AI research in Switzerland is organised around the Swiss Federal Institute of Technology in Lausanne (EPFL) and the Swiss Federal Institute of Technology in Zurich (ETH Zurich). Since December 2023, they have led the Swiss AI Initiative, which brings together more than 800 researchers according to its website. It computes on Alps, the supercomputer of the Swiss National Supercomputing Centre (CSCS) in Lugano. Institutes outside the federal institutes of technology and two public agencies complete the picture.

The discoveries made by these laboratories are followed as they happen in our Swiss Research section. This page describes the framework: who does the research, with which programmes, what computing power and what money.

Who does AI research in Switzerland?

The two federal institutes of technology concentrate most of the resources. Each has its own centre: the ETH AI Center in Zurich and the EPFL AI Center in Lausanne. On 3 October 2024, EPFL announced that the two institutions were jointly founding the Swiss National AI Institute (SNAI), a national institute built on these two centres.

According to that announcement, the SNAI draws on the expertise of more than 70 professors specialising in AI across the country. It works with the CSCS and with the Swiss Data Science Center (SDSC). Outside the federal institutes of technology, two older institutes hold a special place: Idiap in Valais and IDSIA in Ticino.

PlayerLocationAffiliationPublished benchmark
Swiss AI InitiativeLausanne and ZurichEPFL and ETH Zurich, operated by the SNAIMore than 800 researchers, more than 10 institutions (initiative website)
Swiss National AI Institute (SNAI)Lausanne and ZurichETH AI Center and EPFL AI CenterFounded in October 2024 (EPFL, 3 October 2024)
CSCSLuganoNational computing centre, SNAI partnerAlps, more than 10,000 GH200 chips (initiative website)
Idiap Research InstituteMartignyIndependent foundation, affiliated with the EPFL doctoral schoolMore than 180 staff, 17 research groups (Idiap website)
IDSIALuganoUniversità della Svizzera italiana (USI) and SUPSISix research areas (SUPSI website)

What is the Swiss AI Initiative?

The Swiss AI Initiative is the two institutions' joint programme for foundation models, the large models that serve as the basis for generative AI. It was launched in December 2023, with a press release dated 4 December 2023, the SNAI announcement recalls. Its website mentions an initial allocation of more than 10 million hours of computing time on graphics processing units (GPUs), provided by the CSCS.

The SNAI has operated it since its creation. According to the October 2024 announcement, the models are meant to serve health, sustainability, science, education, robotics and augmented reality. The stated goal is to build models with more than 50 billion parameters, including a first national foundation model for languages. Researchers are to release their models as open source whenever possible, so that SMEs and start-ups can use them.

The initiative's website presents it as the world's largest open science effort for foundation models. That is its own assessment, without any figures for comparison. Computing time is obtained through regular calls: the fourth major call for projects was open from 3 August to 14 September 2026.

What role do the CSCS and its Alps supercomputer play?

The CSCS provides the initiative's computing power. The Lugano centre operates Alps, which the initiative's website describes as one of the world's leading AI supercomputers, with more than 10,000 Nvidia GH200 chips. According to the technical sheet cited by 2clics in its feature "Swiss language model: Apertus and the other projects", Apertus was trained on 4096 of these chips.

The CSCS claims a role that goes beyond supplying machines. "Apertus is not a classic case of technology transfer from research to product," said its director Thomas Schulthess, a professor at ETH Zurich, at the model's launch on 2 September 2025.

Which AI research institutes exist outside the federal institutes of technology?

According to its history page, the Idiap Research Institute in Martigny was founded in 1991 as the Dalle Molle Institute for Perceptual Artificial Intelligence, through a collaboration with the Dalle Molle Foundation. It became an independent foundation in 1996 and in 2002 developed Torch, a free scientific computing software presented as a forerunner of PyTorch. It has been affiliated with an EPFL doctoral school since 2009 and today has more than 180 staff spread across 17 research groups.

