For much of 2026, betting on the AI boom has been one of the easiest trades for investors to understand.
Buy the chipmakers. Buy the biggest AI beneficiaries. Buy the companies spending heavily on AI infrastructure.
That trade is still alive, but the mood around it is changing.
Bank of America’s latest fund manager survey shows that long global semiconductors remains the most crowded trade in markets , with 39% of investors naming it as their top crowded position.
That sounds bullish at first.
But there is an important detail: the number has fallen sharply from 82% in July.
Investors are not abandoning the AI trade altogether. They are becoming much more selective about how much they are willing to pay for it.
THE AI TRADE IS LOSING SOME OF ITS CROWD
Global semiconductors remain far ahead of the next most crowded trades.
- 39% of investors said long global semiconductors was the most crowded trade
- 14% pointed to short China stocks
- 12% pointed to long Magnificent Seven stocks
But compare today’s 39% with the 82% recorded in July , and the change in sentiment becomes much clearer.
The semiconductor trade is still the market’s favourite crowded position. It is just not generating the same level of conviction it did a few months ago.
And that matters because crowded trades can become vulnerable when investors start questioning the assumptions behind them.
THE BIG QUESTION: WHEN DOES AI SPENDING PAY OFF?
This is becoming the central debate for AI stocks.
The world’s biggest technology companies are spending enormous amounts on data centres, chips, networking equipment and other AI infrastructure.
For investors, the question is no longer simply:
“Will AI grow?”
Most already believe it will.
The tougher question is:
“Will the returns from all this spending arrive quickly enough to justify today’s valuations?”
The latest BofA survey suggests investors are becoming less comfortable with that equation.
52% of fund managers now expect at least one AI hyperscaler to announce a capex cut over the next six months.
That is up from 30% in October .
At the same time, a record 38% of respondents believe companies are overinvesting, while only 16% want companies to increase capital expenditure.
That is a significant shift in the conversation.
“PAUSE AI CAPEX” IS THE MESSAGE INVESTORS ARE WATCHING
BofA strategists described the survey’s biggest message as:
“Pause AI capex.”
That does not necessarily mean investors expect the AI boom to collapse.
It means the market is starting to ask companies to prove that their spending is translating into revenue, profits and ultimately shareholder returns.
This distinction is important.
AI infrastructure spending can continue to grow while investors simultaneously become more cautious about the stocks benefiting from that spending.
In other words, a strong AI industry does not automatically mean every AI-related stock will keep rising.
THE AI BUBBLE QUESTION IS BACK
Investors are also becoming more concerned about what could go wrong.
When BofA asked fund managers about the biggest tail risk to markets:
- 34% said an AI bubble
- 31% said a hyperscaler credit event
That second concern is particularly interesting.
The AI infrastructure buildout requires enormous amounts of capital. If companies increasingly rely on debt to finance that expansion, higher borrowing costs or weaker returns could create pressure further down the chain.
AI hyperscaler bonds were identified as the most likely source of a credit event by 38% of respondents.
Private credit followed at 30% .
So the concern is moving beyond stock valuations.
Investors are beginning to think about how the AI boom is being financed, and what happens if expected returns take longer to arrive.
HIGHER YIELDS ARE ADDING ANOTHER PROBLEM
The AI investment boom is happening against a less friendly interest-rate backdrop.
Treasury yields have been rising, while the Federal Reserve’s hawkish commentary has kept pressure on borrowing costs.
That matters for companies building expensive AI infrastructure.
When financing becomes more expensive, projects that looked attractive at lower interest rates can become harder to justify.
It also gives investors another place to put their money.
The BofA survey found that investors have been reducing their underweight in bonds, while respondents said a 10-year Treasury yield of 5.8% would be enough to encourage a shift from stocks into bonds.
That creates a higher hurdle for expensive growth stocks.
NVIDIA IS AT THE CENTRE OF THE DEBATE
No company is more closely tied to the semiconductor and AI infrastructure story than NVIDIA.
That makes Nvidia particularly important as investors reassess the AI trade.
The company is facing several sources of uncertainty at the same time.
Leadership uncertainty following Jensen Huang’s reported departure has added another layer of concern around the company and the wider semiconductor sector.
There are also new export restrictions affecting advanced AI chips , creating additional uncertainty around future demand and market access.
And then there is the biggest question of all:
How much longer can AI-related spending keep accelerating at the current pace?
Nvidia’s upcoming earnings will therefore be watched for more than just revenue and profit numbers.
Investors will be looking closely at what the company says about AI demand, hyperscaler spending and the sustainability of the current infrastructure cycle.
TECH IS STILL POPULAR, JUST NOT AS POPULAR
One of the most interesting findings from the survey is that investors have reduced their technology overweight by the most since 2006.
That does not mean everyone has suddenly turned bearish on technology.
Instead, it suggests investors are becoming less willing to own the sector at any price.
That is an important difference.
The AI story can remain intact while the AI trade becomes less attractive from a valuation perspective.
This is where the next phase of the market could get interesting.
THE BROADER ECONOMIC PICTURE IS NOT AS DARK
There is also some good news hidden inside the survey.
50% of investors now expect a soft landing , compared with 42% who expect no landing.
Only 3% expect a hard landing.
So investors are not necessarily preparing for a major economic collapse.
The concern is more specific:
Can AI companies continue spending at extraordinary levels if economic growth remains steady but the returns on that spending take longer than expected?
That is a much more nuanced risk than simply “AI is over.”
WHERE ARE INVESTORS MOVING?
The survey also offers some clues about positioning.
Average cash levels fell to 3.6% from 3.9% , showing that investors are not simply sitting on the sidelines.
At the same time, a record 23% net of investors said gold is undervalued.
Investors also reduced their underweight in bonds by 13 percentage points, although they remain net underweight.
So the picture is not one of outright panic.
It looks more like portfolio repositioning.
Investors are looking for ways to stay invested while reducing exposure to the most crowded and expensive parts of the market.
THE BIG TAKEAWAY
The AI trade isn’t dead.
In fact, semiconductors remain the single most crowded trade among global fund managers.
But the enthusiasm has clearly cooled.
The market is moving from:
“How much can AI spending grow?”
to:
“When will all this spending generate enough returns to justify it?”
That shift could be extremely important for AI stocks.
The next leg of the AI trade may not be driven simply by bigger capex numbers. Investors may start rewarding companies that can prove the return on that spending , while becoming less forgiving toward companies whose valuations depend mainly on future AI growth.
And that is why the next round of earnings from the biggest AI companies could matter so much.
The AI boom may still have plenty of fuel. The question is whether investors are willing to keep paying the same price for it.