On June 1, 2026, Digitimes reported that Taiwan had secured preferential treatment under US Section 232 tariffs for most of its exports — after months of negotiation with Washington — while the proposed tariff-free quota regime for Taiwanese chips remains unresolved (Digitimes). The split outcome is revealing. Taipei got clarity on the goods that matter least to the bilateral relationship and continued ambiguity on the one — semiconductors — that defines it.
What Taiwan actually won
The relief is real and worth naming precisely. Under preferential treatment that took effect retroactively from May 1, 2026 and was announced on May 28, Taiwanese non-semiconductor exports are now capped at a 15% reciprocal tariff rate — the same band granted to the European Union, Japan and South Korea, and a fraction of the roughly 52% stacked rate China faces (Taipei Times). Auto parts and wood products are capped at 15%; steel, aluminium and copper derivatives used in aircraft components revert to their pre-tariff MFN rates near 1.12%. For Taiwan's auto-parts sector, worth more than NT$270 billion, the difference between a 15% and a 25% wall is the difference between competitiveness and exclusion.
That relief flows from the US-Taiwan reciprocal trade framework whose memorandum of understanding was signed on January 15, 2026. Under it, the US applies the higher of its MFN rate or 15% to Taiwanese goods, while Taiwan eliminates or reduces 99% of its own tariff barriers and pledges large directed purchases: $44.4 billion in LNG and crude oil, $15.2 billion in civil aircraft and engines, and $25.2 billion in power equipment through 2029 (USTR). This is, in trade terms, a fairly conventional managed-access bargain. It closed.
The case for the chip tariffs — stated fairly
The semiconductor piece sits on different legal and strategic ground, and the case for treating it differently is genuinely strong. The January 14, 2026 presidential proclamation that imposed a 25% Section 232 duty on advanced computing chips rests on a stark finding: the United States manufactures only about 10% of the semiconductors it requires, even as chips underpin national defense and sixteen designated critical-infrastructure sectors (The White House). A country that cannot make the components inside its weapons and data centers has a real vulnerability, and tariffs that push fabrication onshore are a coherent — if blunt — response. Crucially, the measure was narrow: it targeted a defined band of high-performance accelerators such as Nvidia's H200 and AMD's MI325X, and exempted chips bound for US data centers, R&D, startups and civil industrial use. That is restraint, not a sledgehammer.
The deal Taiwan negotiated alongside it is more proportionate still. Rather than a flat duty, Taiwanese firms building new US fabs may import up to 2.5 times their planned capacity duty-free during construction, and completed projects may import 1.5 times their US output without paying Section 232 duties (Global Policy Watch). In exchange Taiwan committed roughly $250 billion in direct investment and a further $250 billion in credit guarantees toward US chip manufacturing. A quota tied to actual onshoring is a far smarter instrument than a blanket tariff: it rewards the behaviour the policy wants instead of simply taxing trade.
Where the policy goes wrong
The problem is not the design. It is that the design is not finished. As of early June, Commerce has not released the product list defining which semiconductors and derivatives the quota regime even covers. Vice Premier Cheng Li-chiun put the cost plainly, saying Taipei hopes for confirmation before implementation to "reduce uncertainties facing our businesses investing in the US" (Taipei Times). That uncertainty is the policy failure. Capital-allocation decisions for a fab run to tens of billions of dollars and lock in for a decade; a firm cannot size a 2.5x duty-free import allowance against a "planned capacity" baseline that has no published definition.
The leverage is being applied with a hammer that hasn't stopped swinging. Commerce Secretary Howard Lutnick has publicly floated tariffs "up to 100%" for Taiwanese and South Korean firms that fail to commit to US production. As a negotiating posture that may extract investment, but as industrial policy it is self-defeating: the surest way to slow the very fab construction the tariffs are meant to induce is to make the rules governing those fabs unknowable. Investment hates variance more than it hates cost.
Taiwan accepted a quota framework in good faith and prepaid with half a trillion dollars in commitments. Washington owes it a defined rulebook in return.
There is also a strategic asymmetry worth flagging. The 25% chip tariff and the threat of more are aimed at the one trading partner whose advanced-node output the US economy cannot currently replace. TSMC is not a marginal supplier to be disciplined; it is the load-bearing wall of the AI build-out. Treating that relationship as leverage to be squeezed indefinitely confuses a one-time onshoring win with a durable supply position the US does not yet have.
The fix is administrative, not legislative
The encouraging news is that the non-semiconductor track shows the system works when it commits. Months of talks produced a clear 15% schedule, a retroactive effective date, and a defined product scope. None of that required new legislation — only a decision to publish terms and stick to them. The semiconductor quota deserves the same finish: a published derivative-product list, a defined capacity baseline, and a sunset on the open-ended tariff threats. Proportionate regulation is not weak regulation. It is regulation a business can actually plan around.