphere where Europe holds a central position globally. Europe’s share of global electronics production, unsurprisingly, is the highest in the segments where Europe is well-positioned. The EU automotive industrial base is very competitive, and the EU automotive electronics sector is the global leader, accounting for 27% of global automotive electronics, ahead of China (20%) and North America (18%). The EU also holds a strong position in industrial applications, producing 20% of the world’s electronics in that segment. The EU generates 22% of the global aerospace/defense/security electronics and 19% of global medical electronics.
The upcoming Horizon Europe and the Key Digital Technologies Joint Undertaking (KDT JU) should play a pivotal role in positioning Europe at the center of Edge AI by pooling public and private resources and capitalizing on Europe’s strengths in embedded electronics. SEMI Europe welcomes the proposal to extend the scope of future electronics-related EU research and to integrate software, photonics, bioelectronics and flexible electronics into KDT JU. However, the Joint-Undertaking should remain centered on Europe’s competitive electronics design and manufacturing technologies including semiconductor materials and equipment, microchips, FD-SOI, advanced packaging, MEMS, sensors and imagers that enable IoT, supercomputing, rapid data processing at the edge and hyper-connectivity at low power. Furthermore, KDT JU should focus on requirements for transportation and medical since the industries are key end users in Europe. Given the extended scope foreseen for KDT JU, compared to ECSEL JU, SEMI Europe supports the proposal to double the budget to €10 billion in KDT JU, twice that of ECSEL JU.
Establish a technology-centric and pan-European Testing and Experimentation Facility for Edge AI, supported by the Digital Europe Programme
SEMI Europe fully supports the Commission’s plans to establish Testing and Experimentation Facilities through the Digital Europe Programme. The Commission should consider Europe’s key strengths in prioritizing the establishment of a technology-centric and pan-European TEF for Edge AI hardware (AITEF) and federate the synergies of key European research and technology organizations (RTOs) and companies in the field of micro- and nano-electronics under a common platform.
AITEF will play a vital role in accelerating the development and market uptake of embedded electronics that enable major Edge AI applications, amongst others, in smart health, mobility, manufacturing and agriculture. AITEF is envisaged to serve as Europe’s overarching platform for demonstrating and validating pre-commercialized AI technologies. This will allow taking the results of earlier R&D projects out of the lab, and elevate them from technology readiness levels (TRL) 3-5 to 5-7, shortening the time-to-market of new products. This involves making demonstrators; performing yield and statistics analysis on small but relevant volumes of devices; validating them to meet the needs of various AI use cases in Europe; and building prototypes that are ready for market uptake by companies.
It is important to underline that AITEF, supported by the Digital Europe Programme, will not launch new R&D programmes. AITEF’s main focus will be deployment of pre-commercial innovations that are already funded by Europe’s R&D programmes such as Horizon 2020 and ECSEL JU and available at Europe’s leading RTOs specializing in micro- and nano-electronics. Such non-commercialized innovations, however, require testing and validation before market introduction. Therefore, AITEF will play an instrumental role in connecting Europe’s R&D with manufacturing capabilities and accelerating the transition from lab to fab. It is the ambition of the AITEF to leverage the Edge AI innovation of Europe RTO powerhouses to accelerate innovation in Europe.
AITEF should be designed to incorporate complementary testing and experimenting capabilities of RTOs and future products and services of the microelectronics supply chain including design houses, equipment, materials and integrated device manufacturers (IDMs) and foundries located in Europe. AITEF’s role, therefore, will be to develop, test and validate Edge AI hardware and streamline the production of new electronic components by IDMs and foundries in Europe.
Expand AI hardware manufacturing through Important Projects of Common European Interest
While Europe’s public-private-partnerships in R&D are globally admired, the EU and Member States should, in parallel, launch bold investment programmes that support capital-intensive, risky and complex manufacturing activities in Europe. Investing in AI-related research alone does not necessarily translate into keeping the innovation bar high. A strong European AI ecosystem requires the continued coupling of microelectronics research and manufacturing activities.
