Why Investors Mispriced Artificial Intelligence
Value Destruction at Galactic Scale
2nd August 2026
The biggest repricing in artificial intelligence may have almost nothing to do with falling demand.
For decades, investors rewarded businesses that could defend scarcity.
Scarcity created pricing power.
Pricing power created durable cash flows.
Durable cash flows created extraordinary valuations.
The formula was simple.
Build something difficult to replicate.
Protect the moat.
Compound the economics.
Artificial intelligence appeared to fit that pattern perfectly.
The smartest models would command the highest prices.
The largest compute clusters would dominate the market.
The biggest balance sheets would eventually own intelligence itself.
That assumption justified one of the largest capital expenditure cycles in modern history.
It also justified some of the largest valuations.
The market wasn’t simply buying technology.
It was buying decades of expected pricing power.
Then the economics changed.
The Assumption
For much of the AI race, intelligence appeared to be the scarce resource.
Every improvement demanded more GPUs.
More data.
More infrastructure.
More capital.
If intelligence remained scarce, pricing power would naturally follow.
Investors built discounted cash flow models around exactly that assumption.
Demand appeared limitless.
Economic rents appeared durable.
The future looked obvious.
Until it wasn’t.
Intelligence Became a Commodity
Over the past year, something remarkable has happened.
Open-weight models have rapidly narrowed the capability gap.
Inference costs have collapsed.
Distillation has accelerated.
Continuous learning has improved.
Smaller models now achieve results that previously required enormous infrastructure.
Every new release has compressed the economic life of the previous one.
The demand for intelligence never weakened.
The scarcity did.
Markets don’t reward abundance with monopoly pricing.
They reward abundance by compressing margins.
The repricing has only just begun.
Investors Mispriced Duration
This is where the real valuation error sits.
Investors didn’t misunderstand demand.
They misunderstood duration.
Future cash flows are only valuable if competitive advantages survive long enough to earn them.
Artificial intelligence is dramatically shortening those competitive cycles.
Advantages investors expected to last ten years increasingly last months.
Every model release reduces the remaining economic life of the previous generation.
The asset hasn’t disappeared.
Its half-life has.
Was This Really Value Destruction?
Markets describe this as hundreds of billions of dollars of value destruction.
I’m not convinced that’s what happened.
Nothing disappeared.
The intelligence still exists.
In many ways, there is more intelligence available today than ever before.
The value simply migrated.
Away from producing intelligence.
Towards organising it.
Towards resolving uncertainty.
Towards becoming the trusted endpoint that intelligent systems rely upon.
History has seen this pattern before.
Electricity became abundant.
The grid became valuable.
Compute became abundant.
Cloud platforms became valuable.
Intelligence is becoming abundant.
Reliable resolution becomes valuable.
The Next Repricing
Markets rarely recognise structural change immediately.
They anchor to the economics of the previous cycle.
That is exactly what appears to be happening now.
Many investors are still valuing artificial intelligence as though intelligence itself will remain scarce.
The evidence increasingly suggests the opposite.
Intelligence is becoming abundant.
The economic rents are beginning to migrate elsewhere.
The next repricing may not come from another breakthrough model.
It may come from investors recognising that the industry has already changed.
When that happens, balance sheets, multiples and capital allocation strategies may all need to be reconsidered.
The Resolution Economy
The market spent the past three years trying to identify who would own intelligence.
The next three years may be spent discovering that nobody does.
Intelligence is becoming infrastructure.
Reliable resolution is becoming the scarce resource.
That distinction changes everything.
Commodities compete on price.
Scarcity competes on trust.
The companies that outperform over the next decade may not be those producing the most intelligence.
They may be the ones that consistently transform abundant intelligence into trusted decisions while performing the least unnecessary computation.
History may eventually describe this as the moment artificial intelligence became a commodity.
I suspect it will be remembered differently.
The moment value quietly left the models…
…and began flowing into trusted resolution.
That transition has already started.
Most investors simply haven’t marked their models to it yet.