Chris Wood warns AI capex binge may burn billions as markets turn against Big Tech spending

The artificial intelligence (AI) boom is entering a more unforgiving phase as investors begin questioning whether Big Tech’s unprecedented spending will generate adequate returns or merely consume billions of dollars in cash. Jefferies’ Head of Global Equity Strategy Chris Wood said markets are now responding negatively to increases in capital expenditure, a warning signal for hyperscalers that have committed vast sums to AI infrastructure. While announced results have yet to indicate an outright decline in spending, deteriorating free cash flow and sharp share price reactions suggest investors are no longer prepared to reward capex at any cost.

Wood’s long-standing view is that the “hyperscalers will end up blowing a lot of money on their capex binge” and that AI could resemble the airline industry more than the winner-takes-all economics of the internet era.

The warning follows sharp investor reactions to earnings and spending plans from some of the world’s biggest technology companies.

Alphabet was punished after turning free cash flow negative in the second quarter of 2026 for the first time since its IPO in 2004, according to Wood’s GREED & fear report.

Meta shares fell as its free cash flow plunged 91% to $784 million in the second quarter, from $8.5 billion in the same period last year. The company also raised the lower end of its 2026 capex guidance, taking the range to $130-$145 billion from $125-$145 billion.

Microsoft provided the contrast. Its shares gained 8% after it maintained calendar year 2026 capex guidance at approximately $175 billion. That figure was adjusted from an earlier $190 billion estimate because of accounting changes related to the useful life of assets and the movement of finance leases to operating leases, which are not included in capex.

Also Read | Chris Wood’s big warning: The specific risk that will finally trigger the end of AI tradeThe divergent market reactions suggest investors are becoming more selective about AI spending. Companies may still be able to commit billions of dollars to infrastructure, but the market increasingly wants evidence that this spending can support revenue and cash-flow growth.

Wood said results announced so far have not signalled a decline in hyperscaler capex, which is why analysts have yet to cut earnings forecasts for companies such as memory chip producers. But the negative response to higher spending represents an important shift in market behaviour.

The continuing unwind in semiconductor stocks has already pushed some companies close to their 200-day moving averages. Wood said the correction could be limited if it merely represents a technical flushing out of leveraged positions accumulated by momentum traders. The bigger risk is that the violent selloff is anticipating an eventual slowdown in hyperscaler spending.

Korea’s AI trade suffers a brutal reversal

The scale of the speculative unwind is particularly evident in South Korea, one of the biggest beneficiaries of the global semiconductor rally.

The Kospi has fallen 40% from its all-time high of 9,385.6 reached on June 19. Foreign investors have sold a net $116 billion of Korean equities so far this year across the cash and futures markets, with technology stocks accounting for $104 billion of that selling.

Assets in domestic leveraged exchange-traded funds tracking Korean equities have collapsed to $17 billion, down 66% from their $50 billion peak on June 22. However, retail margin-loan balances remain elevated at $22.8 billion, only $2.4 billion below their peak in late May.

Dedicated domestic ETFs tracking Korean equities have received net creations of $48 billion this year, providing some counterweight to the foreign exodus.

Korea’s neutral weighting in the MSCI AC Asia Pacific ex-Japan Index has meanwhile fallen to 17.5% from a peak of 24.6% in late June. The sharp decline highlights how rapidly index exposure and foreign positioning can reverse when investors begin questioning the assumptions underpinning a crowded trade.

Wood sees China emerging as the AI winner

Wood continues to believe that China is best positioned to prevail in AI, particularly in the mass consumer market. His thesis does not assume that demand for computing power will collapse. Instead, he expects demand to keep expanding even if the customers and eventual winners change.

That distinction is central to his outlook that AI may continue transforming the economy while still delivering disappointing financial returns for companies funding the infrastructure buildout.

China’s rapidly growing semiconductor industry also provides a striking counterpoint to the selloff elsewhere. CXMT, the country’s leading DRAM manufacturer, surged 500% after listing. Its market capitalisation reached $523 billion, briefly making it the most valuable company listed in mainland China.

CXMT’s valuation exceeded Industrial and Commercial Bank of China’s $410 billion market capitalisation and was just below Hong Kong-listed Tencent’s $547 billion.

The extraordinary debut came during a week in which global memory stocks remained under intense pressure, demonstrating that investor appetite for AI has not disappeared. Instead, capital may be rotating towards companies offering lower starting valuations or greater exposure to China’s domestic technology ecosystem.

The critical question is no longer whether AI demand will grow, but who will capture the economics of that growth. Wood’s warning is that the companies spending the most may not necessarily emerge as the biggest winners and the market has started demanding proof before financing the next phase of the capex boom.