NextFin News - The Trump administration is preparing a fresh round of tariffs on imported semiconductors, Commerce Secretary Howard Lutnick confirmed this week, built around a simple bargain: manufacture in America, or pay the duty. Speaking at the SEMICON Taiwan exhibition in Taipei, Lutnick said the administration is weighing a new round of levies designed to lure more chipmaking to the United States. The announcement is the latest escalation in a tariff architecture that has already layered a 25 percent duty on certain advanced computing chips since January 15, 2026, and a threatened 100 percent charge on overseas-made chips unless companies commit to U.S. production.
The question for investors is not whether the tariffs arrive — they are coming — but whether the exemption structure can actually move chipmaking back to American soil, or merely reprice the global supply chain while AI demand keeps the cycle aloft.
The Situation: A Tariff Regime, Not a One-Off Levy
The new round is the second phase of a plan the White House set out in a January 14, 2026 proclamation issued under Section 232 of the Trade Expansion Act of 1962. That proclamation found that the United States fully manufactures only about 10 percent of the chips it consumes — a decline from roughly 37 percent of global semiconductor output in 1990, according to Commerce Department materials — while consuming about one quarter of the world's semiconductors. It laid out a two-phase response. Phase one imposed a 25 percent ad valorem duty on a narrow category of advanced computing chips, effective January 15, 2026, with exemptions for chips imported to support the buildout of the U.S. technology supply chain, including data centers, repairs, research, startups, consumer electronics, and public-sector projects. Phase two, triggered after trade negotiations conclude, calls for broader tariffs at a rate the proclamation describes only as "significant."
What Lutnick has now confirmed is the shape of that second phase. Rather than a flat across-the-board duty, the administration is designing relief around investment: foreign chipmakers that manufacture in the United States would face little or no charge, while those that do not would pay. The model was previewed in August, when President Trump said he would impose a tariff of "approximately 100 percent" on chips and semiconductors, then added: "But if you're building in the United States of America, there's no charge." Companies that have "made a commitment to build or if you're in the process of building, as many are, there is no tariff."
The stakes are concentrated in a handful of companies. Taiwan Semiconductor Manufacturing, the world's largest contract chipmaker, has committed $165 billion to U.S. investment, announced in March 2025, and added another $100 billion in July 2026, bringing its total Arizona pledge to $265 billion across ten fabs, two advanced packaging facilities, and a research center. Intel, Micron, Texas Instruments, and GlobalFoundries operate or are building domestic fabs. Nvidia, AMD, Broadcom, and Qualcomm are fabless — they design chips but outsource production, mainly to TSMC and Samsung.
Market Reaction: Relief Rallies, Then Jitters
The market's behavior around each tariff headline shows exactly how investors are reading the policy. When news broke on August 27 that the administration was considering a broader round of semiconductor duties — potentially covering servers, laptops, and gaming hardware — the semiconductor sector sold off: the VanEck Semiconductor ETF (SMH) fell 2.74 percent the next day, Nvidia dropped 4.31 percent, and TSMC slipped 2.33 percent. The move was not a panic about lost demand; it was a repricing of policy risk.
The rebound was just as telling. By September 2, SMH had recovered 0.96 percent to close at $550.48, Nvidia gained 2.57 percent to $224.41, Micron added 2.71 percent to $956.08, Intel rose 1.46 percent to $90.05, and TSMC advanced 0.56 percent to $415.50. The iShares Semiconductor ETF (SOXX) closed up 0.65 percent at $501.44. The pattern is consistent with the August episode, when clarification that companies "building in the United States or have committed to build" would be exempt sent TSMC's U.S.-listed shares up roughly 5 percent in premarket trading, with Nvidia, AMD, Micron, and Texas Instruments all finishing higher.
In other words, the market has learned to treat each tariff escalation as a negotiation, not a shock. Investors are pricing the exemption, not the levy — a read that holds only as long as the exemption criteria remain permissive enough for the dominant players to qualify.
