Moving Supply Chains Out of China: To Cut Costs, or to Spread Risk?

Let me give the conclusion first, then make the case: this wave of supply-chain shifting is driven mainly by "spreading risk," not by "cutting costs"—but cost decides "where the capacity goes." And the overwhelming majority of it is a "China+1"-style hedge, not an "exit from China." Let's break it down.

1. The driver: from "where is it cheapest" to "where is it safest"

The most direct evidence of motive comes from what companies themselves say and do. A June 2025 Bank of America Institute survey covering 1,029 companies put it bluntly: globalization used to mean "producing in the cheapest country," and under today's geopolitical fragmentation it has become—

"producing in the safest place: a risk-management-oriented allocation of capital."

Corporate survey data points the same way: in Bain's 2024 survey, 81% of executives said they were actively reducing their dependence on China, up from 63% in 2022. The dominant framing at the policy level isn't "cutting costs" either, but "de-risking"—introduced by European Commission President von der Leyen in March 2023 and explicitly defined as "reducing national-security risks while retaining the benefits of a globally interconnected economy," and quickly adopted on the U.S. side (Sullivan, the G7 Hiroshima summit). It is a distinct concept from "decoupling (a complete economic severance)": de-risking doesn't seek to cut ties, only to avoid putting all the eggs in one basket.

What triggered this risk narrative was tariffs (U.S.-China bilateral rates briefly hit 125% in 2025), export controls / technology restrictions, and the "single-source" fragility exposed by the pandemic. One detail showing it's "not purely about tariffs": as early as 2019, Apple ordered at least 15–30% of its capacity moved out of China—before the new round of the tariff war had even begun. Corporate de-risking has, to some extent, been advancing independently of tariffs.

2. This is a "China+1 hedge," not an "exit from China"

The word "moving out" easily makes people picture closing plants and walking away. What's actually happening is far shallower:

• Diversification is concentrated in final assembly, while upstream components and suppliers still depend heavily on China. Rhodium's phrasing: "because global value chains are deeply entangled with China, diversification does not necessarily reduce reliance on Chinese inputs and suppliers in the short to medium term"—the industry calls this "tariff-hopping."
• BCG's 2024 survey: 91% of European manufacturers have adopted China+1; but this is "adding one more," not "subtracting China."

3. The most counterintuitive set of data: trade diversification ≫ investment diversification

If companies were really moving capacity, investment (FDI) should follow. But the data shows "the books moved, the capacity didn't":

Trade share fell 8 points, while actual investment barely moved (−0.7 points). This shows companies are rerouting trade flows and shifting the location of final assembly, not actually relocating production lines. An even more revealing signal: the growth in U.S. imports from Mexico and Vietnam is "highly correlated" with the growth in China's exports to Mexico and Vietnam—that is, Chinese components assembled somewhere else, relabeled, and re-exported. The New York Fed also cautions that the decline in China's share is "exaggerated" by this transshipment; the real degree of decoupling is not as large.

4. Where it's going: Southeast Asia + Mexico, not reshoring to the U.S.

A common misconception is "manufacturing is reshoring to the U.S." The data says otherwise:

• U.S. reshoring has in fact stalled: the Kearney Reshoring Index fell 311 basis points in 2025 and turned negative, reversing two years of positive growth.
• Capacity is flowing to lower-cost places like "Southeast Asia (led by Vietnam) and Mexico," because in the U.S. "qualified labor is scarce and too expensive."

Here "where it goes is decided by cost" becomes clear: capacity didn't return to the high-cost U.S., but went to cheaper Asia and nearshore Mexico.

5. The truth about cost: don't just look at hourly wages, and distinguish by industry

"Move wherever wages are lowest" holds true for only some industries:

• Low-end labor-intensive (footwear/apparel, simple assembly): seeking both low wages and tariff arbitrage—cost is still the main driver.
• High-end electronics / semiconductors: wages are only a small slice. ITIF data shows India's electronics-manufacturing labor runs about $2.19 per hour (even lower than Vietnam or Mexico), but "labor accounts for only about 5% of total semiconductor assembly-and-test cost." So whether this end moves depends on supply-chain security, government subsidies (India's PLI, U.S. CHIPS-style), customer/government requirements, and ecosystem depth—not hourly wages.

In other words: the lower-end it is, the more the motive tilts toward cost; the higher-end it is, the more purely it's a risk/strategic motive.

6. Tariffs: a "live, broad-based, negotiable" major risk

Tariffs are the direct catalyst of this round, but their nature is worth seeing clearly:

• Large in magnitude, and volatile: in 2025 U.S.-China bilateral tariffs briefly spiked to 125% (about 145% total on the U.S. side, including the fentanyl tariff), fell to 10% under a May truce (90 days), and were later extended to November 2026.
• The gap is widening: per Rhodium's October 2025 data, China's trade-weighted average tariff rate into the U.S. is about 41%, versus Vietnam's 18%, Thailand's 16%, and Malaysia's 11%—this "China vs. alternatives" gap is precisely the force pushing capacity toward Southeast Asia.
• But it's broad-based, not aimed at China alone: in 2025 the U.S. "reciprocal tariffs" started at 10% globally and ran 10–41% by country (India 25%, Vietnam about 18–20%). Even the China+1 destinations themselves get hit with tariffs—which is exactly what shows companies aren't doing pure tariff-cost arbitrage, but responding to structural geopolitical risk.

7. Why China can't be emptied out: the ecosystem moat

"Full withdrawal" is rare because China's structural advantages are irreplaceable in the short term:

• About 30–31% of global manufacturing value-added is concentrated in China; even seemingly sizable FDI flows into Vietnam and Bangladesh can shift only a small fraction of China's output.
• Supplier density, cluster effects, the engineer dividend, enormous sunk investment, and the "in China, for China" domestic market—CEPR/RIETI's study of 2009–2022 Japanese-firm data states plainly: "full relocation is economically unprofitable, and geopolitical risk does not systematically foreshadow firms' wholesale exit from China."

The result: even as diversification accelerates, China's global export/manufacturing share may hold steady or even keep rising. This combination of cost + ecosystem advantages is the fundamental reason "the exit hasn't happened."

Conclusion: one sentence, and an industry-by-industry read

The "intent" of this wave and the "relocation of final assembly" are dominated by risk/geopolitics; but "where it goes" is still decided by cost (cheaper Asia and Mexico, not the U.S.); and China's cost and ecosystem advantages are exactly why the "exit" hasn't happened.

A reasonable read for the next 3–5 years: the trend will continue (McKinsey says the reorganization of trade along geopolitical lines "has been running in the data for nearly a decade"), but the form remains a "China+1 hedge" rather than an "exit"; upstream diversification will be far slower than final assembly; and the back-and-forth of tariffs (truces, extensions) will make the pace lurch rather than move in a straight line out.

Main sources: Rhodium Group, "Irrational Expectations" and "A Diversification Framework for China"; Bank of America Institute, "Reshoring vs. Friendshoring" (June 2025); CEPR/RIETI (Japanese firms 2009–2022); CSIS, "A Closer Look at De-risking"; Kearney Reshoring Index; ITIF semiconductor report; U.S. Congressional Research Service CRS R48549; World Bank; McKinsey MGI. The data has undergone multiple rounds of adversarial verification, distinguishing "factual data" from "analytical judgment."