The Warning Behind the Headline
Switzerland already holds formidable AI infrastructure. The CSCS Alps supercomputer, operated by ETH Zurich in Lugano, delivers 10 exaflops of AI-optimised BF16 compute — one of Europe's most capable academic machines. ETH Zurich and EPFL together hold more than 150 AI-linked professorships. In September 2025, the Swiss National AI Institute debuted Apertus, the country's first open-source multilingual language model, trained entirely on Alps across six million GPU hours and since downloaded over one million times.
Against that backdrop, the Swiss Science Council's (SSC) May 12, 2026 formal recommendation reads as a structural alert rather than a crisis call. The SSC — the Federal Council's statutory advisory body on science, higher education, research, and innovation — concluded in its report "Synergise. Strategise. Realise.: SSC recommendations for AI computing infrastructure in the ERI domain" that Switzerland's AI compute assets are strategically under-coordinated, funded on project timescales incompatible with infrastructure demands, and at genuine risk of falling behind competitors unless the Federal Council acts.
Three Concrete Proposals
The SSC's recommendation has three interlocking components.
A multi-tiered computing architecture. Rather than concentrating resources in a single national facility, the Council envisions an interoperable, layered system spanning regional compute clusters, national infrastructure, and international connectivity. The design principles specified — flexibility, scalability, efficiency, interoperability, digital sovereignty, and data lifecycle management — signal that the proposal is not simply about procuring more GPU racks. It is about governing how compute is provisioned and coordinated across Switzerland's federated institutional landscape, potentially linking into European-level resources.
An independent strategic board. The report proposes that the Federal Council establish an expert commission with genuine authority over infrastructure planning and evolution — responsive to national research requirements and global technical trends rather than to any single institution's interests. Switzerland's academic computing has historically been negotiated between the ETH Domain, the Swiss National Science Foundation, and cantonal universities. That structure works for established disciplines; it moves slowly when infrastructure needs to evolve at AI timescales.
A shift in funding logic. The most pointed element of the report is its explicit rejection of short-term project grants as the vehicle for building infrastructure. The SSC states plainly that sustained, long-term financial commitment is essential — that project-cycle funding is structurally incapable of generating durable compute capacity. Researchers cannot plan multi-year training runs or commit to open-model programmes when access to compute expires with the grant period.
The Regulatory Backdrop
The SSC recommendation arrives as Switzerland has been deliberately constructing a non-EU AI regulatory posture. In February 2025, the Federal Council adopted a technology-neutral, principles-based approach — declining to replicate the EU AI Act's comprehensive rulebook in favour of targeted amendments to existing sectoral law. On March 27, 2025, Switzerland signed the Council of Europe's Framework Convention on Artificial Intelligence, with implementing legislation expected by end of 2026.
This light-touch posture is credible and deserves credit. Swiss businesses are among Europe's most AI-optimistic: 52% of Swiss companies expect returns on AI investments within one year, above the European average, and 73% of executives anticipate significant AI revenue contribution by 2030. The Federal Council's logic — that existing legal frameworks covering data protection, financial markets, and copyright are sufficient to govern most AI applications without new horizontal legislation — is a legitimate position, not regulatory passivity.
The Steelman and Its Limits
Critics of the SSC recommendation have a coherent counter-position. The 2025–2028 ERI policy framework allocates CHF 29.2 billion across the full education, research, and innovation envelope — a substantial national commitment. The Swiss AI Initiative adds CHF 20 million specifically for AI research through 2028, supplemented by direct CSCS compute-access grants. Switzerland's federated model — ETH Zurich, EPFL, and cantonal universities competing and collaborating without a central planning apparatus — has produced consistently world-class research outputs. The track record of decentralised excellence is not fictitious.
The problem is time horizons. The Alps system was planned and funded over years before the current AI wave crested. The next generation of research infrastructure — clusters optimised for multi-modal training, federated architectures for privacy-sensitive scientific data, interoperability with exascale European resources under EuroHPC — requires the same multi-year commitment cycle. Without a national strategy that sets clear priorities, coordinates institutional plans, and sustains funding across political cycles, Switzerland risks watching its infrastructure advantage erode precisely as the EU's EuroHPC Joint Undertaking, the UK's AI Research Resource, and US NAIRR-style programmes accelerate their own roadmaps.
The independent board proposal is where the SSC's thinking is sharpest. An expert commission with a real mandate can translate evolving technical requirements into budget requests that political processes can evaluate — without being captured by any single institution's interests. That structural independence is not bureaucratic overhead. It is what makes long-term infrastructure commitments credible across funding cycles.
What the Federal Council Should Do
The SSC's report is a formal recommendation, not a directive — the Federal Council is not obligated to act on any particular timeline. The practical question is whether a national infrastructure strategy becomes a deliverable in the current legislative period or is deferred to the next ERI dispatch after 2028.
Switzerland has the research talent, the institutional foundations, and the regulatory philosophy to sustain genuine AI leadership. What the Science Council has identified is a coordination deficit that talent and philosophy alone cannot close. Apertus exists because Alps existed. Whether future Swiss models of comparable ambition can be built on Swiss infrastructure — or whether the next generation of Swiss researchers migrates compute budgets abroad — depends on whether the Federal Council converts the SSC's three recommendations from advisory text into funded policy.
Switzerland's choice is not between innovation and regulation. It is between investing in the shared infrastructure that makes innovation productive, or free-riding on the past investments that built Alps in the first place.