Frontier AI does not inherit the fields it transforms
A common response to commoditization is that the frontier continually moves. Routine coding becomes cheap, but frontier AI discovers drugs. Drug discovery becomes cheap, but frontier AI designs materials. The provider always occupies the next high-value layer.
This can sustain a valuable business. But it does not justify inheriting the historical revenue of each field. When frontier intelligence enters a field, it may reduce the field's cost structure.
Historical field spend ≠ post-transformation AI TAM
If a $100 billion professional process becomes a $20 billion automated process, the provider competes for some share of the $20 billion plus any newly created demand. The missing $80 billion is productivity released to the rest of the economy, not automatically provider revenue.
The frontier-provider thesis therefore requires a race: new monetizable frontiers must open faster than transformed fields become cheaper, competitive or compiled into lower-cost execution.