Rerouted, Not Returned: What Tariffs Changed in US Manufacturing Footprints

A Business Case with a Factory in It

Picture a footprint programme approved in the spring. Its benefits case rests on three assumptions: that a share of Asian-sourced volume will move to a US plant within two years, that the plant will be staffed from the local labour market, and that tariff protection will hold long enough to justify the capital. The programme board sees a reshoring story. The ledger underneath it holds something less solid: a supplier quote, a capex line in a plan and a de-risking workshop.

This paper argues that such a case confuses three different things, and that the confusion is now expensive. The evidence reviewed here says that tariff and industrial policy have changed where US manufacturers source far more than where they produce. Stated intent, in the form of capex plans, quotes and de-risking talk, runs ahead of completed domestic footprint change. Transformation cases sized on announced reshoring, rather than on a longer, more complex multi-country network under a policy regime that keeps moving, are likely to overstate domestic capacity, speed and benefit.

That is a claim about business cases, not about the sincerity of manufacturers or the direction of travel. Reshoring is happening in places. The question is how much of it has been delivered, and what leaders should do about the difference.

What the Flows Show: Rerouting

The best-evidenced finding concerns trade, not factories. The McKinsey Global Institute, working from customs data for calendar 2025, reports that US-China trade fell by around 30 percent, with more than $165 billion of trade pushed away from the corridor. Other sellers replaced about two-thirds of the gap¹. The same report finds limited broad-based evidence of firms bringing production home or relocating it to nearby partners, and notes that Canada's and Mexico's shares of US trade declined¹. The setting was a sharp policy shift: the average effective US tariff rate rose from 2.4 percent in late 2024 to about 22 percent in early April 2025, then eased to about 15 percent by year-end¹.

Kearney's 2026 Reshoring Index, built from trade and output data, points the same way from a different angle. Direct imports from mainland China fell by $135 billion, while the other thirteen Asian low-cost economies gained $194 billion. Imports from the fourteen Asian low-cost countries and regions rose by $60 billion, lifting their ratio to US manufacturing gross output to 14.15 percent from 13.29 percent. Over the same year manufacturing gross output fell by 0.4 percent, and total manufactured imports rose from about $2.85 trillion to roughly $2.98 trillion². The index itself improved, from -115 to -86, but Kearney notes it remains in negative territory, meaning net offshoring².

Mexico is the instructive case. Kearney finds that much of Mexico's gain in computers and electronics is concentrated in downstream assembly, and that Mexico's own imports from Asia rose 17.7 percent in 2025². A product that now carries a Mexican label may still depend on Asian boards and components. The supply chain has been re-architected, and it may be longer rather than shorter.

Both sources carry limits that should travel with the finding. The MGI analysis covers goods only, in nominal dollars, with some economies annualised from partial-year data, and US and Chinese records of bilateral trade differ by over $110 billion¹. Kearney's figures are dollar-denominated and its survey sample is not described in the text². These are reasons to hold the conclusion at medium-high confidence, not to discard it.

What Intent Looks Like, and What It Does Not Prove

Intent is easier to find than delivery. The 2026 survey by the Reshoring Initiative and Regions Recruiting, of 249 respondents, reports that 63 percent of OEM respondents will make US capital investments to support reshoring or domestic expansion in 2026-27, and that the share of contract manufacturers currently quoting reshoring projects roughly doubled, from 16 percent to 32 percent³.

Read the next layer down and the picture changes. Among contract manufacturers asked about customers importing from the China and Taiwan region, 60 percent say those customers are at least discussing de-risking. Only 19 percent say customers are actively transitioning away; another 22 percent say customers are planning ahead, and 19 percent describe informal discussion only³. The distance between discussion and transition is the empirical core of this paper. Kearney's executive survey shows a parallel pattern: 75 percent of respondents reported switching country of origin to other low-cost countries, while 20 percent reported looking at greater reliance on domestic manufacturing².

The survey is useful and it is also an interested-party source. Its publisher is an advocacy body, the report offers no response rate, fielding dates or weighting, and its contract-manufacturer responses skew towards fabricated metal products³. Its authors state that they believe the sample to be representative; no sampling method is given to support that. One figure in it should be handled with particular care. The report says 65 percent of OEMs that have reshored or are reshoring are satisfied with the results in 2026, against 96 percent in 2025³. The questionnaire changed between years, so this is not evidence that reshoring underperforms, and this paper does not use it that way.

The table separates what each source type can and cannot carry.

Why the Gap Opens: Reversible Moves First

The pattern is not mysterious once the decisions are laid out in the order a manufacturer faces them. A tariff shock raises the landed cost of Chinese-origin goods. The cheapest responses are reversible: switch origin to another low-cost economy, build inventory, request domestic quotes, announce capex. Kearney reports that 86 percent of surveyed executives built inventory in the first quarter of 2025, a move that unwinds by design². A domestic plant is the opposite kind of decision. It needs stable assumptions about tariffs, a qualified local vendor base and skilled technicians, and it must be justified over a payback period longer than the policy cycle.

Policy volatility makes waiting the rational choice. In February 2026 the Supreme Court held that tariffs imposed under the International Emergency Economic Powers Act were unlawful, and a Section 122 duty took effect on 24 February for 150 days, with Section 232 duties remaining in place⁵. That was a second policy break within a year. Fifty-seven percent of respondents in the Reshoring Initiative survey name policy uncertainty as their primary challenge³, and in Kearney's survey only 18 percent of CEOs said they were very confident of an acceptable return from re- or nearshoring, down from 47 percent the year before². When the cost of being wrong is a stranded plant and the cost of waiting is a reroute, intent persists and commitment is deferred.