The Dalle Molle Institute for Artificial Intelligence (IDSIA), in Lugano, is attached both to the Università della Svizzera italiana (USI) and to the Department of Innovative Technologies of the University of Applied Sciences and Arts of Southern Switzerland (SUPSI). According to the SUPSI page, it conducts fundamental and applied research with industry. Its areas range from machine learning and neural networks to natural language processing, computer vision and autonomous robotics. It is headed by Andrea Emilio Rizzoli.

Who funds AI research in Switzerland?

The ETH Domain directly funds the Swiss AI Initiative. According to the announcement of 3 October 2024, the ETH Board decided to allocate 20 million francs to the Swiss AI Initiative for the years 2025 to 2028. EPFL, ETH Zurich and third-party funds supplement this funding.

The Swiss National Science Foundation (SNSF) funds fundamental research through projects. In an analysis published on 17 July 2025, it found that the AI-related research it funds increased markedly between 2011 and 2024.

Innosuisse, the Swiss Innovation Agency, instead funds projects carried out with companies. Its 2025 annual report, published on 15 April 2026, reports a federal contribution of 305.1 million francs, plus 28.7 million as a transitional measure linked to Horizon Europe. Applications for national innovation projects reached around 1100, roughly 35% more than the previous year. Innosuisse cites a flagship initiative devoted to AI in the life sciences. On 19 September 2025, it selected four flagship projects from its 2024 call on AI in the life sciences, focused on human health.

Private and foreign funding supplements this public base. Swisscom, the country's largest telecommunications operator, is the strategic partner of the Swiss AI Initiative and contributes to the funding of Apertus, according to EPFL. Google made a donation of 500,000 francs to the ETH AI Center in 2026, according to the Neue Zürcher Zeitung (NZZ). EPFL's LogSync neural chip, for example, is part of a 6.6 million franc grant from the US National Institute of Mental Health.

Key figures

  • More than 800 researchers, including 70 professors specialising in AI, from more than 10 institutions (Swiss AI Initiative website).
  • More than 10 million GPU computing hours at launch, in December 2023, followed by 20 million francs allocated by the ETH Board for 2025 to 2028 (EPFL, 3 October 2024).
  • More than 10,000 GH200 chips in Alps, at the CSCS (initiative website).
  • 305.1 million francs in federal contribution to Innosuisse in 2025 (2025 annual report).
  • More than 180 staff and 17 research groups at Idiap (Idiap website).

What does this research actually produce?

The most visible result is Apertus, a fully open large language model. EPFL, ETH Zurich and the CSCS released it on 2 September 2025, in two sizes of 8 and 70 billion parameters. According to the announcement, it was trained on 15,000 billion tokens in more than 1000 languages, 40% of them non-English data, including Swiss German and Romansh. A version 1.5 followed on 24 July 2026.

The initiative's computing power also serves other disciplines. On 28 October 2026, EPFL presented spora, a tissue atlas of cancer, whose AI models were trained with resources from the Swiss AI Initiative. Like LogSync, this work remains at the research stage: no clinical application has been announced.

What are the weaknesses of Swiss AI research?

The first weakness lies in trained talent leaving. According to the NZZ, as reported by ICTjournal on 26 October 2026, many graduates join American groups, in Switzerland or in the United States. Andreas Krause, director of the ETH AI Center, sees this as a risk to digital sovereignty, as detailed in our article on AI specialists in Switzerland.

The SNAI itself makes training a priority. Its 2024 announcement estimates that only a few hundred people in the world know how to develop the largest foundation models. Moving from the laboratory to the market poses another difficulty. In our feature on Swiss robotics and artificial intelligence, Jan Kerschgens, co-director of the Swiss Center for Digital Sovereignty, said that "it is neither ideas nor talent that are lacking, it is domestic growth capital".

What we don't know yet

  • The funding of the Swiss AI Initiative after 2028: the ETH Board allocation announced in October 2024 ends on that date.
  • The projects selected in the fourth major call, which closed on 14 September 2026, do not yet appear on the initiative's home page.
  • The number of AI graduates leaving Swiss research: according to the NZZ, no precise figure exists.

Sources