When AI Makes Thinking Cheap, Scarcity Moves to Energy, Land, and Materials

Every few months, a new quote goes viral: AI will collapse prices, the old economic system will break, GDP won’t matter, scarcity is over. The emotional punch is understandable. If a model can write, code, design, translate, and reason on demand, it feels like the world should become “free.”

But there’s a cleaner way to think about what’s actually happening in 2026:

Scarcity doesn’t disappear. It relocates.

AI is a productivity engine. It makes many kinds of information work cheaper. Yet it does not repeal physics, permits, logistics, land constraints, grid stability, or mineral processing capacity. The bottleneck simply moves from “human bandwidth” to “physical delivery.”

What AI really makes cheap

AI compresses the cost of tasks that are mostly information transformation: drafting, summarizing, customer support, basic research, translation, first-pass design, and a long list of white-collar workflows. These jobs used to be constrained by time: a person can only read so much, write so much, think so much in a day.

Now the constraint is different. The first draft is no longer scarce. The scarce thing becomes trust (is it correct?), accountability (who owns the outcome?), and distribution (who gets attention?). That’s why you can see “AI deflation” in certain services without seeing a universal collapse in all prices.

What AI does not make free

Even if AI makes decision-making and design faster, real-world output still has to pass through hard constraints:

energy (power generation, grid capacity, storage), materials (copper, aluminum, steel, lithium, nickel, rare earths), land (location and connectivity), and infrastructure (ports, rail, roads, water systems, data centers, maintenance crews).

You can generate infinite versions of a factory blueprint. You cannot instantly generate the factory’s transformers, permits, water access, logistics lanes, and reliable power. The physical world has lead times. It also has politics. Both matter.

The counterintuitive part: efficiency often increases total demand

There’s a common mistake in “AI will cause price collapse” arguments: they assume higher efficiency means less total resource use. History often runs the opposite way. When something becomes cheaper and easier, people do more of it.

AI lowers the cost of planning, coordination, and operations. That can accelerate scale: more automation, more compute, more deployment, more “projects that used to be too expensive to attempt.” The result can be a world where the information layer deflates while the physical layer experiences scarcity pressure and volatility.

In plain terms: we may get cheaper “thinking,” while “power and throughput” become the new hard currency.

The real power shift: the ability to turn bits into atoms

If AI is the new production force, then the production means are the systems that let you ship reliably in the real world. The winners are less likely to be “the people with the smartest model” and more likely to be organizations that can do four things consistently:

(1) Secure energy (cheap, reliable, scalable).
(2) Secure materials (mining, refining, supply chains).
(3) Secure land and connectivity (where projects can physically exist).
(4) Operate under governance (permits, audits, security, responsibility).

This is why, in practice, “AI strategy” quickly turns into energy strategy, infrastructure strategy, and procurement strategy.

So does GDP become meaningless?

GDP measures monetized transactions. If AI internalizes capabilities (a service becomes automated, a workflow becomes cheaper), GDP can undercount some welfare improvements. That part is real.

But governance doesn’t go away. Cities and states still need to measure: fiscal capacity, employment transitions, service accessibility, infrastructure reliability, energy security, and supply-chain risk. The right conclusion is not “GDP is useless,” but: we need a richer dashboard.

A practical takeaway for governments and cities

In Gov-Tech and urban operations, the biggest mistake is to treat AI as a presentation layer—dashboards, shiny demos, “city brain” language. The useful work is boring and structural: building closed loops that convert signals into action, then action into learning.

If you can run discover → triage → dispatch → resolve → review → improve faster and more transparently, you get a real productivity gain. If you only generate more reports, you get noise.

Closing

AI may make many forms of knowledge work cheap. It won’t make the physical world free. Scarcity will migrate—toward energy, land, materials, infrastructure, and the organizations capable of delivering at scale.

The future is not “infinite output equals money doesn’t matter.” The future is harsher and simpler: those who control reliable production systems will matter more.

Internal links

Tags: #AI #Energy #Materials #Land #Infrastructure #GovTech

Mr. Qizhi