The survey also suggests the pause is not tied to one ruling. Sixty-four percent of respondents reported no change in new factory construction, facility expansion or supply chain reshoring after the February decision³. That is consistent with commitment being held back by something broader than any single measure, though the report does not test the point.

The Mechanism in a Business Case

An illustrative sequence shows how the gap enters a programme. The figures below are invented to show the mechanism; they are not drawn from any source or company.

A manufacturer sources a component family from a single Asian supplier. After a tariff increase, the programme team builds a case to move 40 percent of the volume to a US contract manufacturer within 24 months. The case assumes the domestic plant qualifies in 12 months and reaches full rate in 24. In practice the supplier qualification takes 20 months, technician hiring lags, and the tariff basis for the original saving changes before the plant is running. Meanwhile the procurement team has quietly moved a quarter of the volume to a second Asian supplier as a hedge.

At month 24 the dashboard reports reshoring progress at a fraction of plan. The sourcing mix has changed substantially, but towards a second Asian origin. The network now has more nodes, longer lead times and more inventory. The benefits case, measured against its own headline, has failed. Measured against what the organisation did, it has produced a different asset: a more distributed network that nobody sized, funded or governed as one.

This is the chain the evidence supports: policy shock, reversible reroute, deferred irreversible commitment, longer network, benefits case measured against the wrong asset.

The Strongest Objection: Lag and Cycle

The sharpest counter-argument is that this paper mistakes lag for failure. Factories take years to plan and build, so 2025 data cannot show what 2025 intent will produce. On that reading the early indicators, quotes doubling and strong survey capex intent, are the leading edge of a delayed but real wave. There is also a cyclical signal: IoT Analytics describes a strong manufacturing sector and a cyclical uptick, yet concludes that it is too early to call a reshoring boom⁴.

The objection deserves weight and it sharpens the conclusion in three ways. First, it moves the claim from outcome to instrument: the aggregate data cannot yet falsify reshoring, so a business case must not treat reshoring as proven either. Second, it identifies what would change the judgement. IoT Analytics shows US manufacturing construction peaking at $239 billion in June 2024 and falling 21 percent since, with computer, electronic and electrical construction down 44 percent from its peak as CHIPS Act building fades. Excluding that category, spending rose 5.6 percent in nominal terms between February 2025 and March 2026, or about 2.3 percent after inflation on IoT Analytics' own arithmetic⁴. That is modest growth rather than a surge, and it is the series to watch. Third, a cyclical upswing is a different thing from footprint change. Output rising in a stronger economy would look the same in the data as output rising because a plant opened, until the construction and import-ratio series separate them.

The thesis would weaken if real construction and equipment investment outside semiconductors rose materially, if domestic output's share of US demand increased, if the Asian low-cost import ratio fell, or if a like-for-like repeat of the survey showed active transition well above 19 percent. None of those is established today. That absence is not proof of the contrary.

Limits of the Judgement

Four boundaries apply. First, the evidence is thinner on completed firm-level moves than on flows: no source reviewed here provides a ledger of plants opened, suppliers qualified or sourcing share shifted, so the gap between intent and completion is inferred by contrasting survey intent with aggregate flow and output data. Second, parts of the evidence are interested-party, and the survey's sample is small and self-selected. Third, the argument does not extend to sectors with long-lead industrial policy support, such as semiconductors, pharmaceuticals, defence or energy equipment, nor to firms that have already completed reshoring, nor to footprint decisions outside the United States. Fourth, the analysis does not establish what happened to tariff policy after the 150-day Section 122 period was due to expire on 24 July 2026⁵; any programme relying on the current regime should check it directly.

Some things this paper does not say should be stated plainly. It does not say that reshoring has failed or is not happening. It does not say manufacturers' plans are insincere. It does not say that growth in ASEAN exports is mainly transshipment. The claim is narrower: the measurable change so far is a re-architected network, and the evidence for a returned one is not yet there.

What Follows for Programme and Portfolio Decisions

For transformation, programme and portfolio leaders the consequence is a change in what a footprint case is allowed to assume. Three practices follow from the evidence.

1. Separate intent, committed and delivered footprint in reporting. A quote or capex line is intent. A signed contract or a funded build is commitment. A qualified supplier or a plant at rate is delivery. A benefits case should draw on delivery and price commitment at a discount.
2. Gate funding on delivered milestones, tied to triggers. Because the policy regime has already broken twice in a year, staged commitments that release on observed conditions, such as qualified vendors and stable duty assumptions, cost less than a single large bet.
3. Treat the longer network as the asset. If the realistic outcome is a multi-country structure, then Tier-2 origin visibility, multi-origin supplier options, technician pipelines and total-cost-of-ownership capability are the investments that survive either policy outcome. The Reshoring Initiative survey finds 40 percent of OEMs using total cost of ownership in sourcing decisions, up from 30 percent³.

The test for any footprint case: which line of the benefits calculation would still be true if every announced domestic project slipped by two years and the tariff basis changed?

The useful question for a steering committee is therefore not whether reshoring is happening. It is which part of the case depends on a plant that does not yet exist.