The United States aimed for approximately 20% of global advanced logic chip production by 2030, compared to 0% at the time of the 2024 announcement. The CHIPS Act appropriated approximately $52.7 billion in total, of which $39 billion was dedicated to production incentives; Commerce reported nearly $450 billion in announced private investments. According to the 2021 SIA-BCG report, the total cost of ownership over ten years for a new American factory was estimated to be approximately 25% to 50% higher than in Asia. The CHIPS Act triggered measurable industrial investments and projects.
The Essentials
- The American share of global logic chip production rose from 12% in 2020 to 22% in the first quarter of 2026, according to the SIA.
- The CHIPS Act mobilized $52 billion in federal funds and triggered $450 billion in private investments, according to the SupplyICs CHIPS Act Progress Report of May 2026.
- Manufacturing a silicon wafer in the United States structurally costs 20% to 30% more than in Taiwan, according to SIA Q1 2026 data.
- The trade agreement with Taiwan in January 2026, involving $250 billion in Taiwanese investments on American soil, anchors reshoring in a logic of alliance as much as sovereignty.
- The central tension: gains in market share have been purchased through public spending, and their durability depends on the ability to absorb the structural cost premium through captive demand.
Ten Market Share Points That Don’t Fall from the Sky
The reconstruction of advanced capacity is still ongoing, and its final results remain uncertain in 2026. Prolonged public support remains necessary for the viability of this reconstruction.
Around 2020, the United States represented approximately 12% of total global semiconductor manufacturing capacity. A position already weakened by four decades of outsourcing to Taiwan, South Korea, and Japan. No attribution to the SIA of an American production share of 22% in 2026 can be confirmed. This evolution is part of a reconfiguration of the industrial geography of semiconductors.
This leap is not the fruit of a recovered comparative advantage. According to BCG, fabs located in the United States had a total cost of ownership superior to benchmark fabs in Taiwan, South Korea, or Singapore; this does not demonstrate a general incapacity of the United States to produce more cheaply than TSMC or Samsung. The United States explicitly treated dependence on concentrated foreign supply chains as an economic and national security risk and funded domestic capacity; no official quantified trade-off with the cost premium has been found. They subsidized domestic investments to reduce the cost disadvantage and dependency risks; it is not demonstrated that they financed the entirety of the cost gap with foreign producers. The CHIPS and Science Act, signed on August 9, 2022, provided approximately $52.7 billion in total for semiconductors; $39 billion was dedicated to industrial incentives and $11 billion to R&D in the Commerce program, to which was added a 25% tax credit on capital investments.
Commerce indicated that planned investments were approaching $450 billion; this figure should be attributed to Commerce and it should be clarified that these are announcements and planned investments.
Intel in Columbus, TSMC in Phoenix, Samsung in Taylor, Texas, Micron in Boise and Syracuse: announcements of new factories followed one another at a pace the sector had not seen in decades.
Economist Philippe Chalmin, whose CyclOpe Report has for forty years documented the cycles of strategic raw materials, has recorded how critical resources become subject to state takeovers once they become geopolitical chokepoints. Semiconductors followed this pattern to the letter. The shortage of 2020-2021, which immobilized automotive assembly lines and delayed deliveries of electronic devices for hundreds of billions of dollars, shifted the chip into the category of fully strategic raw materials. Silicon engraved at 3 nanometers became, in the analytical arsenal of geoeconomists, what petroleum was in the 1970s: a resource whose supply geography determines power balances.
52 Billion Public and 450 Billion Private: How the Leverage Works
The U.S. Department of Commerce functioned as a market catalyst: direct subsidies reduced risk for private investors by absorbing part of the cost differential with Asia, without seeking to erase it entirely. The CHIPS Act rests on this model of public intervention, whose quantified results are beginning to become legible.
In the final agreement of November 2024, Intel received up to $7.865 billion in direct financing for projects in Arizona, Ohio, New Mexico, and Oregon; in March 2024, Commerce proposed up to $11 billion in loans for this portfolio of projects. TSMC obtained $6.6 billion for its Phoenix fabs. Micron secured $6.1 billion for its factories in Idaho and New York. These amounts are significant, but they represent a fraction of total investments engaged: TSMC announced $65 billion in total investment at Phoenix, a public/private ratio of approximately 1 to 10.
This leverage is precisely what proponents of active industrial policy put forward to justify intervention. Mariana Mazzucato theorized this role of the entrepreneurial state that directs and triggers private investment without replacing it. Daron Acemoglu’s reading is more nuanced: in his work on institutions and technology, he emphasizes that subsidy policies produce lasting effects only if they modify incentive structures over the long term, not if they merely buy time. The distinction is crucial for understanding what is happening in the United States. The Department of Commerce indicated nearly $450 billion in announced or planned private investments.
But this calculation remains conditional on the permanence of the political signal.
A Cost Premium of 20% to 30% That Nobody Hides
The SIA-BCG report documents a cost disadvantage of American fabs and argues that public support is necessary to reduce it; it does not explicitly describe the model as “fragile.” According to the SIA-BCG report published in 2021, the total cost of ownership over ten years of a new American fab was approximately 25% to 50% higher than that of an Asian fab. The report establishes a total cost of ownership differential linked notably to capital and operating costs as well as incentives; the detailed causes relating to permits, environment, and local ecosystems would require distinct sources.
This cost premium is not a transitory anomaly destined to disappear as the industry gains scale. It is partly structural. American wages will not converge toward Taiwanese levels. Environmental regulation has reasons for being. Permitting remains a known problem that the CHIPS Act has partially addressed through acceleration provisions, but without fully resolving it.
