AI and capital.

Investment in AI broadly follows three layers. Infrastructure — chips, data centres, power and networks — absorbs the largest sums and behaves increasingly like traditional infrastructure, with long horizons and heavy upfront commitments. Models and platforms sit in the middle, where a small number of well-funded companies compete at great expense and outcomes are highly concentrated.

Applications form the third layer: companies that use AI to serve a specific industry, function or customer. Entry costs are lower and the field is far wider, but so is competition, and defensibility is harder to establish. Each layer therefore carries a different risk and return profile, and serious investors are explicit about which one they are underwriting.

Access to a capable model is no longer an advantage in itself. Investors look for durable advantage beyond the model: proprietary data that improves the product over time, distribution that competitors cannot easily replicate, and a position in the customer's workflow that is costly to replace.

Unit economics receive close attention, including the cost of compute, which can erode margins that look healthy at small scale. Investors ask how gross margin develops as usage grows, how dependent the company is on a single model provider, and whether there is a credible path to revenue rather than a sequence of pilots. Clarity on these points often matters more than the ambition of the vision.

Much of AI does not fit conventional fund horizons. Infrastructure projects and deep technology often need patient capital that can wait for returns over many years. Family offices and other long-term investors are well placed to provide it, and many have built the expertise to do so.

Increasingly, these investors act directly: investing in companies and projects themselves, or co-investing alongside venture and infrastructure funds rather than only committing to them as limited partners. This brings them closer to founders and operators, and makes trusted relationships and first-hand judgement more valuable than ever.

Serious capital moves on trust. Large commitments rarely begin with a cold approach or a crowded event; they begin with an introduction from someone both sides respect, and continue through conversations where information can be shared candidly.

This is why curated, trusted rooms matter. Pitch events optimise for exposure: many companies, brief presentations, a broad audience. Deal rooms and one-to-one meetings optimise for fit: fewer parties, prepared in advance, with the time and confidentiality to explore terms. The path from introduction to transaction is shorter when the right people meet under the right conditions.

AI and capital is one of the four themes of Davos Catalyst, alongside AI infrastructure, enterprise AI, and sovereign and defence AI. The summit brings investors, family offices, founders and chief executives into one room, with private deal rooms and a personally curated meeting programme in place of stage presentations.

Davos Catalyst takes place from 12 to 15 January 2027 at the Rätia Center in Davos Platz. Participation is by invitation or application only.