A Council, Not a Ministry, Makes the Case
On May 12, 2026, the Swiss Science Council (SSC) — the Confederation's independent advisory body on education, research and innovation (ERI) policy — published Synergise. Strategise. Realise., a formal recommendation that Bern adopt a long-term national strategy for AI computing infrastructure (wissenschaftsrat.ch). The council wants a tiered, interoperable system spanning regional to international scale, an independent expert commission appointed by the Federal Council to steer it, and — the recommendation the SSC leans on hardest — funding that outlasts individual grant cycles. As the council puts it, "short-term project funding is insufficient to build sustainable capacities."
That is not an abstract complaint. Switzerland's flagship AI/HPC system, the Alps supercomputer at the Swiss National Supercomputing Centre (CSCS) in Lugano, cost roughly CHF 100 million to build and now runs on an annual public operating budget of CHF 37 million (swissinfo.ch). CSCS director Thomas Schulthess has already flagged the ceiling this creates: "We are a subsidised infrastructure, and subsidies don't scale." Alps currently serves around 1,000 researchers, down from the 1,800 who used its predecessor Piz Daint, even though Alps offers roughly 20 times the compute — a sign that demand for AI-scale cycles is outrunning even a newly built system. The SSC's diagnosis, that grant-by-grant funding cannot support infrastructure with a multi-year buildout and multi-decade relevance, is a fair reading of that data.
The Steelman: Sovereignty Is Not a Slogan Here
Before pushing back, it's worth taking the SSC's sovereignty framing seriously rather than dismissing it as boilerplate. Switzerland sits outside the EU and has direct experience of what happens when access to European research infrastructure becomes politically contingent: after the collapse of the EU-Swiss framework agreement in 2021, Switzerland lost associate status in Horizon Europe and, with it, its seat at EuroHPC, the EU's joint undertaking for high-performance computing. That access was only fully restored on November 11, 2025, when Switzerland became EuroHPC's 38th participating state, regaining use of the network's 11 supercomputers — including JUPITER, LUMI and Leonardo — and its growing set of AI Factories (eurohpc-ju.europa.eu). A country that just spent four years locked out of its neighbours' compute has a legitimate reason to want infrastructure it fully controls, governed by a body that answers to Bern rather than Brussels. Domestic capacity is also the more defensible position if Switzerland ever wants to train models on sensitive health, finance or public-sector data that it is unwilling to move across borders — the SSC's own language about "knowledge security and data lifecycle management" points at exactly that concern.
Where the Case for Duplication Gets Thin
The timing, though, undercuts the strongest version of the sovereignty argument. Switzerland's EuroHPC readmission gives it access to a network the EU has been building at a scale no single Alpine country can match on its own: 19 AI Factories are already operational across the bloc, backed by €10 billion in EuroHPC funding through 2027 and a further €20 billion InvestAI facility earmarked for up to five AI "gigafactories" with roughly 100,000 processors each (digital-strategy.ec.europa.eu). Building a fully sovereign, full-stack Swiss alternative to that — on top of the CHF 100 million Alps already cost and its CHF 37 million annual draw on public funds — would mean re-funding, at national scale, capacity Switzerland can now access through membership dues instead. For a small, trade-dependent economy whose comparative advantage has historically been plugging efficiently into larger research and standards networks rather than replicating them, that is a case that needs to be made explicitly, and the SSC report — at least in the public reporting on it — does not name EuroHPC or explain how a new Swiss tier would relate to the one Switzerland just rejoined.
There's a governance risk too. The SSC wants the Federal Council to appoint an independent expert commission to oversee infrastructure design. That is a reasonable response to fragmented decision-making, but Switzerland's HPC capacity already sits under the ETH Board, which supervises CSCS and funds Alps. A new commission layered on top needs a tightly scoped mandate — set the long-term funding envelope, ensure interoperability with EuroHPC tiers — or it becomes another coordination body slowing exactly the kind of multi-year commitment the SSC says is missing.
The Proportionate Version
The part of the SSC's recommendation that should survive scrutiny is the funding-continuity argument: multi-year, non-project-tied money for the tiers Switzerland genuinely needs to own — university and regional-level compute, plus the governance and data-handling layer that only a domestic body can run. The part that needs more justification is full-stack duplication of a European network Switzerland has just spent five years working to rejoin. Bern should fund the commitment the SSC is right about and treat EuroHPC access as the top tier of the stack it's building, not a hedge to be replaced.