The "AI slowdown theory" collides with the 5% "global asset pricing anchor," sharply escalating tech risks! Wells Fargo redraws the U.S. stock investment landscape and lowers the S&P 500 target.

According to Zhiyin Finance APP, US stocks and even global stock markets appear to be undergoing a stress test: can the AI-driven profit growth trajectory withstand valuation compression, with the semiconductor sector being the first to take a hit. Ohsung Kwon, Chief Equity Strategist at Wells Fargo, has lowered the year-end target for the S&P 500 from 7,950 to 7,700, leaving just about 1.1% upside compared to Monday's US market close. He also downgraded the technology sector from “overweight” to “equal-weight.” This Wall Street giant now favors software sectors benefiting from the AI application boom over the currently hottest semiconductors and has upgraded the more defensive healthcare sector.

The strategist is increasingly concerned that years of profit expansion have pushed market expectations to near historical highs. Meanwhile, AI capital expenditures, state policies on data center construction, and ongoing fiscal and monetary policy uncertainties are piling up. Notably, Wells Fargo’s chief equity strategist is not greatly worried about profits in 2027 but is mainly on alert for the potential hit to 2028 profits if capital expenditures related to AI data centers slow down. Thus, this adjustment amounts more to a reassessment of the long-term growth and valuation of US equities.

The recent “AI slowdown theory” sweeping global stock markets has quickly priced these worries in—AI leaders such as Anthropic and OpenAI jointly called over the weekend to slow the pace of frontier AI model development. On September 12, Anthropic CEO Dario Amodei called for the deceleration of frontier models' capability improvement, so that safety research and protective measures can catch up; OpenAI CEO Sam Altman and industry leaders like Elon Musk of SpaceX and Tesla soon voiced their support.

The first trading day after Anthropic CEO Amodei’s groundbreaking “slowdown theory”—on September 14—saw “AI chip super-giant” Nvidia’s stock drop around 3.4%, and the global semiconductor benchmark—the Philadelphia Semiconductor Index—plunge about 6%, a rare magnitude, highlighting the pricing-in of risks arising from the AI slowdown debate.

AI slowdown combined with surging 10-year US Treasury yields: US and global stocks now compete on their ability to deliver profit growth amid the AI frenzy

The market has thus begun to reassess expectations for large-scale training investments, model rollout pace, and future compute procurement growth. On September 14, the Philadelphia Semiconductor Index fell sharply by about 6%; South Korea’s KOSPI Index—known as a global AI computing power investment barometer—fell over 3% on Monday and dropped another 0.85% on September 15, closing at 6,627.26, its fourth consecutive daily decline.

For the semiconductor sector—the core beneficiary of the global trillion-dollar AI computing infrastructure boom—valuation may be adjusted sharply and in advance, even if current orders remain strong, as long as investors downgrade expectations for the rate and duration of follow-up orders’ growth.

Meanwhile, the “anchor for global asset pricing”—the 10-year US Treasury yield—rose intraday to 5.012% on September 14, touching its highest level since 2007 before retreating to 4.960%. By the close of US markets on September 15, the 10-year yield held above 5%, closing around 5.04%, again marking the highest since 2007. This “anchor” is a key reference for long-term risk-free US dollar rates, impacting corporate financing costs and the discount rate used to convert future profits into present value. Rising energy prices and inflation have driven yields higher, forcing tech stocks to grapple with both rising funding costs and adjustments to long-term growth expectations.

Another change Ohsung Kwon cited: in the past three months, the number of government orders pausing AI data center development in the US has surged by 175%. From an investment mechanics perspective, pricier financing and slower project delivery extend the payback period for some large-scale AI investments and worsen market fears that AI infrastructure-related orders may fail to materialize, with free cash flow potentially turning negative.

By contrast, Wall Street bulls’ main reason for optimism about a US (and global) bull market is that the unprecedented AI infrastructure boom and the flood of AI applications into a wide range of industries continue to underpin corporate profit expansion. Goldman Sachs has publicly projected a year-end S&P 500 target of 8,000 points, based on earnings growth driving the index higher while the valuation multiple remains stable; Yardeni Research President and Chief Investment Strategist Ed Yardeni maintained his 8,400 point target on September 12 while raising his level of caution for bearish scenarios; Michael Purves, CEO and founder of Tallbacken Capital Advisors, raised his year-end target from 7,400 to 8,500, citing strong, sustained, and broad-based earnings growth and allowing for moderate P/E expansion to add further upside potential.

