The Economics of Abundance

Why competition and optimisation change where value lives.

3rd August 2026

The Assumption

For most of the AI race, investors believed intelligence itself would remain scarce.

Build the smartest model.

Acquire the most GPUs.

Train on the most data.

Raise the most capital.

The company that produced the best intelligence would capture extraordinary economics.

Scarcity would protect margins.

Margins would justify infrastructure.

Infrastructure would justify capital.

It appeared to be a familiar technology story.

Produce something scarce.

Charge accordingly.

Recent developments suggest something very different may be emerging.

Intelligence itself may be becoming increasingly difficult to monopolise.

The Recursive Loop

The idea behind the technological singularity is well known.

Humans build an artificial intelligence.

That intelligence becomes capable of improving AI itself.

It designs a better model.

That better model designs an even better one.

Each generation accelerates the next.

Capability compounds.

But there is another consequence that receives far less attention.

Every iteration also discovers better ways of producing intelligence.

Better architectures.

Better routing.

Better post-training.

Better memory.

Better priors.

Better distillation.

Better compression.

Different mechanisms.

One economic outcome.

More intelligence from less computation.

Recursive self-improvement does not simply increase intelligence.

It increases intelligence per unit of capital.

The Optimisation Principle

Every intelligent system faces the same pressure.

Produce better decisions.

Using fewer resources.

Everything we describe as progress in artificial intelligence follows this pattern.

Better priors reduce search.

Memory reduces recomputation.

Distillation compresses capability.

Routing eliminates unnecessary inference.

Recursive self-improvement discovers more efficient algorithms.

Different techniques.

One objective.

Increase the amount of intelligence produced for every unit of computation.

As every competitor optimises towards the same destination, the cost of producing intelligence continues to fall.

Abundance is not an accident.

It is the natural consequence of optimisation.

Intelligence Becomes Cheaper

Each improvement does more than increase capability.

It reduces the amount of work required to achieve that capability again.

Yesterday’s frontier becomes today’s baseline.

Problems that once required enormous computation become routine.

The marginal cost of intelligence begins to fall.

Not because intelligence becomes less valuable.

Because it becomes easier to produce.

This is how abundance emerges.

Capital Still Thinks In Scarcity

Financial markets have spent decades rewarding scarcity.

Scarce oil.

Scarce chips.

Scarce bandwidth.

Scarce compute.

Scarce intelligence.

When scarcity persists, producers capture extraordinary economics.

Capital naturally flows towards expanding production.

That logic has underpinned hundreds of billions of dollars of AI infrastructure investment.

But recursive improvement changes the equation.

If each generation requires less computation than the last to achieve similar or better capability, the economics begin to invert.

Demand for intelligence may continue growing rapidly.

The amount of infrastructure required to produce each unit of intelligence may not.

Those are very different things.

Competition Accelerates Abundance

Technology alone does not create abundance.

Competition does.

Every breakthrough rapidly becomes the starting point for the next competitor.

Open-weight models spread capability.

Distillation compresses frontier performance into smaller systems.

New architectures reduce computation.

Recursive self-improvement accelerates discovery.

Every improvement becomes tomorrow’s baseline.

As more organisations produce increasingly capable intelligence, pricing power weakens.

Markets compete away scarcity.

The marginal cost of intelligence trends towards its cost of production.

Not because intelligence becomes less useful.

Because it becomes increasingly difficult for any single organisation to monopolise it.

Abundance is not simply a technical outcome.

It is an economic one.

The Token Was Never The Product

As the marginal cost of intelligence approaches zero, another assumption begins to fail.

That intelligence itself is the product.

For years, the AI industry has largely monetised tokens.

Inference.

API calls.

Compute.

But if intelligence becomes abundant, charging for the production of another token becomes increasingly difficult to defend.

The economic unit begins to shift.

Not from one model to another.

From intelligence itself…

To the outcomes that intelligence creates.

Abundance Changes Where Value Lives

History follows a familiar pattern.

When production becomes abundant…

Value migrates elsewhere.

Electricity became abundant.

The grid became valuable.

Compute became abundant.

Cloud platforms became valuable.

Artificial intelligence appears to be approaching the same transition.

As intelligence itself becomes abundant, producing another capable model becomes progressively less differentiated.

The scarcity moves.

Not to intelligence.

To trusted resolution.

The New Bottleneck

When intelligence is inexpensive, the problem is no longer generating another answer.

It is determining which answer should be trusted.

Which supplier.

Which diagnosis.

Which investment.

Which recommendation.

Which action.

As search disappears, trusted defaults replace exploration.

Every reliable decision removes unnecessary computation.

Every trusted recommendation reduces uncertainty.

Every successful resolution becomes reusable.

Eventually, the most valuable systems are no longer those that generate the most intelligence.

They are the ones that consistently produce the right outcome.

The Resolution Economy

Competition ensures intelligence keeps improving.

Optimisation ensures intelligence keeps getting cheaper.

Abundance ensures value migrates elsewhere.

As intelligence becomes easier to produce, producing intelligence becomes less differentiated.

Competitive advantage migrates.

From computation…

To confidence.

From intelligence…

To trusted outcomes.

From producing answers…

To consistently delivering the right ones.

The first phase of AI rewarded producing intelligence.

The next phase is likely to reward organising it into trusted outcomes.

Markets have spent decades learning how to value scarcity.

Artificial intelligence may now require them to learn how to value abundance.