Over the past decade, the importance of data-driven research has grown, supported by artificial intelligence (AI) methods such as machine learning. Adequate computing infrastructure is essential for data- and compute- intensive analyses. The Swiss Science Council (SSC) has examined whether Swiss academia has the necessary resources to meet its current and future needs. Through interviews with organisations in Switzerland and abroad, the SSC has gathered information on current computing infrastructures, cloud services and future plans and visions. Swiss higher education institutions (HEI) are actively considering how to adapt their systems to meet rising computational demands. This is considered strategically important for the future of the Swiss education, research and innovation (ERI) system, and several challenges have been identified.
Predicting resources is difficult because future computing needs and the effects of optimisation are difficult to forecast. HEI face challenges as their computing infrastructures are limited, become outdated quickly and are less competitive than those in industry. As industry increasingly leads AI research, academia risks falling behind and must find ways to remain competitive. Existing academic computing resources are often considered inadequate, prompting some experts to collaborate with industry to gain better access. The Swiss AI Initiative is perceived as benefiting the ETH domain more than it serves as a national initiative associated with the most advanced computing resources.
HEI emphasise the importance of planning data access, storage and archiving alongside computing infrastructure planning. Concerns about data privacy, protection, sovereignty and costs lead to a preference for in-house computing. Swiss natural resources are considered insufficient or too expensive for the kind of largescale supercomputing seen internationally. Retaining senior AI talent is challenging in a competitive job market, which has resulted in a brain drain from academia to industry.
Based on the results of the national and international interviews conducted, Switzerland’s diverse academic landscape currently appears suboptimally prepared for future research requiring intensive computing resources, such as AI. However, the design of future-oriented, high-performance computing capacities is being addressed at political and strategic levels.1, 2, 3, 4, 5 As this field is subject to rapid change and ist importance for Switzerland cannot be doubted, the council recommends the development of a long-term AI infrastructure strategy and to create a tiered computing infrastructure system based on this national strategy. This requires establishing a national strategic board and securing adequate funding. For this to happen, the council recommends as immediate action: Declaring the development of a computing infrastructure for academic research a national task for which new mechanisms must be found, both in terms of design and financing, in order to master the challenges as identified. Make it accessible and equally usable for the diverse research, innovation and education landscape, in line with ist needs and compatible with economic strategies, so that economic benefits would also be targeted, for example, through the integration and servicing of SME. To this end, the relevant stakeholders (e.g., higher education institutions, federation, cantons) must engage in dialogue in order to develop, coordinate, establish and finance the necessary mechanisms at national level with a long-term commitment.