Though their valuation assumptions differ, all three focus on whether corporate profits can keep materializing. According to the formula “index points = EPS × P/E,” as long as earnings growth offsets the decline in the valuation multiple, indices can keep rising. The crux between Wells Fargo and the bulls is how long this offsetting capacity can last.

From a technical perspective, leading-edge model development and the commercial usage of deployed models are two correlated but non-synchronous growth curves. A trained model can continue to serve programming, customer service, financial analysis, and enterprise knowledge management. As the user base, task frequency, and complexity increase, inference demand can still expand. Agentic Workflows turn a single query into multi-step model calls, information retrieval, tool execution, and results validation. Anthropic’s public engineering practices have demonstrated this operational model. One key inference: even if frontier model improvement slows, enterprise application penetration can still drive cloud and inference revenue growth. However, incremental semiconductor orders depend on the utilization of existing compute, inference efficiency, and expansion plans—they are not directly interchangeable with application income.

Thus, the more critical profit expansion path is the spread of AI gains from infrastructure providers to application platforms and AI-using firms. Software companies generate revenue through paid features and workflows; cloud platforms earn from compute and model services; other businesses can improve profits by shortening R&D cycles, raising sales conversion, and reducing redundant labor. Only if these gains exceed the costs of inference, systems integration, and human review will productivity improvements translate into sustainable revenues and cash flow.

Goldman Sachs also identifies the ability of AI investment to deliver lasting profits as key to whether earnings growth can continue. Higher yields and the Fed’s FOMC keeping rates “higher for longer” will likely further widen performance gaps within the AI sector: companies that can demonstrate customer payments, profit growth, and capital returns are better able to support their valuations; those depending primarily on long-term growth expectations need stronger performance evidence.

Wells Fargo “hits the brakes” by lowering S&P 500 target: Tech risk reshapes US equity bull market outlook

Just as Wall Street giants are unanimously calling for a long-term bull market in US stocks led by the AI boom, Ohsung Kwon from Wells Fargo has opted to lower the year-end S&P 500 target, arguing that the decade-long cycle of profit growth will inevitably slow and risks in the technology sector of US and global equity markets are accumulating.

This chief strategist, one of the few on Wall Street to cut forecasts in recent weeks, dropped his target from 7,950 to 7,700. The new target means there’s only slightly more than 1% upside compared to the index's Monday close. Kwon is also more cautious on technology, downgrading it from overweight to equal-weight, citing rising risks ahead of midterm elections—especially as opposition to data center development intensifies.

In his latest research note, Kwon wrote: “In the past three months, the number of active government orders pausing data center development across the US has increased by 175%—a trend we expect to continue, especially as frontier AI labs voice safety concerns.” He added that he believes the likelihood of Democrats winning across the board in the elections is high, which further increases those risks, as the party typically advocates for regulation of artificial intelligence.

As shown above, Wells Fargo lowered its S&P 500 target—strategist Ohsung Kwon estimates the index will rise only 1.1% by the end of 2026.

Before Kwon cut his target, others on Wall Street successively raised theirs. Both Bank of America and Tallbacken Capital Advisors raised their year-end S&P 500 targets on Monday, while JP Morgan and Wall Street veteran Ed Yardeni did so in August.

Meanwhile, the rift between the Trump administration and AI industry leaders has further intensified market pressure, with investors worrying whether these infrastructure investments will deliver returns. Developers including Anthropic and OpenAI have been calling for a slowdown of cutting-edge AI technology to avert disasters, while the US President seeks to keep America ahead in this development race.

Kwon began to take a cautious view on US stocks earlier this month, warning that the AI capital expenditure cycle could enter its late stage in 2027. He prefers software over semiconductors and predicts semiconductor stocks may retest July lows.

The outcome of the midterm elections could benefit the healthcare sector; Kwon upgraded the sector from equal-weight to overweight. A Democratic victory in the Senate or House could create conditions to restore the additional subsidies under the Affordable Care Act, boosting hospitals and health insurers with significant Medicare exchange-related business.

Kwon believes earnings per share exceeding expectations could drive the market higher, but this year’s profits are already at cyclical highs.

Kwon wrote: “This is one of the strongest cycles of EPS growth in history. By 2027, the decade’s annualized EPS growth is expected to reach 14%—a level exceeded only by the post-World War II bull market in the 1950s.”

He sees little risk for next year but warns that an AI spending slowdown will put 2028 earnings at risk.

Kwon estimates current equity allocations are at 72%, the highest since 1969, but according to his calculations, the optimal level should be around 60%. He notes that the gap between actual allocations and the model-implied level exceeds that of even the most feverish moments of the dot-com bubble.