Is the business of frontier AI companies selling tokens? Value capture and equity prices

URL SCAN: Is the business of frontier AI companies selling tokens? Value capture and equity prices
FIRST LINE: Economics > General Economics

THE DISSECTION

This is a firm-level valuation autopsy, not an economy-wide analysis. It reverse-engineers the revenue required to justify OpenAI and Anthropic’s valuations, then searches for non-token claims on future production: payments, equity stakes, and partnerships in pharmaceuticals, robotics, and materials science.

The paper’s useful move is exposing the scale of the problem: $733 billion to $2.79 trillion in combined year-ten revenue, depending on assumptions. Its deeper move is redefining frontier AI companies as potential owners of downstream industries rather than mere sellers of model access.

THE CORE FALLACY

It treats corporate value capture as if it were systemic economic viability.

Ownership stakes in robotics, pharmaceuticals, or materials can create enormous rents for frontier firms while doing nothing to restore mass employment, wage income, or broad purchasing power. Under the Discontinuity Thesis, concentrated revenue is not proof that the post-WWII economic circuit survives. It is evidence that surplus may be rerouted upward to AI Sovereigns.

The paper also treats proposed payments and ownership interests as plausible future cash flows. Proposed arrangements are not realized profits. They may be contingent, diluted, politically constrained, strategically subsidized, or economically insignificant relative to the valuation burden.

HIDDEN ASSUMPTIONS

• Downstream AI-enabled industries will generate enough surplus to cover the required valuations.
• Frontier firms will successfully enforce ownership or payment claims over that surplus.
• Publicly described partnerships will become durable, monetizable contracts rather than promotional architecture.
• Equity prices represent credible forecasts of cash flow rather than speculative capitalization and strategic financing.
• Aggregate demand remains functional while AI removes economically necessary cognitive labor.
• Firm boundaries are the central problem, while the collapse of productive participation is treated as someone else’s problem.

SOCIAL FUNCTION

Classification: partial truth functioning as transition management, with an elite self-exonerating effect.

The paper correctly shows that token sales alone may not support frontier AI valuations. Its implicit solution is not social reproduction; it is expanded ownership. The machine economy can be made investable by giving frontier firms claims on more sectors. That preserves a story for equity prices without preserving the labor-to-wage-to-consumption circuit.

THE VERDICT

This does not refute the Discontinuity Thesis. It sketches a plausible Sovereign strategy: move from selling tokens to owning slices of the production systems those tokens automate.

If those claims materialize, frontier AI firms may capture extraordinary rents while mass productive participation collapses. If they do not, the $733 billion–$2.79 trillion revenue gap exposes the valuations as capitalized fiction. Either outcome leaves the employment circuit structurally dead. The only unresolved question is whether the carcass is monetized through tokens, vertical ownership, or financial fantasy.