Is the US Stock Selloff Over? Wall Street Debates AI Bubble as S&P 500, Chip Stocks Diverge for 2027

TradingKey - Following the U.S. midterm elections, the artificial intelligence trade abruptly suffered massive capital outflows. On November 5, the S&P 500 tumbled 4.4%, the Nasdaq Composite Index fell 5.5%, and the Philadelphia Semiconductor Index plunged about 12%, with Nvidia (NVDA), AMD (AMD), Micron Technology (MU), and SanDisk (SNDK), among other AI chip stocks, bearing the brunt of the selloff. The Nasdaq thereby entered a technical correction, while debates surrounding data center regulation, financing costs, and returns on AI capital expenditures also quickly heated up.

In the face of this round of U.S. stock selloffs, Wall Street has not reached a consensus. Morgan Stanley, Yardeni Research, and Cathie Wood believe the market is primarily experiencing an unwinding of crowded positions and that the long-term bull market is not over; Bank of America and JPMorgan, however, judge that chip stocks may still need to wait for downward revisions to earnings expectations before truly bottoming out. More pessimistically, Capital Economics and Bank of America strategist Michael Hartnett warn that slowing AI capital expenditures and deteriorating free cash flow could cause U.S. stock performance in 2027 to be significantly weaker than market expectations.

Bulls See US Stocks Just Correcting, AI Trade Has Not Lost Fundamental Support

Cathie Wood is not pessimistic about recent market volatility, arguing that ongoing investor concerns over AI spending, debt, and free cash flow actually indicate the market has not yet entered a full-blown mania akin to the final stages of the 2000 dot-com bubble. During that tech bubble, investors virtually stopped discussing risk; today, by contrast, the returns on AI investments are constantly being questioned, and this "wall of worry" could allow the bull market to persist longer.

Wood believes that robotaxis, AI agents, and automation platforms remain in the early stages of commercialization, characterizing the current market as merely taking "two steps forward, one step back." She expects the U.S. equity rally to continue into 2027, though she concedes that rising interest rates and corporate financing costs could pose major headwinds. In a volatile market, she favors taking modest profits during rebounds and using pullbacks to add to high-conviction assets.

Morgan Stanley Chief U.S. Equity Strategist Mike Wilson similarly views this drop as a correction rather than the end of the bull market. He believes investors can gradually increase risk exposure over the next two to three weeks and maintains his year-end target of 8,000 for the S&P 500. Ed Yardeni of Yardeni Research expects the index to climb to 7,900 by year-end, while Jonathan Golub of Seaport Global has set an 8,200 target. The average year-end target among Wall Street strategists stands at roughly 7,973, indicating that most institutions have not abandoned their bullish stance over a single-day slump.

Goldman Sachs strategist Rich Privorotsky believes AI stocks have been sufficiently sold off, noting that cleared-out positioning and compressed valuations could set the stage for a subsequent rebound. While a Democratic victory heightens permitting risks for data centers, it also removes the uncertainty of the election itself; if companies fear future regulatory tightening, they may instead pull forward planned investments. In addition, constraints on new computing capacity could increase the scarcity value of existing infrastructure, relatively benefiting companies that already possess vast computing resources, such as Amazon, Microsoft, and Alphabet.

Some institutions also consider this downturn to be more akin to capital-driven deleveraging than a sudden evaporation of AI demand. JPMorgan noted that the recent AI stock selloff coincided with concentrated profit-taking by hedge funds, with gross exposure among equity long/short funds dropping about 8% from recent highs—the steepest contraction since 2016. Lombard Odier likewise attributed the correction to the unwinding of crowded trades, maintaining that enterprise AI adoption and cloud computing demand remain robust.

Bears Warn AI Spending Unsustainable, S&P 500 May Fall Another 12% by Year-End 2027

Compared with Wall Street bulls, Capital Economics takes a distinctly cautious view on AI investment returns. Vicky Redwood, the firm's chief economic adviser, projects that capital expenditures by a handful of hyperscale cloud computing companies could reach $800 billion in 2026—roughly double the prior year's level—and exceed $1 trillion in 2027.

Redwood estimates that to justify such massive infrastructure spending, tech companies' AI-related revenue from external customers would need to increase roughly tenfold by 2030. While not entirely impossible, this implies that AI must replace existing businesses, business models, and labor on a large scale within a short timeframe—at a pace far exceeding what is currently observed.

