The "AI slowdown theory" collides with the 5% "anchor of global asset pricing," and tech risks surge! Wells Fargo redraws the U.S. equity investment landscape, downgrading its S&P 500 target.

U.S. stocks and even global stock markets appear to be undergoing a stress test of whether the AI-boom-driven earnings growth trajectory can withstand valuation compression, with the semiconductor sector bearing the first impact. Ohsung Kwon, chief equity strategist at Wall Street financial giant Wells Fargo, cut his year-end S&P 500 target to 7,700 from 7,950, leaving only about 1.1% upside from Monday's U.S. market close, and downgraded the technology sector to "equal-weight" from his previous "overweight" rating. The Wall Street giant now prefers software, which benefits from the AI application wave, over the currently hottest semiconductors, and raised its rating on the defensive healthcare sector. The strategist is increasingly worried that years of earnings expansion have pushed market expectations near historic highs, while AI capital expenditure, policy restrictions on data center construction by U.S. state governments, and fiscal and monetary policy uncertainty have all been mounting recently. Notably, the Wells Fargo chief equity strategist does not show much concern about 2027 earnings; his key vigilance is that a slowdown in capital expenditure related to AI data center construction could hit 2028 profits. Therefore, this adjustment is closer to a re-examination of the forward valuation of the U.S. equity market. The "AI slowdown theory" that has recently swept global stock markets quickly brought the above concerns into trading pricesglobal AI leaders such as Anthropic and OpenAI unanimously called over the weekend for slowing the pace of frontier AI large-model development. On September 12, Anthropic CEO Dario Amodei called for slowing the pace of improving frontier model capabilities so that safety research and protective measures can keep up, and OpenAI CEO Sam Altman as well as SpaceX and Tesla chief Elon Musk and other industry leaders subsequently expressed support. On the first stock market trading day after Anthropic CEO Amodei made the heavyweight "slowdown theory" call, namely September 14, shares of "AI chip superpower" Nvidia fell about 3.4%, while the global semiconductor bellwetherthe Philadelphia Semiconductor Indexunusually plunged about 6%, highlighting that the market is pricing in risks brought by factors such as the AI slowdown discussion. AI slowdown combined with the 10-year U.S. Treasury yield shock: U.S. stocks and even global stock markets begin to compete over the ability to deliver earnings growth under the AI boom The market thus began to reassess growth expectations for hyperscale training investment, model release pace, and future computing power procurement. On September 14, the Philadelphia Semiconductor Index plunged about 6%; South Korea's KOSPI index, known as the global "AI computing power investment bellwether," fell more than 3% on Monday, then fell another 0.85% on September 15 to close at 6,627.26 points, its fourth consecutive trading day of declines. For the semiconductor sectorthe sector that benefits most from the global trillion-dollar-scale AI computing infrastructure construction boomeven if existing orders remain strong, as long as investors lower expectations for subsequent order growth and the duration of growth, valuations may adjust sharply in advance. Meanwhile, the 10-year U.S. Treasury yield, known as the "anchor of global asset pricing," rose intraday to 5.012% on September 14, the highest intraday level since 2007, before falling back to 4.960%. Thereafter, as of the U.S. market close on September 15, the 10-year U.S. Treasury yield stood firmly above 5%, closing near 5.04%, continuing to rise to its highest level since 2007. The "anchor of global asset pricing" is an important reference for long-term U.S. dollar risk-free rates, affecting corporate financing costs and also the discount rate used to convert future profits into current value. Energy prices and inflation pressures are pushing rates higher, leaving tech stocks facing both rising funding costs and adjustments to forward growth expectations. Another change cited by Wells Fargo chief equity strategist Ohsung Kwon: over the past three months, the number of AI data center development moratorium government orders taking effect in the United States has increased sharply by 175%. From an investment mechanism perspective, more expensive financing and slower project implementation will lengthen the cash flow payback period for some hyperscale AI investments, and will also escalate market concerns about the fulfillment of AI infrastructure-related orders and fears that free cash flow growth trajectories will turn negative. By contrast, the main basis for Wall Street bulls to remain optimistic about the bull market in U.S. stocks and even global stocks is that corporate earnings expansion still has support under the unprecedented AI infrastructure boom and the wave of AI applications penetrating all industries. Goldman Sachs publicly forecasts a year-end S&P 500 target of 8,000, with a core framework of earnings growth driving index gains and valuation multiples roughly flat; Yardeni Research president and chief investment strategist Ed Yardeni still retained an 8,400 target on September 12 while raising vigilance about bearish scenarios; Michael Purves, CEO and founder of Tallbacken Capital Advisors, raised his year-end target to 8,500 from 7,400, arguing that earnings growth is strong, sustained, and broad-based, and that moderate P/E expansion could add further upside. The three have different valuation assumptions, but they share a common focus on whether corporate profits can continue to be delivered. According to the relationship "index level = earnings per share price-to-earnings ratio," as long as earnings growth is sufficient to offset a decline in valuation multiples, the index can still rise; the core disagreement between Wells Fargo and the bulls is precisely how long this