Economist Steve Hanke says AI won't destroy most jobs because it costs more than hiring humans

TEXT START: A hot potato: A top economist has said that AI will not destroy most of the world's jobs, and his rationale seems totally logical: doing so would simply be far more expensive than employing humans.

1. THE DISSECTION

The article converts a legitimate short-run constraint—AI is expensive, resource-intensive, and often overhyped—into a long-run defense of mass employment. It treats job replacement as an all-or-nothing event in which companies must fire everyone and purchase an AI substitute. That is not how automation works. Firms automate profitable tasks first, then expand as costs fall and systems improve.

The rehiring examples show deployment failure, not human economic invulnerability. Early systems can be overpriced, unreliable, or badly integrated. Those are lag effects, not proof that the underlying substitution mechanism is dead.

2. THE CORE FALLACY

Hanke compares the total cost of an AI system with the wage of a human worker, as though the relevant question were whether AI is cheap in absolute terms. The relevant question is whether AI can produce a given output more cheaply, reliably, and at greater scale than the human alternative.

High capital and energy costs may slow adoption. They do not preserve the mass employment circuit if AI achieves durable cost-performance superiority. They can instead concentrate productive power among firms that control capital, energy, infrastructure, and models. That strengthens the Sovereign class; it does not rescue labor.

The claim also confuses “AI will not replace every worker immediately” with “AI will not eliminate economically necessary work for most people.” The first may be true. The second does not follow.

3. HIDDEN ASSUMPTIONS

• AI’s current resource costs will remain permanently high rather than declining through efficiency, scale, and specialized hardware.
• Human wages, benefits, management overhead, errors, turnover, and availability are fixed and cheap.
• Replacement must occur across entire occupations rather than through task-level automation.
• Temporary AI failures establish a permanent human advantage.
• Scarce energy and physical capital will block AI rather than ration it toward the highest-value applications.
• If consumption is maintained through transfers, productive participation has also been preserved.
• Firms will value human employment as an end in itself when a smaller workforce can produce the same output.

4. SOCIAL FUNCTION

Classification: partial truth functioning as ideological anesthetic and elite self-exoneration.

The article accurately identifies AI’s physical costs and the industry’s promotional dishonesty. Its social function is still reassuring: it turns deployment friction into a claim that the existing labor order will survive, allowing policymakers and institutions to postpone the harder question of what happens when most people are no longer economically necessary.

It also gives capital a convenient narrative. Failed implementations become evidence that workers are indispensable, rather than evidence that the transition is immature and uneven.

5. THE VERDICT

Hanke has a valid cost objection and an invalid systemic conclusion. Expensive AI can still destroy mass employment through selective automation, falling unit costs, and capital concentration. Resource scarcity may delay the break and make AI less freely available; it does not reverse the Discontinuity Thesis.

This is a credible short-run accounting argument miscast as a long-run survival theorem. The article mistakes hospice care for recovery. Under P1–P3, the decisive question is not whether AI is cheap, but whether human labor remains economically necessary. If it does not, the wage-to-consumption circuit is severed regardless of how expensive the machinery is.