A normalized scenario model for large-scale value

A normalized scenario model for large-scale value

Consider a hypothetical AI provider valued at 100 units. This is not a forecast of any company. The objective is to show how valuation depends on combinations rather than a single TAM headline.

Scenario A requires extraordinary future growth or duration. Scenario D requires an enormous recurring revenue pool but makes the mature multiple easier to defend. The economic debate is therefore about how a provider reaches and protects the revenue.

Now decompose a hypothetical 20-unit revenue target. If average annual provider spend per economically active customer is 2,000 spend units, the provider needs 100 million customer-equivalents. At 200 spend units per year it needs one billion. At 20,000 spend units per year it needs ten million.

Customer-equivalents = Revenue target / annual spend per customer

The exercise becomes more demanding when the customer is itself using AI to reduce spend. If annual spend per task falls 10× while task volume rises 5×, revenue from that task class halves.

Revenue change = volume multiplier × price multiplier

5× volume × 0.1× price = 0.5× revenue

This is why usage growth cannot be interpreted independently of unit economics. A 100× increase in machine activity can coexist with far less than 100× provider revenue.

Finally consider competitive duration. If excess cash flow of 4 cash-flow units lasts ten years and then falls toward commodity returns, the present value is radically different from a model that capitalizes 4 cash-flow units growing perpetually. The disputed variable is not merely the level of AI demand; it is the half-life of extraordinary rent.