Investor Steve Eisman sold Alphabet and boosted cash, warning that crowded AI exposure could magnify a market correction.
Artificial intelligence has propelled equities higher, filled corporate spending plans, and changed the bond market.
Investor Steve Eisman now concerns himself that the two seemingly independent trends have turned into one trade.
Eisman told CNBC he decreased his AI risk by selling his longtime stake in Alphabet (GOOGL). He has not bought defensive stocks to replace his Alphabet stake and is keeping cash while he considers what to do next.
He isn’t betting against the market. He frets about what happens if companies can’t get an adequate return on the hundreds of billions of dollars being poured into AI infrastructure.
“I think we have a big correction,” Eisman said when asked what would happen if AI failed to deliver.
He did not estimate how large the correction could be or when it might occur.
Instead, Eisman claimed investors may be more concentrated than they perceive. A traditional portfolio of stocks and bonds could nevertheless be closely linked to the AI boom, with big tech companies dominating major indexes and AI infrastructure fueling significant corporate borrowing.
Official index data backs up the wider concentration problem. The 10 largest businesses in the S&P 500 make up around 36.4% of the index, according to S&P Dow Jones Indices.
The Federal Reserveindicated in its May financial stability report that they remained near the upper end of their historical range. The Fed’s survey of market participants identified AI-related issues as one of the biggest potential threats to financial stability.
Steve Eisman sold Alphabet as AI spending accelerated
Eisman said he had possessed Alphabet so long he couldn’t recall when he first got it.
His sale wasn’t necessarily a gloomy call on Google’s operating company. It was a method to decrease exposure to a market increasingly driven by AI expenditure and expectations.
The tension shows in Alphabet.
The tech giant reported revenue of $119.8 billion for the second quarter, up 24%, and operating income of $40.8 billion, up 30%. Google Cloud sales soared 82% to $24.8 billion, suggesting the AI infrastructure is already fueling significant commercial development.
The spending needed to achieve that increase is also becoming increasingly difficult to ignore.
Alphabet spent $44.9 billion on capital expenditures in the quarter, mostly on servers, data centers, and networking equipment. That was above its operating cash flow of $39.1 billion and left quarterly free cash flow at negative $5.9 billion.
Management boosted its 2026 capital-expenditure guidance to a range of $195 billion to $205 billion and predicted that spending would rise significantly again in 2027. Higher depreciation, electricity, and data center running costs are all likely to weigh on future profitability.
These data points assist in explaining Eisman’s reluctance.
Alphabet is seeing significant sales and profit growth, but investors will have to weigh if those returns can continuously beat the cost of building up the infrastructure.
Related: Alphabet’s biggest AI fear may be fading
Eisman isn’t saying AI technology will be useless.
He suggested it may be an outstanding technology without every firm or investment around it being successful. That difference is significant because stock prices reflect earnings, not just technological progress.
He also questioned ever more complex agreements whereby AI developers, cloud providers, and chipmakers fund or assist each other’s infrastructure buys.
Eisman said Nvidia (NVDA) will probably sell its chips at a profit. His fear is that there is not enough end-customer profitability in the larger system to justify all the investment going into those sales.
The AI trade reaches beyond Alphabet stock
Alphabet isn’t the only firm throwing its weight around.
In the third quarter, Microsoft (MSFT) spent $31.9 billion in capital expenditures, with around two-thirds of that going toward shorter-lived assets like graphics processing units and central processing units. Infrastructure spending drove free cash flow of $15.8 billion.
Microsoft also stated sustained investment in AI and greater product usage eroded its companywide gross margin. The cloud capacity continued to lag behind customer demand, a sign that the spending is related to real need and not speculative building.
What investors should watch
- Whether AI revenue continues growing faster than infrastructure costs.
- Whether free cash flow recovers after the current construction surge.
- Whether cloud demand remains greater than available computing capacity.
- Whether customers become less willing to pay for expensive closed AI models.
- Whether market gains broaden beyond the largest technology companies.
That is the central divide for investors.
The more bullish view is that demand for AI computing is so strong that Alphabet, Microsoft, Nvidia, and other companies can expand into their infrastructure expenditures.
More AI:
- Workers just sent AI companies an ultimatum
- Palantir CEO has a blunt verdict on OpenAI and Anthropic
- Elon Musk pulls no punches with AI rivals as Grok 4.5 debuts
The negative view is that spending is outstripping profitable consumer demand, leaving enterprises with excess capacity, more depreciation, and poor returns.
Eisman sees cash as an option, not a permanent strategy
Eisman has not rotated into consumer staples or any other typically defensive sector.
What investors seem to desire right now, he added, is either AI stocks or no theme exposure at all. Just because slow-growing defensive companies look safer doesn’t mean that buying them presents appealing upside.
For now, he’s sitting in cash and waiting for more clarification.
The posture implies prudence, not a blatant bet against the market. Eisman acknowledged that the question of AI profitability will take time to settle.
So his caution is less about declaring a top now and more about knowing your portfolio’s risk.
Investors who own broad market funds may think they’re diversified across hundreds of companies. But the biggest stocks and a huge chunk of the index’s success are still tied to AI infrastructure, cloud computing, and demand for semiconductors.
Bonds may not offer as much separation as expected when technology companies are borrowing to finance the same expansion, Eisman argued. The interview did not provide data supporting his specific characterization of AI-related bond issuance, so that claim should be treated as his assessment.
The stock market result may eventually be determined by a relatively simple test.
AI does not need to dazzle users. It has to generate enough sustainable income, profit, and cash flow to fund one of the biggest investment cycles in business history.
That, Eisman figures, might let the market keep chugging higher.
The danger is that too many investors have made the same assumption without realizing how much of their portfolio depends on it.