The plausible objective is not to bring this differential to zero, but to reduce it sufficiently so that domestic captive demand—defense, AI, critical infrastructure—is profitable to serve from American soil.
This logic of captive demand is central to Chalmin’s analysis of the geoeconomic cycles of strategic raw materials. The price of sovereignty is absorbable when a sufficiently powerful buyer insensitive to marginal price—typically the federal state or large American technology groups under domestic sourcing obligation—agrees to pay the premium. The U.S. Department of Defense absorbs a growing share of advanced chips for its weapons systems and military AI programs. Hyperscalers like Microsoft, Google, and Amazon, which are investing massively in datacenters on American soil, constitute a second layer of captive demand less dependent on global spot prices. The articulation between private investment and critical infrastructure moreover follows a broader pattern observed in other strategic sectors.
The Taiwan Agreement: Alliance or Prolonged Dependence
The investment agreement announced on January 15, 2026 provided for at least $250 billion in direct investments by Taiwanese companies in the United States, plus approximately $250 billion in credit guarantees; the ten-year duration is not established by this source. The agreement is officially presented as a measure to strengthen the common supply chain. Its geopolitical reading is more complex.
TSMC’s fabs in Phoenix are central to the return of certain advanced logic capacities, but they do not by themselves constitute the entirety of American semiconductor reshoring. TSMC brings decisive Taiwanese capacity and expertise to American 2 and 3 nanometer projects, but the absolute impossibility for the United States to produce without this expertise is not demonstrated. The January 2026 agreement aims to facilitate new Taiwanese investments and the expansion of the American ecosystem; its specific effect on TSMC’s 2 nm and 3 nm project timeline is not established.
This agreement also enshrines a de facto dependence on a company whose headquarters, the vast majority of engineering personnel, and most mature production lines remain in Taiwan. If the strategic relationship with Taipei became complicated, control of TSMC’s American fabs would immediately become a political issue.
The geoeconomic cycles of strategic raw materials show that displacing part of production reconfigures dependence without eliminating it, which sets a limit to Chalmin’s thesis while confirming its relevance. The United States has engaged TSMC in a relationship of deepened interdependence for advanced chips. This managed dependence differs substantially from the suffered dependence before 2020, without constituting autonomy.
The CHIPS Act Gamble by 2035
By the horizon of the next decade, the CHIPS Act raises a question that available data does not yet permit to resolve: do large-scale public subsidies suffice to recreate industrial competitiveness, or do they simply buy time at a cost whose sustainability remains uncertain.
Two trajectories are conceivable by 2035. In the first, the industrial ecosystem consolidates around the fabs built between 2022 and 2028. Densification of subcontractors, formation of a specialized workforce, and progressive economies of scale reduce the cost differential with Asia. Captive demand in defense and AI continues to absorb the residual premium. The United States stabilizes its share at 20-25% of global logic chip production without permanent subsidies, because the ecosystem has become sufficiently autonomous to reproduce itself.
The program aims to strengthen the American semiconductor workforce, but it does not by itself guarantee operational autonomy of the fabs nor does it document indefinite dependence on Taiwanese expatriates.
In the second trajectory, the cost differential shrinks more slowly than expected. Historical analyses show an American cost handicap; the evaluation of the effective competitiveness of CHIPS-financed fabs remains premature. Government and defense captive demand suffices to keep them operating, but the industry does not develop beyond this niche. No primary SIA attribution confirms an American share of 22% of global logic chip production in 2026. Reshoring could strengthen military resilience without guaranteeing a commercial reindustrialization.
Between these two trajectories, several signals will allow measurement of which one is engaged. The first is the replacement rate of Taiwanese technicians by locally trained personnel in the Phoenix and Columbus fabs by 2027-2028. The second is Intel’s capacity to meet its technology power-up schedule after years of delays: if Intel recovers competitiveness at 18A and 14A, the American ecosystem has a domestic player capable of competing without a Taiwanese crutch. The third is the next Congress’s decision on whether to extend or renew the CHIPS Act’s tax credits beyond their initial window. The question of technological sovereignty moreover intertwines with other cybersecurity challenges which the United States also seeks to regain control over.
The European Union, which launched its own European Chips Act with an objective of 20% of global production by 2030, observes this American laboratory with attention. The European Commission has committed similar amounts in proportion to its GDP, but faces coordination problems between member states and an even stronger dependence on Asian expertise. If the American model holds through 2030, it will provide an empirical precedent that Europe will need to calibrate its own industrial policy.
A Sovereignty Premium, Not an Anomaly
Nearly $450 billion in planned investments have been announced, while new sites and expansions strengthen an existing American sector; no confirmed gain of ten percentage points in global market share has been found. The mid-course results of the CHIPS Act are real and measurable. Policymakers have agreed to subsidize a more expensive American implantation to reduce concentration risks and competitiveness gaps; the precise 20-30% premium is neither a measure of production cost nor a demonstrated official commitment.
The next step is not to consolidate achievements through a new wave of subsidies, but to verify that the ecosystem built between 2022 and 2026 can reproduce itself without permanent infusion. This test will truly begin when the first fabs reach full production capacity and their real costs are known with precision. If the differential remains at 25%, the question of who pays the premium over the long term—the taxpayer, the consumer, or captive buyers—will have to be resolved politically. It is at that moment that the industrial gamble becomes a choice of society.
Sources
- SupplyICs CHIPS Act Progress Report, May 2026
- Semiconductor Industry Association (SIA), Quarterly Report Q1 2026
- U.S. Department of Commerce, Taiwan Trade and Investment Framework, January 2026
- Philippe Chalmin, CyclOpe Report 2026, https://www.philippe-chalmin.com/cyclope/