🤚 The Open-Palm Paradox

In 1865, English economist William Stanley Jevons observed something profoundly counterintuitive: when coal-burning engines became more efficient, total coal consumption increased. The cheaper it got to use energy, the more energy people used. He called this the “rebound effect.” We call it Tuesday in the AI industry.

The Jevons Paradox has officially arrived in artificial intelligence, and it brought receipts. The cost of running an AI inference has plummeted — API prices have dropped by more than 90% over the past two years. Models that once required data center-scale budgets now run on hardware you can buy at a consumer electronics store. Intelligence, as a commodity, has never been cheaper.

And yet — and yet — total AI spending is not declining. It is exploding. Global AI infrastructure investment is projected to exceed $300 billion in 2026. Compute demand is growing faster than compute supply. GPU shortages persist despite multiple new chip entrants. The paradox isn’t theoretical anymore. It’s in the quarterly earnings reports.

👐 The Two-Handed Reality Check

Here’s the mechanism, stripped of academic language: when intelligence gets cheaper, it doesn’t replace existing use cases at a lower cost. It creates entirely new categories of use that didn’t exist before. A legal document review that cost $50,000 in billable hours and wasn’t worth automating at $10 per API call becomes absolutely worth automating at $0.10 per call. Multiply that logic across every industry, every workflow, every tedious cognitive task that was previously “not worth the trouble.”

The math is ruthless:

• A 10x reduction in inference cost doesn’t produce a 10x reduction in spending — it produces a 100x increase in usage
• Every new efficiency milestone unlocks markets that were previously priced out of AI entirely
• The cheaper the model, the more creative people get with applications — and creativity scales faster than cost savings

This is why every company that has tried to “save money with AI” has ended up spending more on AI. The savings are real — they just get immediately reinvested into doing more things with the technology. Your AI budget didn’t shrink. Your AI ambitions grew to fill the budget you freed up.

🌿 The Gentle Awakening

The labor implications are where this gets genuinely interesting — and where the paradox offers its most counterintuitive prediction. As AI gets cheaper, demand for human work doesn’t necessarily decrease. It restructures.

When cognitive automation becomes cheap enough, entirely new workflows emerge that require human judgment, orchestration, and oversight. The universe of “work worth doing” expands. Every AI deployment needs someone to design it, deploy it, manage it, iterate on it, explain its outputs to a skeptical VP, and apologize when it hallucinates the company’s quarterly revenue.

Research has consistently shown that as the cost of cognitive work falls, human attention moves up the value chain — toward judgment, decision-making, and the uniquely human talent of telling a machine what to do and then disagreeing with its answer. The machines get cheaper. The humans who direct them become more valuable. Jevons would have found this delightful, assuming he could have been explained what a GPU is.

👑 The Crown Verdict

The Jevons Paradox tells us something the AI industry desperately needs to hear: efficiency does not lead to less consumption. It leads to more. The cost of intelligence is collapsing, and demand is responding exactly as 160 years of economic theory predicted — by going vertical.

If AI becomes 10x cheaper again — and it will — the question isn’t whether demand will increase. It’s whether our infrastructure, our institutions, and our power grids can absorb the demand that cheap intelligence creates. We built a product so efficient that using it consumes more resources than the inefficient thing it replaced. Jevons is somewhere nodding, coal dust on his lapels, whispering: “I told you so.”

The companies building AI models aren’t selling efficiency. They’re selling demand creation. And if history — specifically, 1865 to present — is any guide, that demand has no ceiling. Only a floor that keeps dropping.

Inspired by AI Just Proved Jevons Paradox by Peter Diamandis.

“We made intelligence 90% cheaper and somehow spent 300% more on it. The economists aren’t surprised. The CFOs are inconsolable.” — The Slap of Wisdom Department of Paradoxical Economics, drafting a budget that already needs revising

Your paradox is showing. Consume wisely.