Switzerland and AI: tiny but mighty

While the US and China battle for AI supremacy, Switzerland is charting a different course - one that prioritizes quality over scale, public-private collaboration, and ethical AI development.

Meet the Stargate Storm. President Trump and his administration have shouted loud and clear their intention to lead the GenAI race, through heavy investments and an industry-led approach. On the other side of the globe, China has made massive waves in the news and financial markets with the introduction of DeepSeek, a GenAI model that is just as powerful as its American competitors (ChatGPT, Claude, Gemini, etc.) but less energy-hungry and based on an open-source approach. And right in the middle stands Switzerland, a tiny spot on the world map. With its numerous pole positions in innovation rankings, can it take its place in GenAI?

The newly created American tech alliance will have an undeniable impact on innovation, with the top tech firms likely to push their own agenda forward. Switzerland has so far taken a more nuanced approach to innovation, combining the efforts of both the public and private sectors to foster innovation in a balanced and collaborative way.

Technological advancements are backed by heavy contributions from both sectors. With R&D expenditure amounting to 3.3% of GDP, Switzerland ranks among the top five countries globally. Two-thirds of this spending comes from the private sector. Other crucial contributions come from academia, in the shape of talent growth, patents and the creation of startups. These contributions are supported by government initiatives such as the Swiss National Science Foundation providing grants to promising researchers, Présence Suisse, Swisstech and Switzerland Global Enterprise - all of which promote Swiss startups on the international stage.

Specific entities have been created to facilitate collaboration and support the country's vision for digital sovereignty. EPFL and ETH Zurich are teaming up with other universities to tap into a growing network of over 200 AI experts nationwide. The aim is to tackle the societal challenges brought about by AI. One of their main missions is to develop GenAI models such as large language models (LLM) specifically for Swiss society, with the ambitious aim of building a safe, trustworthy and transparent AI.

A Swiss-made LLM

"We're working on a model that is multilingual, transparent and open-source, and better suited for our Swiss private and public institutions." says Martin Jaggi, professor of machine learning at EPFL and member of the Swiss AI Initiative Steering Committee. "If you look into current models out there, they're mostly trained in English. If we take Meta's LLama, English makes up around 90% of its data. Our model is currently being trained in more than 1,000 languages."

Aside from being multilingual, another aspect conveying the Swissness of the model lies in its trustworthiness and transparency. The Swiss AI Initiative wants to remain transparent about how data is used and processed, and to ensure compliance with the relevant regulations in Switzerland and Europe. Full disclosure on data inclusion and exclusion characteristics is a key goal as well. "Models inherit all their strengths and weaknesses from the training data, so we want to be very explicit in making this transparent," says Jaggi.

The federal government has an ongoing plan to keep Switzerland at the forefront of computing capacity dedicated to scientific research. To that end, the Swiss National Supercomputing Center started building Alps - a supercomputer that's currently ranked seventh out of the top 500 supercomputers worldwide and that's the second most powerful supercomputer in Europe. As a public asset, Alps is open to the broad community of researchers in Switzerland and beyond. The Swiss LLM will be available by this summer and will undergo approximately three million GPU hours of training.

Switzerland may not replicate Silicon Valley's fast-paced, profit-driven model, but does it really need to? Its value-driven approach to innovation contrasts sharply with the US and China's strategies, offering a unique blend of quality, ethical consideration and collaboration.

While Switzerland cannot compete with the scale of investment and computation of the US, it has chosen quality over quantity. "Switzerland may not replicate Silicon Valley's fast-paced, profit-driven model, but does it really need to? Its value-driven approach to innovation contrasts sharply with the US and China's strategies, offering a unique blend of quality, ethical consideration and collaboration." says Patrik Wermelinger, Chief Investment Promotion Officer of Global Switzerland Enterprise. This positions Switzerland as a distinct and sustainable player on the global innovation stage. "With its strong focus on R&D, computational excellence and talent cultivation, Switzerland is forging a path that aligns innovation with long-term societal benefits." says Wermelinger.

Jaggi adds: "The DeepSeek news is encouraging for us here, in Europe. It's an indication that you can produce a high-quality model without the means of big multinational companies."

As the world seeks more sustainable and inclusive models of technological advancement, Switzerland stands out as a beacon of possibility. The fast-paced technology industry shows us that everything can change from one minute to the next - but one thing is certain: Switzerland's long-standing history of innovation is far from over, and its global impact will continue to grow.

References

This article was published in the March 2025 issue of Dimensions, an EPFL magazine that showcases cutting-edge research through a series of in-depth articles, interviews, portraits and news highlights.