TSMC Abroad: What Subsidies Can and Cannot Buy

Everyone wants their own fab, and almost nobody understands what that requires. This story covers the Arizona difficulties, the Japanese success, the German project, the cost differentials and why the leading edge is not moving — part of the Taiwan Company Stories hub.

Why did governments start funding fabs?

Because the pandemic chip shortage demonstrated that a small number of Asian facilities determined whether cars, medical equipment and consumer electronics could be produced anywhere, and that no domestic policy lever existed to change it.

The response was the largest industrial subsidy wave in decades: American, European, Japanese, Korean, Indian and Chinese programmes committing enormous sums to domestic semiconductor capacity, with the leading foundry as the most desirable investor.

The strategic objective is supply security rather than commercial return. Governments are effectively buying insurance, and the price is the difference between what a fab costs to build and operate locally versus in Taiwan.

What went wrong in Arizona?

Schedule delays, cost overruns and workforce friction. The project ran behind its original timeline, faced disputes over labour practices and training approaches, and required bringing substantial numbers of experienced engineers from Taiwan to establish operations.

The underlying issues were cultural and structural. Fab construction and operation in Taiwan relies on a workforce and contractor base accustomed to particular schedules, urgency norms and problem-solving approaches, and the American labour market operates differently in ways that affected both construction and ramp.

Costs are substantially higher than equivalent Taiwanese facilities — construction, labour, and the absence of a nearby supplier and service ecosystem all contribute. Subsidies offset part of this; the operating cost differential persists indefinitely.

Why did Japan go better?

Because Japan retained a genuine semiconductor supply base. Materials suppliers, equipment makers, chemicals producers and skilled industrial workers all exist there, and the manufacturing culture is closer to Taiwan’s in its expectations around precision and continuous operation.

The Kumamoto project moved quickly, benefited from strong prefectural and national support, and had an anchor customer in Sony’s image sensor operations plus automotive demand from Denso and others. It demonstrated that overseas fabs can work where preconditions exist.

The Japanese case also had realistic technology targets, focusing on mature and specialty nodes serving identified local demand rather than attempting to replicate leading-edge capability — a scoping decision that made success far more likely.

What is the German project for?

Serving European automotive and industrial demand with specialty and mature processes, in partnership with Bosch, Infineon and NXP, supported by European Union and German funding under the bloc’s chip strategy.

The rationale is customer proximity rather than technology leadership. European automakers and industrial firms consume large volumes of exactly the mature-node silicon the fab will produce, and they experienced the shortage acutely enough to support local capacity.

The challenges are European: construction costs, energy prices, permitting timelines and a labour market with limited semiconductor manufacturing experience. Dresden has a genuine microelectronics cluster, which helps considerably.

Why does the leading edge stay in Taiwan?

Because it depends on the ecosystem rather than on the fab. Leading-edge manufacturing requires materials suppliers, equipment service engineers, specialized subcontractors, packaging partners and a research organization, all operating within hours of each other and iterating continuously.

That density exists in Hsinchu, Taichung and Tainan and nowhere else at comparable scale. A single fab abroad, however advanced, cannot replicate it, and process development requires the pilot lines, research fabs and engineering population concentrated at home.

The strategic dimension reinforces the technical one. Keeping the frontier in Taiwan preserves the island’s indispensability, which is a national interest as well as a corporate one — the argument examined in the chip geopolitics story.

What do the economics actually look like?

Overseas fabs cost meaningfully more to build and operate than Taiwanese equivalents, with estimates of the gap varying by location and methodology but consistently substantial across construction, labour, utilities and support services.

Subsidies cover construction capital but not ongoing operating cost differentials, which means overseas capacity produces structurally more expensive wafers indefinitely. Someone pays that: customers, the manufacturer’s margins, or continuing public support.

The company’s approach has been to price overseas capacity accordingly and to be publicly clear about the cost reality, which has occasionally created friction with governments expecting price parity alongside supply security.

What does this mean for customers?

Choice and cost. Customers requiring domestic production for regulatory, procurement or risk reasons can obtain it and will pay more; those optimizing for cost will continue buying Taiwanese wafers.

Defence, government and critical infrastructure buyers fall into the first category by mandate. Consumer electronics generally falls into the second, since price competition leaves no room for a geographic premium.

