AI Perspectives
The AI capital supercycle: where venture money is really flowing
AI now dominates venture capital. We map where the money goes — infrastructure, models and applications — and why the application layer is underpriced.
Artificial intelligence has become the centre of gravity for global venture capital. A large share of every new dollar now goes to AI companies, and a handful of mega-rounds for model labs and infrastructure providers capture the headlines.
That concentration hides an opportunity. Capital is crowding into the most capital-intensive layers — chips, data centres and foundation models — where returns depend on enormous scale. Meanwhile, the application layer, where agents meet specific customers and specific workflows, is still priced like ordinary software.
We believe the next decade's outsized returns will come from companies that turn general-purpose models into specialised agents for neglected markets. These businesses need less capital, reach revenue sooner and compound proprietary data with every customer.
Frequently asked questions
Why invest in AI applications rather than foundation models?
Applications require far less capital, reach revenue faster and build moats through proprietary data and workflows, while foundation models demand massive scale.
