Andreas Cleve, saus his company Corti is building foundation models specific to healthcare
Coinciding with last week’s AI Action Summit in Paris, new figures from Dealroom point to a massive increase in artificial intelligence investment at a time when VCs have been paring back their dealmaking in other sectors. According to the data, AI companies raised $110 billion in 2024, representing a 62% increase from the year before. However, the bulk of this capital – $80.7 billion to be precise – was directed at US companies, with investment into Europe and China coming in at $12 billion and $7.6 billion respectively. In the case of generative AI, 83% of the money raised went to American startups and scaleups.
So where does all this leave European ventures? Given that much of the money invested has gone into foundational work – such as the development of large language models – largely carried out in the U.S., can Europe’s companies compete?
Well, once the foundational work has been done, there should, in theory, be countless opportunities for startups that can create applications and use cases. That was a view expressed immediately after the AI Action Summit by Robert Lachlan, founding partner of Visionaries Club, a Berlin and Londo-based VC fund.
“Foundational models have absorbed the lion’s share of venture capital funding in 2024, but we believe investment in application layer businesses will skyrocket in 2025,” he said.
Joel Hellermark agrees. He is founder and CEO of Swedish Sana Labs, a company providing an AI platform that enables businesses to reach into their many applications to not only to access data but also to compile research, write reports and prepare documentation. As he sees it, Europe has the talent base to create solutions that deliver a tangible return on investment.
AI You Can Use
“Europe has the talent to take a global position in Applied AI,” he says. “When it comes to foundational research we don’t have the funding but we can take advancements and create models that people want to use.”
To that end, Sana – which has raised $130 million in VC capital to date – has partnered with OpenAI to create its applications. Hellermark says the key to winning business is the provision of products that can demonstrably deliver an ROI to buyers. For instance, Sana says that by using AI agents to research, process and output information, customers can achieve productivity gains of 90%.
But there is more to the European AI story than startups using U.S.-developed language models as the basis for their applications. Jarek Kutylowski is a founder and CEO of translation company DeepL. The company has developed its own LLMs, but as he stresses it has specialised in the information required to build translation services. This, he says, is a differentiator.
Focusing On Industries
“While many translation and AI tools are powered by general-purpose models designed for a wide range of tasks, the DeepL Language AI platform is purpose-built for communication and language use cases,” he tells me in response to emailed questions. “DeepL Translator is powered by our next-generation language model, which leverages proprietary LLM technology finely tuned language.”
As he describes it, the models are tutored by language experts. He argues this creates a knowledge base that tends to be superior to translation platforms powered by more generalised LLMs.
A similar approach is taken by Corti, a Copenhagen-based company providing what it describes as a co-pilot for healthcare professionals. The mission of the business is to use AI to help doctors and other practitioners with patient assessment and clinical decision-making. The company has developed its own LLM. One that is specific to healthcare.
“We have built something from the ground up,” says CEO and co-founder Andreas Cleve. “We are a foundation model vendor, but it is a model that is built for a trillion-dollar market called healthcare. Our LLMs focus on one aspect of reality.”
Cleve says European companies have an opportunity to thrive by working on foundational models that home in on specialist fields. This, he says, is something that plays to Europe’s strengths. “ In Europe, we are good at the hard problems rather than the general, mass-produced product. A Rolex rather than Timex.”
There is perhaps a blurring of the lines here between foundation and application. Businesses working in specialist fields must also enable customers to use the technologies at an application level. To this end, Corti solutions can be integrated via APIs to existing medical systems. DeepL also seeks to bridge the gap.
“We focus on delivering high-impact applications through our Language AI platform. By bridging foundational AI development with enterprise-grade applications, we ensure our technology is both cutting-edge and directly valuable to businesses, governments, and individuals worldwide,” says Kutylowski.
Usability will undoubtedly be a key battleground. Hellermark says AI tools are still at an early stage in terms of the way that users interact with the technology. “There is a massive opportunity to define the UI and the OS for AI. It’s like we are still in 1984 when we have yet to see a truly personal computer. We mainly interact with AI through chat and we have fairly simple integrations with the rest of the ecosystem. Europe has the talent to take a global position in Applied AI,” he argues.
The three scaleups featured in this article have already indentified their routes to market and found customers. For many companies at earlier stage, that challenge lies ahead.
