The Chip Triangle: How TSMC, SMIC, and Intel Ohio Are Redrawing the Geopolitical Map of Semiconductor Power

On a construction site outside Columbus, Ohio, the largest private investment in American manufacturing history has slowed to an uncomfortable crawl. Intel's $20 billion Fab 52 and Fab 62 complex — flagship infrastructure of the CHIPS and Science Act's domestic revival ambitions — missed its 2025 production target and is now projected to begin volume output no earlier than late 2027. Twelve hundred miles away in Phoenix, TSMC's N4 node fab shipped its first Apple silicon wafers in February 2026, a historic milestone that nonetheless arrived two years behind schedule and required a reported $40 billion in cumulative subsidy commitments from both Washington and Taipei to materialise. Meanwhile, in Shenzhen, SMIC quietly disclosed in its Q4 2025 earnings that its N+2 process node — a domestic approximation of a 7-nanometre architecture — had reached meaningful volume yield, a development the US Commerce Department's Bureau of Industry and Security acknowledged only indirectly, by tightening the foreign direct product rule for another tranche of equipment suppliers.

These three sites form what Chris Miller, in his definitive account of semiconductor statecraft, Chip War, calls the "physical chokepoints" of modern geopolitical contest. The chip is not merely a commodity; it is, as Miller argues, the foundational input for every weapons system, every AI model, every surveillance apparatus that states deploy to project power. When Washington invoked the Export Administration Regulations against Huawei in 2020 and extended those restrictions to virtually every Chinese advanced-logic buyer in 2022 and 2023, it was making a choice historically reserved for naval blockades: attempting to deny a rival the physical means of industrial modernity. The consequence is a triangle of industrial competition — Arizona, Ohio, Shenzhen — whose outcome will shape the computational infrastructure of the next thirty years.

TSMC Arizona: Subsidised Dependence or Strategic Anchor?

The narrative Washington prefers is one of triumphant reshoring: American soil, American fabs, American jobs. The narrative that survives scrutiny is considerably more ambiguous. TSMC's Arizona facility is, at its core, a Taiwanese company operating under American subsidy conditions — conditions that include technology-sharing provisions, restrictions on expansion in China, and requirements that TSMC alert the Commerce Department to significant customer order shifts. The foundry has not relocated its leading-edge research and development; its most advanced nodes, currently the N2 and N2P architectures being qualified at Hsinchu, remain in Taiwan. Arizona produces trailing-edge N4, which is competitive but not at the frontier.

This matters because the logic of advanced-node supremacy is cumulative. TSMC's dominance at N3 and N2 is not merely a manufacturing advantage; it is an ecosystem lock-in — EUV lithography tooling calibrated over years, process chemistry refined through hundreds of thousands of wafer runs, customer co-development relationships baked into design rules. The CHIPS Act funds bricks and equipment, but it has not funded the patient, decades-long accumulation of tacit knowledge that makes TSMC's Taiwanese fabs irreplaceable. The Arizona site is better understood as a geopolitical insurance policy than as a genuine transfer of technological sovereignty. If cross-strait tensions were to interrupt Taiwanese production, Arizona could sustain N4-level output for the defence and automotive sectors — but it could not sustain the frontier node production on which AI accelerator chips depend.

The subsidies themselves reveal a further tension. The CHIPS Act's $52 billion allocation has been distributed under conditions that effectively exclude the Global South from any corresponding technology transfer benefit. Countries in sub-Saharan Africa, Southeast Asia, and Latin America that manufacture downstream electronics using advanced chips have no seat at the table in Phoenix or Columbus. The reshoring narrative is a mission-economy intervention that serves a narrow industrial-national constituency while externalising the costs — supply chain disruption, price elevation, access restriction — to the periphery.

SMIC's Quiet Advance and the Limits of Export Control

The SMIC story is one that Washington's export-control architecture was explicitly designed to prevent, and has conspicuously failed to prevent. When the Commerce Department added SMIC to the Entity List in December 2020, the intent was to deny the Chinese foundry access to American equipment and foreign-produced equipment made with American technology, thereby freezing SMIC at the 28-nanometre node and ensuring that China's domestic semiconductor ecosystem could not close the gap with TSMC, Samsung, or Intel Foundry.

