Wells Fargo became the second major bank this week to adopt a cooler stance on U.S. equities, pointing to September's rough track record and election-season uncertainty clouding the AI trade. Kwon's team says investors are questioning how long AI-heavy investment can run at its current pace and expects anxiety about AI-related capital outlays to crest soon. Even so, they note their sentiment indicator still leans constructive after the selloff in mid August.
The analysts highlight two pressure points: potential overbuilding tied to AI and the possibility of additional pauses on data center projects as the midterm elections approach.
On Monday, JPMorgan's trading desk, overseen by Andrew Tyler, the Global Head of Market Intelligence, adopted a short-term defensive stance after Kevin Warsh's hawkish remarks prompted markets to price a higher chance of additional rate hikes this year. That caution met a market already wobbling in AI-adjacent corners.
Chip stocks such as Nvidia, Advanced Micro Devices and Intel were among Tuesday's laggards, along with Magnificent Seven names like Amazon and Tesla. By 11:50 a.m. in New York, the Philadelphia Semiconductor Index was off 1.4% and had touched its lowest intraday mark since Aug. 3. The benchmark has yet to claw back losses from the mid June through July slide, a pattern technical strategists are watching closely.
When experts urge caution, steady investing still wins over time, so download the free Always Be Buying E-Book
"AI demand will continue to accelerate, but the economy/capital will be the constraint," Kwon wrote, adding, "We see a risk that the AI capex cycle will enter its late stage in 2027, unless the economy accelerates enough to warrant further investments. The growth rate is expected to peak next quarter." In the near term, he called Anthropic PBC's S-1 for its upcoming IPO "the most important near-term catalyst."
Kwon favors higher quality over high beta and tilts to software rather than semiconductors - unless Anthropic's filing is so robust that it effectively rubber stamps the AI trade and flips that playbook. Looking further ahead, he calculates that to validate current buildout plans, AI must produce $1.4 trillion in gross output by 2028 and reach $2.2 trillion by 2029. Meanwhile, he and his team see momentum building for data center moratoriums, with more pauses likely as the midterms draw nearer.
September has a reputation for being a grind, and this year the AI story is bumping up against old fashioned constraints like budget cycles and policy risk. If headlines turn choppy, keep an eye on how semis trade versus software, whether Anthropic's numbers clear the bar Kwon set, and whether any new data center pauses crop up. Those clues will tell you if the AI boom still has room, or if it takes a breather while the campaign season plays out.