Tech Sell-Off Highlights Rising Scrutiny Of AI Spending

What happened?

US technology stocks led the broader market lower on Friday, as question marks about the AI buildout continue to overshadow strong third-quarter earnings reports. The Nasdaq Composite fell 2.9% on Friday and 2.0% on the week, while the S&P 500 declined 1.8% and 0.6%, respectively. The yield on 10-year US Treasuries fell 1 basis point on Friday and 7bps on the week to 4.93%.

Broader risk appetite had also been hurt by a further rise in Brent crude oil to USD 118/bbl on Friday amid concerns that the US may resume military strikes against Iran.

The pressure has paused on Monday. Brent crude has fallen more than 4% to below USD 110/bbl, as markets focus on extended Iran-Qatar talks. The tech sell-off, meanwhile, has eased in Asia, with the Kospi rebounding nearly 3%, albeit with some single-company performance dispersion evident within the North Asia AI supply chain. Nasdaq futures are up 0.9% at the time of writing.

What do we think?

The latest earnings season is demonstrating that demand for AI infrastructure remains exceptionally strong. However, this past week’s market reaction reflects a shift in the AI debate. Although earnings expectations are broadly being met, investors are now placing greater weight on concerns about overinvestment, financing, and cannibalization.

First, overinvestment. In its 3Q results, Alphabet (GOOGL) maintained its USD 195-205bn 2026 capex plan but reported negative free cash flow for the quarter and signaled a significant further increase in capex for 2027. Its shares fell almost 13% the following day. Meta lifted the lower end of its 2026 capex range to USD 135-145bn and pledged continued aggressive investment beyond this year. Its shares fell nearly 7% the following day as investors focused on the higher capex floor and negative 3Q free cash flow.

However, although the bar is clearly rising, cloud growth—a key metric in assessing the return on AI investment—remains strong, and accelerating cloud revenues and expanding advance orders for compute capacity point to increased monetization and sustained demand for AI infrastructure.

In its 3Q results, Alphabet’s Google Cloud posted a year-over-year revenue jump of 103%, up from 82% in the prior quarter and 34% in the March quarter. Microsoft (MSFT)’s Azure grew 48% for the three months to September, compared with 43% for the June quarter, while Oracle (ORCL)’s growth in 3Q also quickened to 62%, from 47% in 2Q. All three companies offered a positive outlook for further cloud growth, saying that additional capacity could support the strong demand they are seeing in the pipeline. Microsoft also reported an eight-fold increase in large Copilot accounts, while Alphabet said its Gemini reached 1 billion users.

On financing, the funding of neoclouds (specialized computing-infrastructure providers whose business is primarily renting out GPU compute capacity) has been a particular area of investor concern. Neocloud credit has sold off because investors increasingly see a maturity and funding mismatch: Very large upfront, debt-financed investment is supported by cash flows that arrive later and remain sensitive to utilization, execution, customer renewal, and refinancing conditions.

But neoclouds account for only about 8% of our capex estimate, versus 80% for Alphabet, Amazon (AMZN), Microsoft, Meta (META), and SpaceX (SPCX). This concentration of capex among larger, generally better-funded companies should limit the direct impact of financial stress among more leveraged infrastructure providers on overall capex. We therefore think select project delays, repricing, and consolidation are more likely than a broad collapse in AI spending.

On cannibalization risks, Alphabet's latest results provided evidence of some near-term pressure from AI on its existing business model. Search query growth reached a five-year high, but Search revenue growth decelerated, with management indicating that AI Mode currently monetizes at a lower rate. Intel (INTC), whose shares fell 28.6% on Friday, provides a further example of robust underlying AI demand colliding with monetization and execution constraints while creating headwinds for traditional business lines: AI is boosting server orders, but it is also contributing to component scarcity, which means Intel has been unable to fully satisfy data center demand. Meanwhile, component scarcity has fed into higher PC prices, which are hurting its client business.

What is our view on this?

Overall, the earnings season has reinforced our confidence in the strength of overall AI sector demand and prospective investment. We have raised our aggregate AI capex forecasts to USD 1.0tr in 2026 and USD 1.4tr in 2027 and introduced a forecast of USD 1.5tr for 2028. These estimates are supported by stronger spending commitments from the largest hyperscalers.

Nonetheless, while we do continue to believe the broader trend remains intact, the bar for performance is rising and higher volatility in the AI trade is likely to be the new normal in this stage of the cycle.

We would be concerned that a more sustained turning point for AI investment could be approaching if we see evidence that AI lab revenue growth is slowing meaningfully, if cloud utilization weakens, or if falling AI costs fail to stimulate sufficient new usage to absorb expanding capacity.

We would also reassess our spending outlook if advances in algorithmic efficiency sharply reduced compute requirements, if increasingly large investments fail to deliver commensurate improvements in AI capabilities, or if a significant deterioration in financial conditions lead to constraints in the hyperscalers’ willingness to fund new investment.

How do we invest?

In our base case, we expect AI investment to remain a positive driver of global equity returns in 2027 and do not think investors should abandon the AI theme in response to short-term volatility. But as the market moves from rewarding spending commitments to evaluating monetization, balance sheet strength, and returns, selectivity and portfolio construction are becoming increasingly important.

First, reduce overconcentration. The earnings season is demonstrating that single-stock risks are elevated, and investors should focus on diversifying more broadly across the AI value chain. Within AI infrastructure, we prefer high-quality companies with reasonable valuations and less risk from excess capacity—particularly in the foundry, compute, and power spaces. Elsewhere, we favor financially strong technology platforms and infrastructure providers with pricing power, reasonable valuations, and lower oversupply risk.

Second, rebalance across sectors and asset classes. Investors who have seen equity and/or technology allocations rise in proportion to their broader portfolio over recent years should consider rebalancing to right-size exposure. Within equities, our preferred sectors include industrials, health care, and financials. Regionally, we prefer the Eurozone, emerging markets, and Asia ex-Japan. We also see attractive return and diversification potential in high-quality fixed income. This past week, as stocks have come under pressure, quality bonds have rallied, showing that quality fixed income is once again serving as a portfolio stabilizer. Elevated yields can also provide a cushion against potential future bond market volatility.

Third, use volatility deliberately. Investors should prepare for further volatility and decide in advance how to respond. Market swings can provide opportunities to rebalance, manage downside risk, generate income, or build exposure gradually at predefined prices. Structured strategies involve issuer, liquidity, pricing, and capital-loss risks and should be appropriately sized based on investors' risk tolerance.