The US Manufacturing and Electrification Megatrends Are Here and They’re Way More Than AI | Chris Semenuk

US manufacturing just ended its longest contraction in ISM PMI survey history, with sub-50 readings lasting three consecutive years before crossing above 50 only in the last three months. The recovery is three to four months old as of recording, and Semenuk notes most investors are unaware the sector was even in recession. The three-year downturn was caused by three simultaneous headwinds: a post-COVID de-stocking cycle that left roughly 36 months of excess inventory to work down, higher-for-longer interest rates that caused CFOs to delay capital projects, and continuously changing tariffs that caused manufacturing CEOs to delay expansion decisions. All three headwinds are now dissipating at the same time.

The reindustrialization thesis is not about relocating existing Asian factories to the US but about the US becoming the first choice for incremental global capital investment. US share of global foreign direct investment has risen above 20 percent against a historical range of 10 to 15 percent, and non-residential construction is at all-time highs. Chinese manufacturing labor costs were 30 to 40 times lower than US costs 20 years ago but have since risen markedly, and Semenuk argues total manufacturing cost now includes transportation, IP protection, retooling flexibility, and working capital, making US production cost-competitive for advanced categories. The Biden administration's subsidy strategy and the Trump administration's tariffs together made domestic manufacturing a mandate rather than an option, and Semenuk argues no future administration will campaign on returning manufacturing abroad, making the shift permanent.

Semenuk draws a sharp distinction between reindustrialization and the AI and data center narrative investors typically associate with it. Manufacturing accounts for 26 percent of US electrical power consumption while AI accounts for 6 to 7 percent, making manufacturing the primary driver of the electrification theme. US electricity consumption was flat to negative one percent annually for roughly two and a half decades, meaning no meaningful transmission, distribution, or generation capacity was built during that period. Demand growth is now projected at two to four percent by most estimates and three to five percent by the Energy Information Agency. Conflating reindustrialization with AI spending is, in his view, a significant analytical error.

The grid itself, not generation capacity, is the biggest pinch point. The US grid essentially stopped being modernized in the 1970s and cannot simply absorb large amounts of new generation without risk of overload. PJM power prices went up ten times over a twelve-month period approximately seven months ago, with price caps in place that Semenuk argues have discouraged new generation investment and worsened the shortage. GE Vernova turbine orders placed today will not be delivered until 2031 and require a 25 percent down payment plus proof of permits and an installer, with gas turbine lead times from Vernova, Siemens, and FTAI running approximately four years. GE Vernova's total backlog is north of 90 billion dollars, and Semenuk argues its 80 billion dollar service backlog is more underappreciated than its equipment backlog because every turbine carries a 10-year service agreement. Quanta is described as virtually the only company with its own craft skilled labor force for high voltage projects that does not need to subcontract, and high voltage reinvestment at 765KV is flagged as a major upcoming opportunity with large contracts expected over the next couple of years.

Reindustrialization is entering a second stage in which factories already built are being filled with production machinery. Approximately 70 percent of IIJA infrastructure funding has already been committed, signaling the construction phase is largely complete. Caterpillar's resource industries division is showing order backlog growth not seen since 2012, all three Caterpillar divisions grew simultaneously last quarter for the first time in years, and Semenuk predicts Caterpillar can earn roughly 40 dollars per share annually by 2029 against roughly 20 dollars today. Caterpillar's backlog is north of 60 billion dollars, and order backlogs across industrial businesses are extending two to three years at historically unprecedented levels.

Short-cycle industrial stocks have been left behind for three to five years and are currently trading at trough earnings on trough multiples, while aerospace, defense, and industrial construction names such as Fluor and Primoris are at all-time high valuations. Semenuk's fund is positioned in short-cycle businesses including Rockwell, Emerson, Cognex, Parker Hannifin, Ingersoll Rand, Fastenal, Applied Industrial Tech, Gates, and Eaton, and is avoiding aerospace, defense, and construction names on valuation grounds. Standard one- to two-year EPS estimates cannot be used to value these businesses because earnings have been stagnant, requiring analysis of normalized earnings power three to four years out. He is skeptical of revenue-less companies such as Oklo, NuScale, and Nano, which use the word framework when announcing contracts, contrasting them with profitable established businesses that offer the same thematic exposure without the valuation risk.

Modern regulated utilities are now delivering 8 to 9 percent EPS growth plus 3 to 4 percent dividend yield for roughly 12 to 13 percent total return, compared to the old model of roughly 6 percent total return. Nextera has committed to 8 to 9 percent EPS growth plus 4 percent dividend yield out to 2035. Semenuk notes Nextera trades at roughly 20 times earnings while Constellation trades at roughly 35 times earnings despite selling essentially the same electrons, and that Constellation has not announced a new data center deal in approximately nine months.