When Intelligence Stops Being Scarce

Economics has an irritating habit: the moment we remove one constraint, it immediately reveals the next.

Imagine intelligence becoming infrastructure. Good analysis can be run in thousands of copies. Design options are produced faster than a person can review them. Advice at the level of a strong specialist costs little more than the electricity required for a few minutes of computing.

That sounds like abundance.

So it is time to look for the scarcity.

The bottleneck always moves

When computing was expensive, we economised on computing. When networks were slow, we conserved bandwidth. When a highly qualified engineer was difficult to find, processes were designed around that person’s diary.

If high-quality thinking becomes dramatically cheaper, that diary stops being the main constraint.

You can call not one adviser but a thousand. Let them argue, model scenarios, criticise one another and produce dozens of carefully developed options overnight.

After the brief excitement, it turns out that there is still only one factory. There is still one plot of land. The budget is finite. A day still has twenty-four hours. And, for some reason, the customer is not willing to buy twenty product variants at once.

An abundance of intelligence does not abolish the limits of the world. It simply makes them much more visible.

Physical reality becomes expensive again

A million virtual engineers do not produce a million turbines without metal, energy, machine tools and logistics. A million architectural options do not create a second city centre in which to build another house.

As intellectual work becomes cheaper, value may shift towards energy, materials, equipment, infrastructure, access rights and physical space.

There is something faintly comic about the ending to decades of talk about the knowledge economy: we finally obtain something close to unlimited intelligence and promptly run into copper, concrete and electricity again.

From a systems point of view, however, this is entirely normal. Removing one bottleneck does not make the system unlimited. It exposes the next one.

Data may become more valuable than the algorithm

If strong models are available to many organisations, access to the model itself stops distinguishing one company from another.

What matters more is what everyone else does not have: proprietary data, accumulated context, a history of decisions, feedback from real-world operation, access to unusual assets and measurements.

You can have the perfect analyst, but without good data it will analyse uncertainty with impeccable precision.

Perhaps companies of the future will talk less proudly about which model they use and more about the quality of the world that model is able to see through their data.

Attention becomes a luxury

There is another resource AI does not automatically increase: human attention.

If a system can produce a thousand good reports, who reads them? If twenty plausible strategies can be generated every day, who chooses one? If every employee can write a perfect email in a minute, how many perfect emails is a colleague expected to survive?

We already know this effect from information. When data is hard to obtain, access is valuable. When data is effectively endless, the filter becomes valuable.

The same may happen with intelligence. A shortage of good ideas may be replaced by a surplus of reasonable suggestions.

At that point the expensive person, system or rule will be the one able to say: you do not need to look at these nine hundred and ninety options.

When good answers are effectively unlimited, a good filter becomes more valuable than one more answer.

The rarest resource may be a good question

When an answer is expensive, we admire the person who knows how to obtain it. When the answer is almost free, the formulation of the task suddenly becomes valuable.

Which problem is worth solving in the first place? What should count as a good outcome? Where will change produce a real effect, and where are we merely improving something because we now can?

AI may optimise a given objective indefinitely. But choosing the objective is not simply a computational problem. It contains interests, risk, taste, politics and a view of the future.

Not because writing a sentence is difficult, but because there may be too many possibilities for us to pursue all of them calmly.

A world of cheap intelligence may turn out to be a world of very expensive goals.

Abundance does not abolish economics

We often imagine extremely capable AI as a “solve the problem” button. But most problems are not caused solely by the absence of a good solution.

There are resources, deadlines, competing objectives, ownership rights, physical constraints and people who sincerely want different outcomes.

So making intelligence abundant will not create an economy without scarcity. It will probably change the price list.

Intelligence itself may become cheap. Energy, real-world experience, access, trust, attention and the ability to choose a goal may become more expensive.

And perhaps one day the prestigious claim will no longer be “our system is very intelligent”, but something rarer: “we know exactly why it is doing all this”.

Sources and reference points

The factual reference point below is current as of September 2026. The future scenarios in the article are the author’s assumptions, not a forecast made by the organisation cited.

1. IMF, Global Economic and Financial Implications of Artificial Intelligence, 2026. Source (https://www.imf.org/-/media/files/publications/imf-notes/2026/english/insea2026002.pdf)