Beyond bigger models: why the UK and Switzerland matter in AI’s next phase

Read our London Tech Week blog from Swiss Business Hub UK.

Much of the AI conversation has been driven by scale: larger models, more parameters, faster inference and bigger training runs. Scale will continue to matter. But scale alone is not enough.

The next phase of AI will be shaped by ecosystems that can turn capability into real-world value. That means better tools in areas such as drug discovery, financial services, public services and robotics. It also means the full AI stack needs to work: talent, compute, hardware, data, open science and governance.

This is where I believe the UK and Switzerland have a strong role to play. Both countries have built exceptional talent pools. In Switzerland, ETH Zurich, EPFL, the University of Zurich and IDSIA have long been part of the global AI story. Shane Legg, co-founder of DeepMind, completed his PhD at IDSIA in Lugano. In the UK, world-class universities and research institutions continue to train and attract AI talent from around the world.

Global AI labs have followed that talent. The UK remains one of the world’s leading AI centres, with deep links between research, capital, policy and entrepreneurship. Switzerland has become a magnet for companies and research teams working on machine learning, robotics, computer vision, life sciences and AI safety. OpenAI, Anthropic, Google, Microsoft, IBM and many others have all invested in Swiss AI activity.

Infrastructure is the second part of the story. In the UK, AI hardware companies such as Graphcore and emerging players such as Fractile show that the race is not only about software. In Switzerland, the link is visible through both public infrastructure and talent. Cerebras, one of the most closely watched AI chip companies, counts Jean-Philippe Fricker, an EPFL alumnus, among its co-founders and senior technical leaders.

More importantly, Switzerland made a long-term bet on compute before AI infrastructure became the bottleneck everyone talks about. The Alps supercomputer at the Swiss National Supercomputing Centre in Lugano includes 10,752 NVIDIA Grace Hopper superchips and is one of the world’s most powerful supercomputers. It is also a key part of the Swiss AI Initiative launched by ETH Zurich and EPFL.

That infrastructure matters because the limits of scale are not only technical. They are also scientific. We are already using AI systems we cannot fully inspect or explain. As models become larger and more capable, the gap between what they can do and what we understand about how they do it may widen. Post-training and guardrails are important, but they cannot replace a deeper understanding of what models are trained on, how they are built and whether their development can be reproduced.

Richard Feynman famously wrote: “What I cannot create, I do not understand.” In AI, this becomes a question of trust: how can we fully trust systems if we cannot understand how they were created?

This is why Apertus matters. Developed by EPFL, ETH Zurich and CSCS, Apertus is Switzerland’s fully open foundation model. With a new version expected soon, many will ask whether it is more capable. The more important point is its commitment to open weights, open data, open science and reproducibility. Apertus is not trying to be another closed commercial assistant. Its value is that researchers, companies and institutions can inspect, challenge and build on it.

The UK-Swiss connection is also visible in applied AI. Isomorphic Labs, founded and led by Demis Hassabis, has its headquarters in London and its second location in Lausanne. Its work in AI-enabled drug discovery shows how machine learning can translate into impact in one of the world’s most important sectors. The field gained global visibility when the 2024 Nobel Prize in Chemistry recognised David Baker for computational protein design, and Demis Hassabis and John Jumper for protein structure prediction.

Finally, there is governance. The UK helped put AI safety on the global agenda with the 2023 AI Safety Summit. The global summit series has since continued through Seoul, Paris and India, with Switzerland set to host the next edition in Geneva in 2027, reinforcing Switzerland’s role as a platform for international dialogue on AI that serves people, science and society.

The UK and Switzerland are not just participating in the AI race. They can help define what the next phase should look like: technically ambitious, commercially useful, scientifically rigorous and grounded in trust.

For UK companies, investors, researchers and scale-ups looking to explore Switzerland’s AI, deep tech, life sciences, fintech or digital innovation ecosystem, we are here to help make the right connections.