Understanding the crucial role of maintaining a strong industrial base, countries around the world are making heavy government-backed investments to build domestic fabs. China’s Made in China 2025 initiative, which establishes an Integrated Circuit Fund to support the development of AI-related hardware, calls for $150 billion in funding to replace imported semiconductors with homegrown devices. In June 2020, U.S. lawmakers proposed the Chips for America Act, a bill that would provide $22.8 billion in aid to the U.S. semiconductor industry.
In 2018, the European Commission approved €1.75 billion in public support from France, Germany, Italy and the UK to fund the IPCEI (Important Projects of Common European Interest) Microelectronics. The project seeks to enable technologies for key applications such as autonomous and electric mobility, smart manufacturing and home devices. IPCEI Microelectronics paves the way for developing energy-efficient chips, next-generation power semiconductors, smart sensors with higher performance and accuracy, advanced optical equipment and compound materials beyond silicon.
The cost of R&D and manufacturing in microelectronics is increasing, with a state-of-the-art fab easily costing more than €10 billion. The technological, industrial and geopolitical reality in which the current IPCEI Microelectronics was conceived, compared to the context in which it is now being implemented, has drastically changed. The microelectronics industry in Europe is now contemplating establishing a second IPCEI. SEMI Europe urges the EU and Member States to reinvigorate the manufacturing of microelectronic chips and systems to enable Edge AI in the second IPCEI Microelectronics as well as other IPCEIs planned for clean, connected and autonomous vehicles, smart health, industrial internet of things and cybersecurity.
Prioritize advanced AI skills in the Digital Europe Programme, Erasmus+ and Pact for Skills
Continuous innovation is the oxygen of microelectronics and AI, powering the development of highly customized solutions by a workforce with technical expertise in engineering. In addition, capabilities such as smart applications require workers with growing knowledge in software and data analytics. While the evolution of microelectronics since the 1980s has been swift, education curricula in Europe have not matured at the same pace, opening a gap between the worlds of industry and education and imposing a fragmented school-to-work transition for many young graduates.
Against this background, the SEMI – Deloitte Survey Workforce Development: A Critical Industry Issue shows that the global electronics industry is increasingly reporting difficulty filling open positions in engineering disciplines related AI, machine learning and digitization. Likewise, in 2018, the German industry registered a shortage of 337,900 workers in STEM. Another report indicates a shortfall of 173,000 skilled workers as 89% of STEM businesses struggle to recruit in the UK. If the current trends continue, China and India will account for more than 60% of the OECD and G20 STEM graduates, and Europe will be lagging behind with 8% of STEM graduates by 2030. Europe is behind its peers in developing AI talent as well. The U.S. employs twice as many AI-skilled individuals than the EU, despite the American total labour force being just half the size.
The upcoming Digital Europe Programme’s Advanced Digital Skills Pillar, the newly announced Pact for Skills and Erasmus+ Programmes are well-positioned to accelerate the development of advanced AI skills in Europe. They should also a play a powerful role in diversifying and strengthening Europe’s AI talent pipeline by cultivating skills among underrepresented groups. Future EU education programmes should support work-based learning to help students develop the knowledge and practical job skills needed by businesses. Innovative dual-learning programmes, apprenticeships and industrial master’s and doctorates are shining examples that are already paying off in some parts of Europe.
Work-based learning is vital to remaining competitive in the long run, in particular in rapidly emerging technologies such as AI. In addition, as the conventional path from education to work to retirement is becoming outdated, reskilling and upskilling are essential responses to ongoing technological transformations and labour market transitions. Future EU-funded education programmes should, therefore, support the reskilling and upskilling of adult workers in order to develop the digital skills needed to remain productive and innovative. Such reskilling and upskilling programmes should be supported by integrating massive open online courses into work-based learning or classroom teaching to provide the ideal educational environment.
“The White Paper indicates Europe’s willingness to move forward to capture the huge potential ahead in the rapidly emerging AI technologies,” said Laith Altimime, president of SEMI Europe. “SEMI Europe would welcome the opportunity to discuss the aforementioned policy recommendations with the European Commission. The microelectronics sector is a key enabler and remains ready to work with all interested stakeholders to develop new avenues of growth to reinforce Europe’s AI leadership.”
You can download SEMI Europe response to EU AI White Paper here.
Emir Demircan is Director of Advocacy and Public Policy at SEMI Europe. Contact: firstname.lastname@example.org