The Mechanism: Tariffs as an Investment-Forcing Device
The first-order reading of the policy is straightforward: make foreign chips more expensive, and domestic production becomes relatively cheaper. But that is not actually how this regime works, and treating it as a simple cost shock misses the point. The tariff is not primarily a revenue measure or even a protectionist wall. It is a negotiating and investment-leveraging device — a threat calibrated to extract capital commitments rather than to block imports.
That design is visible in the architecture. The January proclamation explicitly pairs the duty with ongoing trade negotiations, and the Taiwan deal announced January 15, 2026, shows the pattern: Taiwan agreed to facilitate $250 billion in U.S. investment across semiconductors, AI, defense technology, telecommunications, and biotechnology, and in return the U.S. reciprocal tariff on Taiwanese goods would total no more than 15 percent, with future Section 232 duties on Taiwanese semiconductors structured to "reward Taiwanese semiconductor producers that invest in the United States." The duty is the stick; the exemption is the carrot; the object is the factory.
This makes the policy structurally different from a conventional tariff cycle. A cyclical trade measure rises and falls with the business cycle or a negotiating calendar. This one is self-reinforcing: every new investment commitment creates a domestic constituency — construction jobs, supplier contracts, state incentives — that makes reversal politically harder. The CHIPS Act's 35 percent investment tax credit under Section 48D, available only to projects that begin fab construction before December 31, 2026, adds a second clock. Fiscal 2026 is also the final year the Commerce Department receives CHIPS Act funding for incentivizing domestic production. The tariff threat and the tax-credit deadline are pushing in the same direction, and both expire on the same calendar.
The transmission channel runs through capital expenditure, not near-term prices. Building a leading-edge fab takes years and tens of billions of dollars. TSMC's Arizona campus is a 2,000-plus-acre project near Phoenix; Intel's Ohio and Arizona expansions and Samsung's Texas fab are measured on similar timelines. That means the policy's effects are back-loaded: supply-chain companies can remain upbeat in the near term — as they have, citing strong U.S. AI chip demand — while the structural reconfiguration unfolds over years.
The Second-Order Effect: A Two-Tier Global Chip Industry
Push one step further and the exemption structure creates something the market has not fully priced: a two-tier global semiconductor industry. Companies with U.S. fabs, or credible commitments to build them, operate inside a tariff-free zone. Everyone else pays. That split does not map cleanly onto nationality. It maps onto capital intensity and commitment.
Consider the fabless designers. Nvidia and AMD do not own fabs, but they can avoid the duty by committing to use TSMC's, Samsung's, or GlobalFoundries' U.S. production — exactly the relief investors cheered in August, when chip stocks climbed on news that companies "building in the United States or have committed to build" would be exempt. Morgan Stanley analyst Joseph Moore called that outcome "something of a relief." The immediate market read was right for the near term: the exemption blunts the shock.
But the second-order consequence is a concentration of bargaining power in the few companies that can actually build in America. TSMC, with its $265 billion Arizona commitment, becomes not just the dominant foundry but the gatekeeper of tariff relief for the entire fabless ecosystem. Apple, which sources its chips from TSMC, secured exemption after pledging an additional $100 billion in U.S. research and manufacturing. The companies that cannot credibly commit — smaller designers, overseas packaging and testing houses, suppliers in the Philippines, Malaysia, and Vietnam — absorb the cost. Dan Lachica, president of the Semiconductor and Electronics Industries in the Philippines, called the plan "devastating" for his country. That asymmetry is the real distributional story of the policy.
There is also a competitiveness cost that trade hawks inside the administration acknowledge. Tariffs on chips used in U.S.-made servers, laptops, and gaming hardware would raise costs for American manufacturers relative to foreign rivals that can buy the same silicon duty-free. The administration is betting that the AI infrastructure build-out is so demand-driven that customers will absorb higher prices rather than switch suppliers. That bet holds only as long as the AI capex cycle remains strong.
