For years, the idea of “bringing semiconductor manufacturing back” to the United States has been sold as a strategic matter: supply chain resilience, national security, and less dependence on Asia. The problem is that, when the conversation shifts from the geopolitical level to the financial details, numbers that are hard to ignore start to surface. And in the chip industry, what’s uncomfortable tends to impact prices, schedules… and margins.
A cost breakdown attributed to SemiAnalysis and focused on 5 nm wafers in 2025 compares two scenarios: a TSMC factory in Taiwan (Fab 18, phases 1–3) versus another in the United States (Fab 21, phase 1). The implicit headline is tough to hide: producing in the U.S. is much more expensive per wafer, and the gross margin per unit is reduced to a fraction.
The comparison that sums up the problem: total cost and profit margin per wafer
The table shows the total cost per wafer as $6,681 in Taiwan, versus $16,123 in the U.S. That’s roughly a 141% surcharge. And looking at the other side of the coin — the gross margin per wafer — the gap widens dramatically: $10,819 in Taiwan compared to $1,377 in the U.S.
Expressed as a percentage, the chart paints an even clearer picture: the gross margin per wafer drops from 62% in Taiwan to 8% in the U.S. In other words, the profitability per unit becomes almost eight times smaller.
It’s important to emphasize one idea: these figures do not describe “the total profitability” of a company (which depends on many business lines and product mixes), but the economics per wafer for a specific node. Still, as an industrial thermometer, it’s a fever indicator.
The silent factor: depreciation per wafer as an industrial “mortgage”
If there’s one part of the breakdown that explains much of the margin hit, it’s the depreciation per wafer. In Taiwan, it appears as $1,500, but in the U.S. it rises to $7,289. The difference isn’t marginal: it’s the kind of figure that completely changes the game.
Simply put, depreciation reflects how the cost of building and equipping a factory (and its machinery) is “spread out” over its lifespan. Here, a crucial nuance comes into play: scale.
The table indicates a very different monthly installed capacity: 90,000 wafers/month in Taiwan versus 24,000 wafers/month in the U.S. While the American CapEx might not be the highest globally, producing far fewer wafers causes the depreciation “share” per unit to skyrocket. Colloquially, it’s like a factory “weighing more” when its costs are divided among fewer units.
It’s no coincidence that the “cost premium” chart shows depreciation as the most extreme jump: +384% compared to Taiwan. It’s the line that turns a wage debate into an industrial structure debate.
Labor: double the cost, and not just because of wages
The second clear factor is labor costs, which go from $1,800 per wafer in Taiwan to $3,600 in the U.S.: a +100% increase.
It’s not just about wages. In a high-end semiconductor fab, labor costs also depend on availability, shifts, intervention times, and operational culture. A quote attributed to Morris Chang, the legendary TSMC leader, illustrates this stark difference: if something breaks early in the morning, one environment acts “immediately,” while in another, they wait for the next shift. Beyond tone, the message is clear: in high-complexity production, minutes and hours matter, and that cost ultimately appears (directly or indirectly) in the cost per wafer.
Raw materials and variable costs: the gap extends beyond personnel
The table also reveals another surprise: raw materials are double, from $1,520 to $3,040 per wafer (+100%). Additionally, variable costs per wafer increase from $5,181 to $8,833 (+71%).
However, not everything increases: utilities stay at $630 in both cases, and consumables at $760. This suggests that the difference isn’t in “turning on the factory,” but in how it’s supplied, operated, and above all, amortized.
There’s also a jump in overhead, from $471 to $803 per wafer (around +70%). Along with depreciation, the message is the same: fixed and organizational costs penalize heavily when the plant isn’t operating at full scale yet.
CapEx: less total spending doesn’t mean cheaper wafers
Another striking fact: the total CapEx shown is $27 billion for Taiwan and $14.38 billion for the U.S. In main manufacturing equipment (wafer fab equipment), $22.95 billion versus $8.6 billion. In other words, U.S. expenditure on equipment is lower in the figures.
But this apparent “advantage” is neutralized by the actual volume. A factory can cost less yet produce more expensive wafers if:
- it produces fewer wafers per month,
- it takes longer to reach the target throughput,
- or has higher operational and maintenance costs.
The industrial economy holds this paradox: cutting project costs doesn’t always lower the product price.
What does this mean for the market?
From here, the debate moves beyond technicalities into commercial considerations: if manufacturing in the U.S. significantly raises wafer costs, the industry must decide which lever to pull.
In practice, several non-exclusive options are usually considered:
- Higher prices for wafers to customers demanding local production.
- Incentives and subsidies to absorb part of the differential (directly or indirectly).
- Scalability improvements: increasing wafer output and utilization to drive down depreciation per unit.
- Production mix adjustments: reserving certain workloads for specific plants based on urgency, logistics, or regulatory requirements.
This isn’t happening in a vacuum: chips aren’t just made to “gain margins,” but also to ensure supply, meet contractual terms, and reduce geopolitical risks. What these numbers show is that such security comes at a cost—and, at least in this snapshot, that cost is high.
The clear takeaway: manufacturing outside Taiwan isn’t just “copy-paste”
The core lesson is inconvenient for those looking for simple headlines. Building advanced capacity in a new country isn’t just about money and machinery; it’s about creating an ecosystem, operational culture, suppliers, talent, and industrial routines. Until that machinery produces at the pace the economy requires, depreciation becomes a heavy burden.
Nevertheless, the global trend toward diversifying manufacturing seems unlikely to slow. Because, in semiconductors, the final calculation isn’t merely financial—it’s strategic. The lingering question is whether the market—clients, governments, and manufacturers—will accept higher wafer costs for “near-shore” fabrication over many years, or if they’ll demand scale results much faster than industry can currently deliver.
Frequently Asked Questions
Why can manufacturing advanced chips in the U.S. be more expensive per wafer?
Because some cost components, especially depreciation per wafer and labor, can be significantly higher if the plant produces less volume and doesn’t yet operate at a comparable utilization to mature fabs.
What does “depreciation per wafer” mean in a semiconductor factory?
It’s the allocation of the factory’s construction and equipment costs over its lifespan, divided across the wafers produced. Fewer wafers mean each “bears” a larger share of that amortization.
Is the additional cost in the U.S. solely due to wages?
No. The breakdown also shows significant differences in depreciation, raw materials, overhead, and variable costs. Wages influence the total, but they don’t explain the full jump.
Can the cost per wafer in U.S. fabs decrease over time?
In theory, yes: increasing volume, improving utilization, stabilizing the supply chain, and optimizing operations can lower depreciation per unit and reduce some associated costs. The key is sustained scale.
Sources:
- SemiAnalysis (“TSMC 5 nm wafer economics, 2025”, November 30, 2025)
- Analyst publication by Jukan on X, citing data compiled by SemiAnalysis