A certain type of forecast ages poorly, and the Wall Street forecast regarding American manufacturing from late 2024 has aged worse than most. Major desks at the time came to the nearly unanimous conclusion that there were still two or possibly three good quarters left in the post-pandemic industrial bounce. A contraction was already indicated by the ISM Manufacturing PMI’s drift below 50. Spending on building new factories was declining. The sector’s employment was stagnant. The narrative made sense. It simply proved to be incorrect. After seven quarters, the same industry is still growing in unexpected areas and attracting funding for long-term projects.
Both policy and accident contributed to the changes that have occurred since then. Signed last summer, the One Big Beautiful Bill Act increased the advanced manufacturing investment credit from 25% to 35% and made full expensing for new equipment permanent. For any business considering whether to start construction on a manufacturing facility or a transformer plant, that is a significant figure. However, practically everyone in the industry now acknowledges that the AI buildout has been the main driver of demand. Software is not the only thing that powers data centers. They require switchgear, power management equipment, transformers, cooling systems, and other tangible items manufactured in physical locations, frequently by businesses whose names are seldom mentioned in tech publications.
If you know where to look, you can see it on the ground. The construction sites surrounding TSMC’s Arizona campus in Phoenix continue to be active after dusk. A Microsoft data center is currently situated behind chain-link fencing that is almost the length of a county road in Mount Pleasant, Wisconsin, where the former Foxconn site struggled for years. Large transformers, which take 18 to 24 months to manufacture, are reportedly sold out by original equipment manufacturers through 2027. Industry executives have a feeling that capacity hasn’t kept up with the order book and won’t for some time. This feeling is typically voiced in private at conferences rather than during earnings calls.
The majority of the story can be found in semiconductor manufacturing alone. In order to revitalize the U.S. chipmaking ecosystem, businesses have committed more than $500 billion in private investment; by 2032, domestic capacity is expected to nearly triple. It’s not a rounding error. Every business that supplies the steel, chemicals, specialty gases, and testing equipment will be impacted by this structural shift in the chip manufacturing process. Investors don’t seem to think this will end soon. The reshoring figures from early 2025—surveys of more than 500 manufacturers demonstrating a strong desire to resume production—continue to be cited because they remain consistent with actual events.
Beneath the surface, there are reasons to exercise caution. In April, Dow warned that the Middle East conflict had increased the cost of oil and naphtha, making the supply of chemicals more difficult until 2026. Even though prices are staying the same, costs are increasing. Global readings have been inconsistent, and the Pakistani manufacturing PMI recently fell below 50 in April for the first time in months. If the global narrative continues to deteriorate, it’s still unclear if the U.S. story will hold. According to McKinsey’s March intelligence summary, roughly 60% of executives surveyed still anticipate higher profits, which is in line with the previous two quarters but not as optimistic as it was a year ago.
This cycle’s scale isn’t what makes it intriguing. It’s its tenacity. This is not how manufacturing is meant to surprise people. Everyone learns the rhythm of cycles turning, capacity catching up, and demand cooling. It’s difficult to ignore the feeling that something truly structural is going on beneath the noise as you watch this run of strength continue past quarter seven. It might be reshoring. AI infrastructure could be the cause. Perhaps the world has finally realized that someone needs to create the infrastructure that powers the digital economy. For whatever reason, the predictions that called for two quarters are now in the same drawer as all the other forecasts that misjudged the duration of a real shift.
