Taiwan Semiconductor Manufacturing, the world's leading contract chipmaker, is navigating a sharply divergent overseas expansion story — one where its mature-node China fabs are quietly minting cash, its Arizona operations have swung to profitability on AI-fueled demand, and its Japan and Germany facilities continue to absorb heavy startup losses.

The contrast, laid out in TSMC's 2025 annual report, underscores the uneven economics of geographic diversification in the semiconductor industry. Fabs built to serve geopolitical imperatives — rather than pure cost optimization — carry a financial burden that mature, depreciated plants in China do not. For investors tracking TSMC's margin trajectory, the regional breakdown offers a nuanced picture that aggregate revenue figures alone cannot capture.

The headline numbers reinforce the divergence. TSMC's two China fabs in Nanjing and Shanghai together generated approximately NT$39.177 billion (roughly US$1.2 billion) in net profit in 2025, while Arizona swung from a NT$14.3 billion loss to a NT$16.1 billion profit. Japan's Kumamoto fab, meanwhile, saw losses more than double, and Germany's ESMC joint venture continued to bleed ahead of production.

TSMC did not respond to a request for comment beyond what is contained in its annual report.

China's Mature Fabs: Low Costs, High Returns

According to TSMC's 2025 annual report, as cited by Mirror Media, the company's Nanjing plant recorded a net profit of approximately NT$27.606 billion in 2025, while its Shanghai facility contributed NT$11.571 billion — bringing the combined China profit to NT$39.177 billion (approximately US$1.2 billion).

The economics are straightforward. Both fabs focus on 16nm and 28nm mature nodes, primarily serving the automotive, microcontroller unit (MCU), and power-management IC markets. Mirror Media notes that the equipment at these facilities is largely fully depreciated, stripping out a major cost component and allowing revenue to flow almost directly to the bottom line.

Operationally, the Nanjing plant produces approximately 20,000 wafers per month at 16/12nm and 40,000 wafers per month at 28/22nm, according to data previously cited by Guancha — representing roughly 3% of TSMC's total global capacity. That relatively modest share of output is generating a disproportionately high share of overseas profit, a function of cost structure rather than scale.

The profitability of these fabs stands in stark contrast to TSMC's newer overseas investments, where depreciation cycles are just beginning and utilization rates are still ramping. The China operations effectively illustrate the financial ceiling that geopolitically motivated fabs in higher-cost jurisdictions are working toward — but have not yet reached.

Arizona: AI Demand Drives a Profitable Turnaround

TSMC's Arizona operations delivered what may be the most significant earnings inflection in the company's overseas portfolio last year. The facility swung to a NT$16.1 billion profit in 2025, reversing a NT$14.3 billion loss in 2024, according to the annual report.

Mirror Media attributes the turnaround to a combination of full-capacity bookings and favorable product mix. NVIDIA Corp., Apple Inc., Alphabet Inc.'s Google, and Microsoft Corp. have collectively filled the fab's order book, with AI-related workloads commanding premium pricing and delivering peak margins. A brief disruption in the third quarter — stemming from an industrial gas shortage — proved temporary, and the facility closed the year in positive territory.

The Arizona result carries broader strategic significance. TSMC is accelerating its U.S. footprint, with 4nm already in production and 3nm expected to come online on American soil as early as 2027. The U.S. accounted for over 74% of TSMC's total 2025 revenue, with regional sales jumping 42.3% year-on-year to NT$2.83 trillion, according to the annual report.

The speed of Arizona's profitability recovery — from a sizable loss to a meaningful profit within a single fiscal year — will likely be cited by management as evidence that the U.S. expansion model is commercially viable, even amid the elevated cost environment that has historically made American semiconductor manufacturing difficult to justify on financial grounds alone.

Kumamoto: Losses Deepen as Upgrade Plans Take Shape

Japan's Kumamoto fab presents a more complicated picture. Losses at the facility more than doubled in 2025, rising from NT$4.4 billion in 2024 to NT$9.767 billion, according to TSMC's annual report — a deterioration that reflects the ongoing burden of depreciation, staffing, and operational costs ahead of full-scale production ramp.

The widening losses arrive even as TSMC has proposed upgrading its second Kumamoto fab to 3nm capability, a move that would represent a significant step up in process technology for Japan and signal a longer-term commitment to the region. The upgrade proposal suggests confidence in Japan's role within TSMC's global network, but the current financial trajectory indicates that profitability at Kumamoto remains a medium-term target rather than an near-term reality.

The contrast with the Arizona recovery raises a question that investors are likely to probe: whether Kumamoto's loss profile will follow a similar trajectory toward profitability, or whether structural cost differences in Japan — including labor, utilities, and construction — will extend the timeline.

Germany: Cautious Stance as Automotive Demand Waits to Recover

TSMC's European venture, ESMC — a joint fab that broke ground in 2024 and is designed to produce 28/22nm and 16/12nm chips — posted a net loss of NT$688.6 million in 2025, widening from a NT$556.9 million loss in 2024, according to the annual report. The losses remain relatively modest in absolute terms but reflect the early-stage burden of equipment depreciation, utilities, and pre-production operational costs.

More telling than the financials is TSMC's strategic posture toward Germany. Mirror Media and the Economic Daily News have both reported that TSMC Chairman C.C. Wei has emphasized that the ESMC site is oriented entirely toward automotive and industrial chip demand, focused on 28/22nm and 16/12nm nodes, with little prospect of advancing into sub-5nm processes.

Construction remains on schedule, but TSMC has adopted a wait-and-see stance on the mass production timeline. The original 2027 ramp-up target is now described as uncertain, contingent on a clearer recovery in automotive chip demand — a market that has faced inventory corrections and softening end-market conditions. The cautious posture in Germany stands in contrast to the more aggressive capacity push underway in the United States and, to a lesser extent, Japan, suggesting TSMC is calibrating its commitment to Europe against near-term demand visibility rather than geopolitical momentum alone.