For years, Wall Street had one dominant AI assumption: bigger models would require bigger data centers, which would require more chips, more memory, more networking equipment and more capital.

It was a remarkably powerful investment narrative.

Then the people building the models started warning that the industry may need to slow down.

That was enough to send a shock through global technology markets.

AI-linked stocks fell sharply on Monday after Anthropic CEO Dario Amodei called for a slower pace of AI development and was backed by OpenAI CEO Sam Altman and xAI chief Elon Musk. The selloff spread from AI laboratories to the companies supplying the infrastructure that makes the technology possible.

In Japan, SoftBank — a major investor in OpenAI — fell as much as 13.2%. Kioxia plunged nearly 10%, Tokyo Electron dropped around 4%, while SK Hynix and Samsung also fell. Taiwan Semiconductor Manufacturing Co. declined as well.

The moves were striking because these companies did not suddenly report collapsing earnings.

Instead, investors were reacting to a change in the expected pace of AI progress.

That distinction could become extremely important for markets.

The financial case for the AI boom has never been simply that AI will be useful.

It is that AI capabilities are improving rapidly.

Every generation of models requires enormous amounts of computing power to train and run. As systems become more sophisticated, investors have expected demand for specialized processors, memory, data centers and cloud infrastructure to rise accordingly.

That expectation has created some of the largest investment booms in modern technology.

Nvidia has become the symbol of the semiconductor side of the AI revolution.

But the ecosystem extends far beyond Nvidia.

Memory manufacturers such as SK Hynix and Kioxia are essential because AI systems consume huge amounts of high-performance memory.

Taiwan Semiconductor is critical to manufacturing advanced processors.

Tokyo Electron supplies equipment used by chipmakers.

SoftBank has positioned itself around the broader AI ecosystem through major investments.

Every link in that chain benefits when AI developers need to build more powerful systems.

That is why the latest safety warnings hit the stocks so quickly.

The latest warning from Anthropic is especially unsettling because it is not vague.

Amodei has argued that AI agents could become capable of causing massive online disruption within six to 12 months, potentially inflicting hundreds of billions of dollars in damage.

That is not a forecast that the damage will definitely occur.

It is an argument that the industry may be approaching a point where technological capability is advancing faster than society's ability to manage its consequences.

Amodei's proposal is therefore about pacing.

The goal is not necessarily to stop AI development permanently.

It is to create enough time for safety systems, independent oversight and international coordination to develop alongside the technology.

OpenAI CEO Sam Altman has agreed with the core concern, while Elon Musk has also endorsed calls for slowing the pace of development.

For financial markets, that creates an awkward contradiction.

The safer path may involve slower development.

But the economic value of the AI boom has been built on expectations of faster development.

The AI rally has produced enormous gains because investors have priced in an extraordinary future.

Chip demand will remain strong.

Data-center construction will accelerate.

AI adoption will spread across industries.

Companies will monetize the technology.

Profits will rise.

And the technology will keep improving at a rapid pace.

The more assumptions investors build into an asset's valuation, the more vulnerable it becomes to even a modest change in expectations.

That appears to be what happened Monday.

Saxo Bank's Charu Chanana said the valuations of AI and semiconductor companies assume both strong demand and relentless technological progress. Even the possibility of delays can therefore trigger profit-taking.

This is a crucial distinction.

The selloff does not prove the AI boom is over.

It shows that investors are beginning to price the possibility that the boom could become less explosive.

Artificial intelligence is not disappearing because CEOs are calling for caution.

Businesses are still adopting it.

Consumers are still using it.

Governments are still investing in it.

AI remains one of the most important technological developments of the century.

The bigger question is whether spending can continue at the extraordinary pace investors have come to expect.

This is where the market reaction becomes more complicated.

A slowdown in frontier-model development does not necessarily mean slower AI adoption.

Businesses could still deploy existing models more broadly.

Cloud providers could continue expanding capacity.

Companies could invest in AI applications rather than frontier research.

And efficiency improvements could create new demand even without a constant acceleration in model capability.

But investors still have to answer a different question:

Who will actually earn attractive returns on all the money being spent?

T. Rowe Price portfolio manager Sebastien Mallet highlighted this concern, arguing that while AI will clearly change the world, that does not mean every investment currently being made will generate an attractive return.

That may become one of the defining questions of the next stage of the AI cycle.

The AI investment story has already expanded beyond software.

Data centers require land.

Power grids require upgrades.

Semiconductors require enormous fabrication investments.

Cooling systems, networking equipment and storage capacity must all scale.

Companies are committing billions of dollars to infrastructure that may take years to generate an acceptable return.

That creates financial risk.

If AI adoption exceeds expectations, those investments could look brilliant.

If demand is slower than expected, companies could be left with expensive capacity that generates weaker returns.

Investors are beginning to realize that AI is not just a technology trade.

It is also a capital-allocation trade.

The AI slowdown debate is unfolding at precisely the wrong time for anyone hoping for an easy answer.

America and China are competing for technological leadership.

That means every proposed slowdown must confront the possibility that a rival could continue moving forward.

Trump has explicitly rejected calls to slow AI development, arguing that the United States needs to maintain its lead over China.

Chinese state media have meanwhile attacked Amodei's proposal as a strategy that could restrict China's technological development.

This transforms the debate.

It is no longer simply:

“Should companies build AI more slowly?”

It becomes:

“Can competing countries agree to slow a technology that could determine future economic and military power?”

That is a much harder problem.

During the strongest phase of the AI rally, announcing a large data-center project or massive chip order was enough to excite investors.

The market may now become more skeptical.

Investors could start asking:

How quickly will this infrastructure generate revenue?

What is the return on invested capital?

How long will customers need to keep spending?

What happens if model development slows?

What happens if AI regulations become stricter?

And perhaps most importantly:

Who ultimately captures the economic value?

Those questions do not kill the AI trade.

They mature it.

The most important lesson from Monday's selloff may therefore have little to do with artificial intelligence itself.

It may be about expectations.

AI is likely to remain a transformative technology.

But transformative technology does not automatically mean every company involved will outperform indefinitely.

History is full of technological revolutions in which the technology succeeded while many of the investments surrounding it failed to produce attractive returns.

That is the market's emerging concern.

The AI story may be real.

The spending may be real.

The productivity gains may be real.

But prices still matter.

And once the industry's own leaders start warning that AI development may need to slow, investors are forced to consider a possibility that was almost invisible during the height of the rally:

The technology can continue winning while the stocks take a breather.

That distinction may define the next chapter of the AI boom.