The likely equilibrium is a two-tier market where a modest share of production carries a location premium and the majority does not — which is a smaller change to the industry than the subsidy programmes’ scale implies.

What is the honest assessment?

That geographic diversification is happening, is expensive, is slower than announced, and does not shift the technological centre of gravity. Taiwan will retain leading-edge manufacturing for the foreseeable future because the conditions that produced it exist nowhere else.

The subsidy programmes will produce real capacity that provides genuine insurance against specific disruptions, which has value even if it does not achieve the self-sufficiency rhetoric that accompanied the funding.

For the company, overseas fabs convert geopolitical risk into operating cost and turn it into a stakeholder in every major bloc’s industrial strategy — expensive insurance that is nonetheless cheaper than the alternative of being a single-jurisdiction supplier of an indispensable product.

What does a fab actually need beyond the building?

Ultrapure water at enormous volume, absolutely reliable electricity, specialty gases and chemicals delivered continuously, equipment service engineers available within hours, spare parts inventories, waste treatment capacity and hundreds of specialized subcontractors for maintenance and facility operations.

Each of these exists in Taiwan as a mature industry serving many fabs. In a new location they must be created or imported, and the resulting supply chain is longer, more expensive and more fragile than the one a Taiwanese fab draws on without thinking about it.

The equipment service dimension is particularly underappreciated. Advanced tools require frequent specialist intervention, and the difference between an engineer arriving in two hours and in two days is measured directly in lost production — which is why suppliers cluster around fab concentrations rather than distributing evenly.

How do customers view overseas capacity?

As insurance they will pay a limited premium for. Customers with government contracts, defence exposure or explicit supply chain requirements need domestic production and will accept higher prices; commercial customers optimizing for cost generally will not.

This produces a segmented demand picture where overseas fabs serve a specific slice of the market rather than replacing Taiwanese capacity broadly. Utilization depends on whether that slice is large enough to fill the facility, which is the central commercial question for each project.

Some customers have committed to overseas capacity strategically, accepting cost to support diversification. Whether those commitments survive a competitive downturn, when procurement teams face margin pressure, has not yet been tested.

What happens to the subsidies over time?

Construction subsidies are one-off; operating cost differentials are permanent. This creates a structural question that most programmes have not addressed: whether governments will continue supporting facilities whose economics remain unfavourable after the initial capital grant is spent.

Historical precedent in other subsidized industries suggests that political support weakens as attention moves elsewhere, and that facilities built on grants sometimes struggle when the operating reality becomes apparent. Semiconductor programmes may prove different given security motivations, but the pattern is worth noting.

The alternative is that the market segments permanently, with domestically produced chips commanding a security premium in defined applications. That outcome would make the facilities viable and would represent a real change in how the industry is structured.

What is the talent transfer challenge?

Establishing an overseas fab requires sending hundreds of experienced engineers to train local staff and stabilize processes, which depletes home operations and asks people to relocate their families for years to places where the industry has no existing community.

Local hiring faces a different problem: semiconductor process engineering is a specialized discipline that universities in most countries produce in small numbers, so a new fab competes for a scarce national pool while needing hundreds of qualified staff simultaneously.

This is why fab clusters persist. Once a region hosts several facilities, a labour market forms in which engineers can change employers without relocating, which makes recruitment far easier for everyone and makes the first fab in a new region disproportionately difficult.

How should the subsidy programmes be judged?

Against their actual objective, which is supply security rather than commercial return. Judged as investments they look poor; judged as insurance premiums against a specific disruption scenario they look more defensible, and the appropriate question is whether the premium is proportionate to the risk.

The programmes will produce real capacity, real employment and real capability in the host countries, alongside real ongoing cost. Whether that combination represents good policy depends on assessments of geopolitical risk that reasonable people weigh differently.

What does this mean for the industry’s structure?

Higher costs and more capacity than a purely commercial industry would build, distributed for political rather than economic reasons. Overcapacity in specific segments is a real possibility once subsidized projects complete simultaneously across several jurisdictions.

For customers this means more supply options at differentiated prices, and for the industry it means returns diluted by capacity that would not have been built on commercial logic alone — a cost of the security objective that subsidy programmes rarely acknowledge explicitly.

Frequently Asked Questions