offsetting capacity can last. From an underlying technology perspective, the pace of frontier model R&D and the commercial usage of already-deployed models are two interrelated but non-synchronized growth curves. A model that has completed training can continue to serve programming, customer service, financial analysis, and enterprise knowledge management; as the number of users, task frequency, and task complexity rise, inference demand can still expand. Agentic workflows will also expand a single query into multiple rounds of model calls, data retrieval, tool execution, and result verification. Anthropic's publicly disclosed engineering practices have already demonstrated this operating mechanism. One conclusion that can be derived is that even if the pace of frontier capability improvement is constrained, enterprise application penetration may still drive growth in cloud services and inference revenue; but new semiconductor orders also depend on existing computing power utilization, inference efficiency, and capacity expansion plans, and cannot be simply equated with application revenue. Therefore, the more noteworthy path of earnings expansion is the gradual diffusion of AI returns from infrastructure suppliers to application platforms and enterprises using AI. Software companies can earn revenue through paid features and workflow services, cloud platforms charge for computing power and model services, and other enterprises may improve profits by shortening R&D cycles, increasing sales conversion, and reducing repetitive labor. Only when these returns exceed the costs of additional inference, system integration, and human review will productivity gains sustainable earnings and cash flow. Goldman Sachs also regards whether AI investment can generate sustained profits as the key to whether earnings growth can continue. The high-yield environment and the long-term higher-for-longer interest rate backdrop maintained by the Federal Reserve's FOMC may further widen performance divergence within the AI sector: companies that can prove customer payment, profit growth, and return on capital are better positioned to support valuations; companies that mainly rely on forward expansion expectations need stronger performance evidence. Wells Fargo "steps on the brakes" and cuts its S&P 500 target: tech risks rewrite the U.S. equity bull market outlook Just as Wall Street giants were unanimously calling for a long-term U.S. equity bull market under the AI boom, Wells Fargo's Ohsung Kwon chose to cut his year-end S&P 500 target, saying the decade-long earnings growth cycle will eventually slow and that risks in the technology sector of U.S. and even global markets are accumulating. The chief equity strategist is one of the few on Wall Street in recent weeks to cut his forecast, lowering the target to 7,700 from 7,950. The new target implies upside of slightly more than 1% from the index's Monday close. Kwon is also more cautious on the technology sector, downgrading it to market weight from overweight because the upcoming midterm elections are increasingly becoming a risk for the sector, especially as opposition to data center development grows. Kwon wrote in a newly released research report: "Over the past three months, the number of data center development moratorium government orders taking effect across the United States has increased sharply by 175%, and we expect this momentum to continue, especially against the backdrop of frontier AI labs expressing safety concerns." He added that he believes the probability of a sweeping Democratic election victory is high, which further raises the above risk because the party generally advocates regulating artificial intelligence. As shown in the chart above, Wells Fargo cuts its S&P targetstrategist Ohsung Kwon expects the S&P 500 to rise only 1.1% by the end of 2026. Before Kwon's cut, others on Wall Street had been raising forecasts one after another. Bank of America and Tallbacken Capital Advisors both raised their year-end S&P 500 targets on Monday, while JPMorgan and Wall Street veteran Ed Yardeni raised their respective targets in August. Meanwhile, the divergence between the Trump administration and AI industry leaders has further intensified market pressure, with investors worried about whether these infrastructure investments can actually deliver returns. Developers including Anthropic and OpenAI have been calling for slowing the development of the most advanced AI technology to prevent catastrophe, while the U.S. president is pushing for the United States to stay ahead in the development race. Kwon began taking a cautious stance on U.S. stocks earlier this month, warning that the AI capital expenditure cycle may enter its late stage in 2027. The analyst said he prefers software over semiconductors, and that semiconductor stocks are likely to retest their July lows. The midterm election outcome could benefit the healthcare sector, and Kwon upgraded the sector to overweight from market weight. A Democratic victory in the Senate or House elections could create conditions for restoring additional subsidies under the Affordable Care Act, which would boost hospitals and health insurers with a high proportion of business related to health insurance exchange platforms. Kwon said earnings per share growth exceeding expectations could provide upside momentum for the stock market, but this year's earnings are already at a cyclical high. Kwon wrote: "This is one of the strongest EPS growth cycles in history. By 2027, the annualized EPS growth rate over the past decade is expected to reach 14%, a level exceeded only by the post-World War II bull market of the 1950s." He sees little risk next year, but warns that a slowdown in AI spending will put 2028 earnings at risk. Kwon estimates that the current equity allocation ratio is 72%, the highest since 1969, but by his calculation it should be around 60%. He said the gap between the actual allocation ratio and the level indicated by his model is larger than during the most frenzied historical period of the tech bubble.