The Counter-Thesis: Exemptions Neutralize the Bite
The strongest case against the structural read is that the exemption structure is so permissive it neutralizes the policy's own force. Citi analyst Christopher Danely argued that the exemptions mean the semiconductor tariff will have "minimal impact," because U.S.-based semiconductor companies with in-house fabs already have U.S. operations, and fabless companies can commit to using TSMC, Samsung, or GlobalFoundries, which have U.S. capacity. On this view, the tariff raises some costs and a lot of uncertainty without materially shifting where chips are made. The Semiconductor Industry Association, while supportive in principle, has emphasized that it wants to see the exemption details — a signal that the industry believes the devil is in the design.
There is real force to this objection. If the exemption threshold is low — a paper commitment rather than steel in the ground — companies will optimize for tariff avoidance rather than genuine reshoring, and the policy becomes a paperwork exercise with a price tag attached to consumers. History offers precedent: tariff threats have been deployed repeatedly since 2025, and some never materialized in the form first announced. The "approximately 100 percent" chip tariff was threatened in 2025 but did not take effect as a flat levy; it was channeled into the narrower January 2026 proclamation and the investment-linked structure.
But this counter-thesis underestimates the cumulative pressure of the regime. It is not one tariff; it is a stack — the 25 percent advanced-computing duty already in force, the threatened 100 percent levy held in reserve, reciprocal tariffs by country, and anti-circumvention rules that penalize transshipment of goods to mask their origin. The administration has shown it is willing to move from threat to action, and the Taiwan deal demonstrates that the threat extracts real capital: $250 billion in commitments in exchange for tariff relief. The relevant question is not whether any single levy bites, but whether the stack changes the location calculus for the next decade of fab investment. On that measure, the evidence already points one way.
What to Watch: The Falsifying Signal
The structural reshoring thesis stands or falls on one observable metric: the U.S. share of global semiconductor fabrication capacity. The administration's own benchmark, cited in Commerce Department materials, is a rise from roughly 10 percent in 2024 toward 20 percent by 2030. If that share does not move meaningfully higher — if announced investment fails to convert into operating capacity — then the exemption structure is doing what the skeptics say: letting companies paper their way out of the tariff while the geography of chipmaking stays put.
Near-term signals to watch: the detailed exemption criteria the Commerce Department must still publish; whether the 100 percent threat is codified in a proclamation or remains rhetorical; the pace of fab construction starts before the December 31, 2026 CHIPS tax-credit deadline; and the reaction of downstream customers — data center operators, consumer electronics makers — to any pass-through of higher chip costs. A second consecutive quarter of cooling AI infrastructure spending would test the administration's assumption that demand is inelastic enough to absorb tariff-driven price increases.
Scenario-wise, the base case is a staggered rollout in which announced investors — TSMC, Intel, Samsung, Micron, and the fabless designers tied to their U.S. lines — navigate the exemptions with limited disruption, while smaller overseas suppliers in Southeast Asia absorb disproportionate costs. The upside case for the policy is that the exemption stack accelerates fab construction enough to lift U.S. capacity share toward the 2030 target. The downside case is that exemptions prove porous, costs leak into AI infrastructure and consumer electronics, and the policy slows the very technology build-out it claims to protect.
Bottom Line
This is a structural shift in trade policy, not a cyclical tariff skirmish: the United States is using tariff exposure as a lever to relocate semiconductor capital, and the exemption design makes that lever self-reinforcing rather than reversible. The near-term market impact is muted by the investment-linked carve-outs, which is why chip stocks have treated each escalation as manageable. But the long-term consequence is a more concentrated, more politicized global supply chain, where access to the U.S. market is priced not in dollars of duty but in billions of committed capital.
"Reshoring semiconductor manufacturing is a top priority for President Trump, whose policies have already secured hundreds of billions of dollars of investments in this key sector," the White House said.
The tariff is not really a tariff. It is a membership fee for the American chip club — and the bill is coming due in fab commitments, not customs receipts.
Explore more exclusive insights at nextfin.ai.
