The Tariff Dodge Is Unraveling: Why US Firms Are Quietly Heading Back to China

Three years of tariff escalation were supposed to deliver a manufacturing renaissance on American soil. Instead, the data points the other way: tariff-rate convergence between China and its supposed substitutes has erased the arbitrage that justified leaving China in the first place. The clearest evidence comes from Alliance Consumer Group, a Texas flashlight company that, after 2025 tariffs on Chinese goods ballooned past 145%, encouraged its Chinese manufacturer to build a Thai factory and began exploring Vietnam and Cambodia. By mid-2026 the calculus flipped. COO Phil Laster told the New York Times the firm has pulled sourcing back to China: 'Have we pulled back to China? Yes, we have.'

The flashlight gap is one example of a broader pattern documented by the Cato Institute: for product categories including wristwatches, sunglasses, artificial flowers, glassware, vacuum flasks, ceramics, and candles, the China-specific tariff is now identical to the rate applied to Southeast Asian rivals. The expected decoupling premium — the whole economic rationale for moving out of China — has evaporated.

Top-line bilateral trade tells a comforting story: China's share of US goods imports fell from nearly 18% in 2017 to roughly 11% in 2024. But value-added accounting strips out the relabeling trick. PIIE's August 2026 update finds that Chinese value-added content in US imports was about 17.7% just before Trump's first trade war and has barely moved since — implying that the missing 7 points of bilateral share mostly migrated through Vietnam, Mexico, and other transshipment hubs rather than into US factories.

Mexico's imports from Vietnam surged 103% in 2026 to $18.4B in H1 alone — already 82% of full-year 2025 — feeding suspicion of Chinese transshipment routed through Hanoi to dodge Section 301 duties. Vietnam's exports to the US rose 53% YoY in January 2026 alone, reaching $20B, while direct Chinese shipments fell 46% over the same window. The flow has not exited China; it has been re-stamped.

If tariffs were producing a domestic build-out, factory construction would be booming. It is not. Per IoT Analytics' Industrial Macro Pulse of May 12, 2026, US manufacturing construction spending has declined 21% from its June 2024 peak of $239 billion. Strip out the volatile computer/electronic/electrical segment — much of it data-center shell, not traditional factories — and the rest of manufacturing is up only 5.6% since tariffs began, well below what a 'supercycle' would imply.

Kearney's Reshoring Index — a measure of how much US manufacturing import demand is being met domestically — improved from -115 to -86 in 2025, but remains deeply negative: the US still imports far more manufactured goods than it produces. And 59,000 net US manufacturing jobs disappeared between April and November 2025, even as reshoring announcements piled up. The fiscal incentives and political theater are real; the steel-in-the-ground capex is not following.

Headline FDI numbers look healthy: the BEA reported on June 10, 2026 that new foreign direct investment in the US reached $232.2B in 2025, up 49.5% from 2024 — the strongest rebound in three years. The composition, however, is damning for the reshoring thesis. Japan ($50.5B), Germany ($26.7B), and Canada ($23.5B) drove the surge. Greenfield investment from China totaled $80 million — described by ITIF as 'the smallest among major sources,' and the entirety was greenfield (i.e., new US-located capacity), with zero M&A.

German FDI in particular is the canary: Berlin-based multinationals were supposed to be the headline beneficiaries of US industrial policy and tariff walls. Instead they have pulled back sharply. The 'ally-shoring' narrative that reshoring bulls used to support 'Reshoring' capex names has no fresh capital behind it.

The strongest argument against the reversal narrative is Apple's high-profile pivot. By March 2026, the company had moved roughly 25% of iPhone production to India — a 53% increase in India output year-on-year — and publicly aimed to source all US-bound iPhones from India by year-end 2026. But the operational reality is far more fragile than the press releases. In July 2025, Hon Hai (Foxconn) recalled more than 300 Chinese engineers and technicians from its iPhone plants in India, with Foxconn halting further deployment of Chinese staff to those facilities.

The financial logic on Apple's side is also weakening. Apple received an estimated $2.2 billion in tariff refunds in the most recent quarter — a windfall that effectively subsidizes the status quo of China-heavy production. With Walmart collecting roughly $2.9B and Nike around $300M in tariff refunds (per CNN and Forbes, Aug 5–21, 2026), the cost-of-relocation case for retailers and consumer-goods importers has eroded faster than the operational case for staying put has collapsed.

The inversion is real but partial. Where regulatory and quality moats matter — pharmaceuticals, medical devices, aerospace — the reshoring math still works. Roche committed $50B to US manufacturing and R&D over five years in April 2025, with plans for new gene-therapy capacity in Pennsylvania and an AI-focused R&D hub in Massachusetts. Bristol-Myers Squibb has earmarked $40B over five years for US expansion. The 22 new pharma manufacturing sites announced since 2024 cite roughly 44,000 new jobs.

The driver here is not tariffs — it is FDA pre-clearance programs for domestic plants and a Section 232 pharma tariff deadline of July 31, 2026. The asymmetry tells the real story: companies move to the US when domestic production is structurally cheaper or uniquely de-risked; they move back to China when the cost arithmetic dominates. Most consumer-goods manufacturing is in the second category. Most pharma and aerospace is in the first.

Investors who bought into the reshoring supercycle thesis — often via diversified industrials, construction-engineering firms, and tariff-insulated domestic champions — face a reckoning. The companies that benefit from a true reshoring wave are a narrow set: pharma and aerospace names with regulatory moats, plus specialty construction firms servicing pharma/data-center capex. The wider pool of 'US manufacturing' names is at risk of a multiple compression as the underlying capital flows fail to materialize.