menu
If you only looked at funding headlines in Q4, you could be forgiven for thinking AI has never been healthier. The months of October and November delivered a rapid succession of outsized rounds across the US, cutting through almost every layer of the stack. Reflection AI raised a $2bn Series B in October, while Anysphere, the company behind Cursor, secured a $2.3bn round in November. Fireworks AI, Sesame, Uniphore, OpenEvidence, Hippocratic AI, Lila Sciences and EvenUp all raised substantial rounds within the same quarter.
The pattern mattered as much as the numbers. Capital moved quickly and in dense clusters, often into companies that had already raised earlier in the year. Conviction remained high, competition for exposure intensified, and deployment timelines shortened. Whether this ultimately proves to be a bubble or the early phase of a generational shift remains open.
What Q4 clearly showed was a growing gap between the speed of capital deployment and the market’s ability to define durable long-term value.
Looking back at the year as a whole helps explain why Q4 felt so intense. According to Crunchbase, AI accounted for close to half of all global venture funding in 2025, up from roughly a third the year before. That level of concentration is rare, even in strong cycles, and it framed Q4 as a moment of acceleration rather than consolidation.
One of the more striking aspects of the Q4 funding wave was how widely it spread. Large rounds were no longer confined to foundation models or core infrastructure. Application-layer companies were pulled into the same acceleration. The scale of Anysphere’s raise of a massive $2.3 billion Series D in November illustrates how far valuations have moved even for product-led AI companies with clear user traction. In the Nordics Lovable raised its largest funding round so far, a Series B round of $330M in December.
The recent Crunchbase data shows that 58% of all AI funding in 2025 flowed into megarounds of $500m or more. In other words, capital has concentrated rapidly at the very top of the market, reinforcing a dynamic where a small number of companies absorb a disproportionate share of available funding.
Taken together, the Q4 raises point to an ecosystem prioritising speed and positioning. Capital is being deployed to secure exposure early, sometimes ahead of settled business models or competitive boundaries. That does not diminish the quality of the companies involved, but it does raise expectations. As capital intensity increases, the margin for error narrows, and the path from adoption to durable economics becomes more important than ever.
From an investor perspective, this kind of compression reflects strong belief in AI’s long-term importance, while simultaneously increasing pressure on teams to execute in increasingly crowded environments.
Away from software-only related funding narratives, some of the more interesting Q4 signals came from physical AI. Robotics, automation and embodied systems continued to attract attention from founders and investors focused on tangible outcomes and clearer economics. Applying AI in constrained, real-world environments is slower and more complex, but it also creates stronger barriers to entry.
Europe’s industrial robotics base remains central here. Universal Robots in Denmark, ANYbotics in Switzerland, and long-established players such as ABB and Siemens form part of a mature ecosystem already operating at scale across manufacturing, energy and infrastructure. These platforms provide a foundation for integrating more advanced autonomy and model-driven behaviour, grounded in real operational requirements rather than speculative use cases.
A more visible moment arrived in late October when Norwegian–US startup 1X opened pre-orders for Neo, a consumer humanoid robot positioned for household tasks. The announcement generated curiosity, scepticism and no small amount of humour online, which is often a sign that a technology category has entered broader public awareness. According to Sifted, Dealroom data shows humanoid robotics companies raised $3.2bn globally by the end of October 2025. That’s more than the previous six years combined.
Consumer humanoids remain early, but the attention and capital flowing into the category suggest embodied AI is being taken seriously again.
Europe did not mirror the US funding surge in Q4, reflecting a different capital landscape rather than a lack of ambition. Crunchbase data shows that 79% of global AI funding in 2025 went to US-based companies, making the scale of the imbalance hard to ignore.
Against that backdrop, one of Europe’s most consequential Q4 moves came through infrastructure rather than venture rounds. In December, the European Commission and the European Investment Bank signed a memorandum of understanding to advance the €20bn InvestAI initiative, setting the course for up to five AI “gigafactories” across the EU. The focus now shifts to turning ambition into projects that can support real training workloads at scale.
This approach reflects where Europe can realistically compete. In industrial systems, energy, mobility and regulated markets, success depends less on who trains the biggest model and more on who can deploy AI reliably in production environments. For founders building in these environments, shared AI infrastructure is moving from policy aspiration to operational necessity.
Q4 sharpened the picture of where the AI market stands heading into 2026. Capital at the frontier continues to accelerate, valuations continue to stretch, and competition for exposure remains intense. At the same time, some of the most compelling long-term opportunities are emerging outside the most crowded parts of the software stack.
Physical AI sits firmly in that category. It demands patience, engineering rigour and a willingness to work through complexity. It also offers clearer paths to defensibility and real-world impact. The coming years will favour teams that can translate intelligence into systems that operate reliably in the real world, over time and at scale.