Therefore, Capital Economics expects AI investment to continue growing in 2027, but at a significantly slower pace, before potentially declining by about 10% in 2028. Equity markets typically price in shifts in the capex cycle ahead of time; Redwood expects that by late 2027, the S&P 500 Index could be 12% below current levels and down roughly 17% from its October 2026 peak, with tech-heavy indices potentially facing steeper declines.

Bank of America strategist Michael Hartnett, meanwhile, focuses on the concentration and financing models of the AI trade. His report argues that the midterm elections could mark a turning point where AI stock concentration peaks. The market cap share of the top 10 US AI-related companies once reached 41%, matching the tech stock concentration seen during the 2000 dot-com bubble and sitting just below the historical peak of about 44% in the Japanese stock market in 1989.

Hartnett projects that capex by hyperscale cloud companies could surge from $239 billion in 2024 to $1.3 trillion in 2027, while their free cash flow could swing from $239 billion in 2024 to negative $252 billion in 2027. If this forecast materializes, AI investment will become increasingly reliant on bond and equity financing rather than being primarily funded by operating cash flow, and the importance of interest rates and credit spreads will rise accordingly.

JPMorgan similarly cautions that tech companies' actual financing pressures may not be fully reflected on traditional balance sheets. The firm estimates that only about 15% of large AI companies' debt and lease commitments are recorded directly on their balance sheets. As credit spreads widen, investors will need to look beyond revenue growth and scrutinize long-term leases, project financing, and other off-balance-sheet obligations.

This means the greatest risk facing the AI trade is not an immediate collapse in demand to zero, but rather that projected revenues fail to ramp up quickly enough to cover depreciation, interest, and construction costs. Once capital markets demand higher returns, some data center projects may be delayed, with chips, memory, power equipment, and emerging cloud computing companies among the first to feel the impact.

Semiconductor Stocks May Bottom in Early 2027, but Earnings Cuts Remain a Necessary Condition

As of November 5, the Philadelphia Semiconductor Index had fallen approximately 38% from its June 22 peak. By comparison, the four major semiconductor downcycles in 2011, 2015, 2018, and 2022 saw an average decline of about 32% and lasted an average of roughly 253 days.

While the magnitude of this correction has already exceeded the historical average, its duration has been only around 136 days. This suggests that while price risk may have largely been digested, the adjustment in terms of time may still be insufficient. Bank of America believes that this indiscriminate selloff could gradually conclude in the first quarter of 2027, but the market will first need to see actual cuts to earnings forecasts to establish a more reliable bottom.

Valuations have contracted significantly, with the SOX index currently trading at a forward P/E of around 13.7x, below the S&P 500's 17.7x and representing an approximate 22% discount; at the peak of the AI rally, semiconductor stocks traded at a roughly 15% premium over the broader market. Historically, the forward P/E for chip stocks at cyclical troughs has been around 12x, meaning that if earnings forecasts remain stable, valuations still face about 10% to 12% of downside toward historical lows.

Greater uncertainty stems from earnings. U.S. cloud capex is currently running at an annualized rate of around $1.2 trillion, compared with market expectations exceeding $1.5 trillion for 2027. If spending merely remains at current levels, actual capex could fall about 20% short of consensus expectations. Bank of America estimates that under such a scenario, AI semiconductor revenue could face downside risk exceeding 20%, with earnings per share potentially dropping by 35% to 40%.

However, capex cuts may not affect all companies equally. Cloud hyperscalers are more likely to defer land, power, and facility construction first, while next-generation chips delivering higher performance and energy efficiency, such as Vera Rubin, may still be prioritized. JPMorgan notes that long-term supply agreements and customer prepayments could support the earnings of certain memory chipmakers over the next six months or so, indicating that stock prices are front-running a downcycle that has yet to show up in financial reports.

Synthesizing Wall Street views, this wave of U.S. equity selling is neither a routine fluctuation that can be ignored, nor sufficient evidence that the AI bull market is definitively over. In the short term, positioning unwinding, deeply oversold conditions, and corporate buybacks could drive a rebound in the S&P 500; over the medium term, semiconductor and AI infrastructure stocks still need to digest headwinds from earnings downgrades, data center regulations, and rising financing costs.

The critical dividing line for 2027 will be whether AI capex can pivot from 'scaling up' to 'generating returns.' If cloud computing revenue, AI commercialization, and free cash flow keep pace with the rate of investment, this correction could present a buying opportunity within a long-term bull market; if spending continues to rely on debt while earnings and cash flows fail to materialize, even if the S&P 500 remains relatively stable, Nvidia, Micron, SanDisk, and other high-beta AI stocks could face a more prolonged period of valuation resets.