The most important AI companies of the next decade may look far less like software startups and far more like industrial infrastructure companies.

Artificial intelligence is moving beyond chat interfaces and copilots into power grids, robotics, manufacturing, scientific research and industrial systems, where reliability, precision, operational robustness and deployment matter far more than product demos. 

AI should not be treated as a platform shift in the same way as mobile or the browser. The more useful distinction is between companies that merely wrap model APIs and those that use AI as a micro-process inside much broader products or systems. In industrial AI, the model is rarely the product. The advantage comes from the full system around it: data, workflow integration, reliability, deployment and defensibility.

Increasingly, the challenge is no longer just building smarter models, but building the infrastructure and compute systems needed to tackle problems traditional software struggles to solve.

And many of the companies driving that shift are being built in Europe.

To explore where this transition is happening first, Atlantic, KOMPAS VC, Lowercarbon Capital, Maze, Norrsken VC, RAISE Ventures and SET Ventures came together to publish The Future in Focus: Advancing AI.

The report brings together leading early-stage investors across Europe and the US to spotlight the founders and technologies pushing frontier AI into the real economy. 

The first edition of Future in Focus in 2025 featured startups of which more than half raised follow-on funding within 12 months. This second edition focuses on the widening divide between companies merely applying AI and those genuinely advancing the field itself.

That divide is becoming more visible in the market. Many AI-native startups can grow quickly. But if the end-user experience is essentially tokens behind a thin interface, defensibility is weak. The more durable companies are those where AI enables something previously impossible or uneconomic, such as lower-cost expert services, autonomous industrial workflows or real-time operational decision-making. 

Against that backdrop, the report focuses on builders applying state-of-the-art models, proprietary data and frontier infrastructure to some of the hardest problems across energy, climate adaptation, manufacturing, healthcare, governance and scientific discovery.

In many cases, these founders are working on systems where scale is constrained less by software distribution and far more by physics, operational reliability and industrial deployment.

As the report puts it: “While every startup claims to be ‘AI-native’, there is a smaller group of builders who are genuinely advancing the field of AI and applying solutions to real-world problems.”

From NeoForge building AI-driven manufacturing systems that simulate and optimise manufacturing parameters before physical production begins, to Qualia developing infrastructure for the next generation of robotics and vision-language-action models, the report follows founders building at the frontier of industrial and applied AI.

It also features companies like Grunuss applying generative AI and quantum-native simulation to next-generation superconductors, Gridraven bringing probabilistic weather intelligence to energy infrastructure, and Alien Intelligence building privacy-preserving infrastructure for enterprise AI and proprietary data access.

Across the report, the same pattern keeps emerging: the next generation of AI is being shaped by proprietary data, infrastructure, deep domain expertise and systems built for real-world deployment rather than demos alone.

AI alone is not a moat. Durable advantage still comes from network effects, switching costs, proprietary data, economies of scale and deep workflow embedding.  

The companies in this report are not simply using frontier models, but building systems that become harder to replace as they accumulate data, integrate into operations and improve with scale.

One of the clearest arguments running through the report is that Europe may be structurally well positioned for this next era of AI.

Many of the industries now being reshaped by AI already have deep roots across the continent: manufacturing, industrial automation, energy systems, scientific research, climate infrastructure and advanced engineering. 

As AI moves beyond consumer interfaces and deeper into complex physical systems, Europe’s combination of industrial depth, engineering talent, research infrastructure and domain expertise could become a major structural advantage. As compute, energy infrastructure and scientific complexity become increasingly central to AI, these strengths may matter far more than many currently assume. 

The report’s conclusion captures the core thesis directly: “Exciting opportunities sit where frontier AI meets real industrial depth.”

But the shift also comes with a harder economic question. AI companies may reach revenue faster than previous software startups, yet inference, compute and deployment costs can make margins fragile. The strongest companies will be those that can show not only technical ambition, but credible unit economics, pricing power and a path to lower marginal costs over time.