Valuing AI Start-ups

With AI start-ups, classic valuation logic often misses the mark: the companies are young, revenues are small or not yet representative, and technology cycles move faster than any planning horizon. Arguing with revenue multiples or DCF models alone misses what investors are actually looking at.

There are three reasons why traditional metrics fail. First, today’s revenue says little about the potential when the product is only just reaching the market. Second, high compute and development costs distort the cost base at a time when the company is deliberately investing. Third, a single advance by the major foundation model providers can devalue entire product categories within months – a risk that appears in no business plan.

What investors really look at

In our processes, we see four recurring valuation anchors: the technical depth and speed of the team; access to proprietary data and the rights to it; the product’s defensibility against the foundation models – in other words, whether the company is building a genuine layer of value creation of its own or merely wrapping a thin shell around other people’s models; and finally, early commercial signals such as pilot customers, intensity of use and willingness to pay.

“What is valued is not today’s revenue, but the credibility of the path to a defensible business model.”

What this means for founders

For founders, this means that the valuation discussion is conducted through the market narrative and the milestones, not the spreadsheet. A realistic plan up to the next financing stage, clearly defined technical and commercial milestones and, in early phases, a suitable structure – such as SAFE or convertible instruments with a valuation cap – move the formal valuation question to where it can be better answered: the follow-on round, with evidence instead of claims.