What has happened instead is instructive about the structural limits of unilateral technology denial. SMIC has, through a combination of reverse engineering, equipment modification, multi-patterning workarounds, and selective acquisition of older Dutch ASML machines (which fall below the EUV threshold), achieved volume yield on its N+2 node. Independent analysis by TechInsights, published in late 2023 on the Huawei Mate 60 Pro chip, confirmed SMIC-manufactured 7-nanometre equivalent silicon in a consumer device — a development that violated no explicit export control rule because SMIC had sourced no controlled equipment after the designation. The Bureau of Industry and Security's subsequent rule tightening on ASML's DUV tools, and on KOKUSAI ELECTRIC's deposition systems, was an acknowledgement that the original control regime had gaps that Chinese engineers exploited methodically.

When the structural incentive to acquire a capability is sufficiently powerful, institutional constraints are treated as engineering problems rather than prohibitions. China's semiconductor programme is driven by a recognition — stated explicitly in the Made in China 2025 documents and every subsequent Five-Year Plan — that dependence on foreign chips is an existential vulnerability. The export control regime has accelerated, not diminished, the political will to eliminate that dependence. SMIC's N+2 achievement is a demonstration of proof of concept; the investment roadmap behind it suggests N+1 and N-1 (5-nanometre equivalent) within a three-to-five-year horizon, assuming continued domestic equipment development from AMEC, NAURA, and the recently capitalised SMEE lithography programme.

Intel Ohio: Mission Economy Delayed, Promises Deferred

Intel's Ohio investment is the most overtly political of the three sites, and the most troubled. The company's decision in late 2024 to pause Phase 2 of the Columbus complex — delaying the second fab pair and the associated research centre — came after Intel's foundry services division posted operating losses exceeding $7 billion in fiscal year 2024. The new leadership team under CEO Lip-Bu Tan has restructured the foundry ambition, prioritising internal product competitiveness over third-party customer acquisition, and has entered into a manufacturing partnership arrangement with TSMC for its upcoming Panther Lake client CPU, effectively outsourcing the most advanced silicon to the company its Ohio expansion was meant to rival.

The political implications are difficult to overstate. The CHIPS Act was sold to Congress on the premise that domestic investment would produce domestic capability — that American taxpayer subsidy would generate American technological leadership, not American real estate hosting foreign-designed processes. The emerging reality is more complex: Intel Ohio may produce chips at competitive nodes by 2028, but it will do so having ceded the manufacturing learning curve to TSMC Arizona for several critical product generations. The risk of public investment de-risking private profit without ensuring public benefit is entirely applicable here. The CHIPS Act grant to Intel, announced at $8.5 billion, was conditioned on milestones that Intel's restructuring may not meet on the original timeline.

What remains at stake in Ohio is not merely one company's industrial strategy. It is the question of whether the United States can reconstitute a domestic semiconductor industrial base after four decades of offshoring — not through market signals alone, which produced the offshoring in the first place, but through the sustained, patient, mission-oriented public investment that built the original Bell Labs and Sematech programmes. The answer, as of April 2026, remains genuinely uncertain.

The Stakes: Who Controls the Computational Substrate?

Chris Miller's central argument in Chip War is that semiconductor leadership is not a commercial advantage but a strategic one — that whichever power commands the production of advanced logic chips commands the ability to build the weapons, the surveillance systems, and the AI infrastructure that define military and economic power in the 21st century. The triangle of TSMC Arizona, SMIC Shenzhen, and Intel Ohio represents three competing visions of how that command should be structured: American-subsidised allied offshoring, Chinese domestic autarky, and American reindustrialisation.

The uncomfortable truth that emerges from examining all three simultaneously is that none of the three models is succeeding on its own terms within its projected timeline. TSMC Arizona is producing chips but not leading-edge sovereignty. SMIC Shenzhen is advancing faster than the export control architecture anticipated but remains two to three nodes behind the frontier. Intel Ohio is behind schedule, undersubscribed by customers, and partially dependent on the very foreign foundry it was meant to supplant. The computational substrate of the 21st century is being contested, expensively and imperfectly, across a triangle whose resolution is likely to be a decade away — and whose shape, when it resolves, will determine the architecture of global power in ways that dwarf the strategic consequences of any single military engagement.

For the Global South, which consumes chips it does not produce and hosts assembly operations it does not control, the contest among these three poles offers no obvious ally. The reshoring of advanced manufacturing to American and Chinese soil simultaneously reduces the geographic distribution of semiconductor capacity and entrenches the dependency relationships that have characterised the global electronics supply chain since the 1980s. The question of who controls the chip is also, inescapably, the question of who controls the terms on which the rest of the world accesses the computational infrastructure of modernity.

The Monexus tech desk reports the semiconductor race without defaulting to the Washington framing of "allied reshoring vs Chinese theft" — a framing that obscures the public-subsidy dynamics, the missed milestones, and the absence of any Global South voice in the redesign of the chip supply chain.