Tariffs are not a factory

On February 20, 2026 the Supreme Court ruled, 6-3, in Learning Resources v. Trump that the International Emergency Economic Powers Act does not authorize the president to impose tariffs — striking down the universal "Liberation Day" duties announced in April 2025. The administration terminated the IEEPA tariffs and partially replaced them under Section 122 of the Trade Act, effective February 24, 2026. Section 122 carries a 150-day statutory limit: those replacement duties expire on July 24, 2026 unless Congress votes to extend them. Tariffs imposed under other authorities — Section 232 on national-security grounds, and Section 301 — were left untouched and are expected to persist.

So the wall that was meant to pull production home is, as this note publishes, eight days from a legal cliff, held up by narrower authorities the courts have not disturbed. That instability is itself part of the story. You do not site a ten-year factory against a tariff that might not survive the summer. The policy pull is real, but it arrives with an asterisk, and capital reads asterisks.

The clean way to test reshoring is to watch what actually gets built. US manufacturing construction spending peaked in 2024 — an annual average of about USD 235.6 billion by the Census Bureau's count — and has drifted down since, roughly 7% lower through late 2025 in nominal terms, and more once inflation is stripped out. The decline is concentrated in one place: spending on computer, electronic, and electrical plants — the semiconductor fabs that drove the 2021 to 2024 surge — fell about 44% from its mid-2024 peak.

Take those out and the rest of manufacturing construction grew about 5.6% in the roughly one year since the tariffs took hold, which after around 3% inflation is close to 2% of real growth. Not a collapse, not a boom — a drift. Kearney's Reshoring Index, which measures how much US manufacturing demand is met at home versus imported, stayed in negative territory through 2025. Manufacturing employment is down about 1% since the April 2025 tariffs took effect. A year of the most aggressive US trade policy in generations produced, on the physical measures, almost nothing that reads as reshoring.

Ask manufacturers directly and the reason is plain. In the Institute for Supply Management's December 2025 survey, 86% of manufacturers said they would pass at least some tariff cost through to customers, while only about 36% said they were actively pursuing domestic production shifts — meaning nearly two-thirds had no intention of reshoring to escape the duties. The ISM's manufacturing survey chair put it bluntly: for many firms it remains, in her words, "cheaper to go offshore" or to shift to a less-tariffed country.

The asymmetry is the whole point. Raising a price is a spreadsheet edit; moving production is a multi-year capital project with a labor problem attached. Faced with a tariff of uncertain lifespan, the rational move for most firms was to absorb or pass the cost and wait — not to pour a foundation. Policy can make offshore production more expensive. It cannot, by itself, make domestic production exist.

Even for the firms that do want to build, the binding constraints are physical, not fiscal. Deloitte's 2026 outlook found trade uncertainty was manufacturers' top concern, named by 78% — but the second wall is labor. Roughly a third of 600 executives surveyed put equipping workers with the right skills at the top of their worry list, and immigrant workers filled nearly one in four US manufacturing production jobs in 2024, a pool now under policy pressure. The trades a new plant needs most — welders, machinists, pipefitters, controls technicians, industrial electricians — are precisely the ones in shortest supply.

On top of the labor gap sit the slower walls: multi-year grid-interconnection queues, a shortage of large power transformers, and permitting timelines measured in years. Applications for new manufacturing facilities fell about 39% year over year by mid-2025, and with capacity utilization stuck in the mid-70s, few firms can justify a greenfield build at all. A tariff changes the reward for domestic capacity. It does nothing about the time, the grid, or the people required to create it.

The money did not sit still — it moved to the parts of the physical economy that could actually absorb it. Data-center construction ran at roughly USD 47 billion annualized by early 2026, up about 31% year over year, with hyperscaler capital spending reported near USD 410 billion for 2025. And inside the factories that already exist, capital shifted from new buildings toward automation: one widely cited projection has mobile robots in warehouses and plants growing from about 103,000 units in 2023 toward 500,000 by 2030, and Deloitte found the share of manufacturers planning to deploy physical AI within two years jumped to 22%, more than double the 9% doing so today.

The pattern is telling. When building new fixed capacity is slow and labor-scarce, capital flows toward density and flexibility — more output from the footprint you already hold — rather than toward pouring fresh concrete you cannot staff. The demand to produce more at home is there. It is being routed away from the thing that takes years and toward the things that take months.

Moduloa's thesis begins from the claim that physical AI erodes the labor-cost advantage that sent production offshore. This year's data adds a sharper and less comfortable point: even when policy manufactures the pull — a 15%-plus tariff wall, permanent expensing, political will — production does not come home on command, because the constraint was never only cost. It is the time, the labor, and the grid capacity needed to stand a factory up. That is exactly the gap a modular, certified, portable model of capacity is meant to close: capacity you can site and re-site faster than a bespoke megafactory, executed by a flexible robotic layer that does not depend on finding a hundred welders in a county that has none.

Read this way, the reshoring data is an argument for the thesis by way of the counterfactual. The trade wall did the demand-side work, and the supply side still could not respond. The honest limits are large and worth stating plainly. Modular certified capacity at scale does not yet exist; humanoid robots are not yet dropping into factories in numbers that matter; and the single biggest driver of physical construction right now is data centers, not the flexible manufacturing this thesis describes. What the year establishes is direction, not arrival — tariffs proved the demand for domestic production is real and policy-backed, and proved just as clearly that the wall alone cannot build the thing behind it. This note scores nothing on the register; it is evidence about the environment those predictions live in, not a test of any dated claim. See the register →

This is a synthesis of government data, analyst reports, industry surveys, and legal analysis. Figures are cross-checked across at least two independent sources; where a number is a single analyst's projection it is marked as such. Key references:

FactCheck.org — Manufacturing construction spending declines under Trump (Census figures) · FRED / US Census Bureau — Total construction spending: manufacturing · IoT Analytics — US manufacturing reshoring boom: what the data says · MarketScale — The reshoring boom that wasn't · Manufacturing Dive — Manufacturers plan price hikes over reshoring (ISM survey) · Deloitte — 2026 Manufacturing Industry Outlook · Perkins Coie — Supreme Court holds IEEPA tariffs unlawful · Baker Donelson — IEEPA tariffs end, Section 122 begins · Thomson Reuters — US tariff authorities after IEEPA · Congressional Research Service — Supreme Court rules against IEEPA tariffs

Limits of this note: construction-spending figures are nominal unless flagged, and market sizing for automation and data centers rests on single-analyst projections that should be treated as directional. The July 24, 2026 Section 122 expiry is the statutory default as of writing and could be extended, replaced, or litigated further. ISM and Deloitte figures are survey-based and reflect stated intent, not audited behavior. Corrections are welcome: shk@moduloa.com.