Semiconductor Overcapacity Talks Face a Double-Counting Risk · Traverse Analysis
Semiconductor Overcapacity Talks Face a Double-Counting Risk
Semiconductor capacity figures can count one factory in several chip categories. The new sectoral talks must distinguish those capabilities before drawing conclusions about excess supply.
The United States wants trading partners to help address industrial overcapacity while it investigates some of those same partners for practices that could warrant U.S. tariffs. The European Union, Japan, Korea and Mexico appear in both the and . Agreeing on what counts as excess capacity will not settle that conflict, but it could narrow the factual disputes on which a joint response depends.
The new semiconductor overcapacity talks face a counting problem before governments agree on a response. The OECD maps a factory's full capacity under each chip category it can produce. Those capability figures overlap. Adding them would overstate physical capacity, potentially distorting the comparison with demand that a common policy would rely on.
The United States and 14 partners have committed to technical meetings before December 2026, including work on a foundational-semiconductor platform. Their October statement calls for nonconfidential information, identification of data gaps, and use of OECD work. The statement specifies neither a common capacity test nor a company submission procedure.
A semiconductor company's trade-policy director can use this interval to prepare a technical evidence brief for officials. It should specify the product market behind the company's claim and explain which capacity belongs in that comparison. Section 301 does not prescribe this particular measurement method.
One fab can appear in several chip categories
Steel offers experience in coordination, but no ready-made semiconductor test. The steel forum's September framework builds on principles adopted in 2017 and an annual information-sharing exercise. Its members share an interest in sustaining market-based steel production, although the framework allows their responses to differ with national circumstances. Semiconductor negotiators still need to establish which products compete and which capacity serves them. A common forum cannot make overlapping capability figures additive.
The OECD's December 2025 capacity study deliberately maps manufacturing capabilities and discloses the overlap. This is not an error in its overall count of physical capacity. The underlying commercial datasets also lack consistent allocation by process technology. Even matching node sizes, measured in nanometers, do not establish that two lines make substitutable chips.
For negotiators, the distinction changes what a table can prove. A capability chart identifies factories with the relevant manufacturing technology. It does not establish how much of that chip is being produced, how much unused capacity could serve that market, or whether other listed capacity would compete for the same orders. A policy brief that silently moves between those questions can turn a useful industrial map into an unsupported estimate of excess supply.
A hypothetical fab-capacity return, as of November 10, 2026. This illustration applies the OECD distinction between capacity and capability to one factory. It assumes a fixed total of 10,000 wafer starts a month and both analog and specialty-memory capabilities. The figures describe no actual company or economy, and product allocation remains unknown.
Reported item
Figure
What it establishes
Physical fab capacity
10,000 wafer starts a month
The facility's stated production ceiling
Capacity at fabs capable of analog production
10,000
Analog capability exists at this facility
Capacity at fabs capable of specialty-memory production
10,000
Specialty-memory capability exists at this facility
Output allocated to either chip category
Unreported
No category-level output total follows from the capability entries
Adding the two capability entries produces 20,000 on paper. It creates no additional physical production. The appropriate response is to retain the shared-fab identifier and mark the missing allocation. Reclassifying the entire amount as harmless spare capacity would be equally unsupported.
Define the market before comparing capacity
The platform has a starting point. The OECD's 2024 semiconductor taxonomy distinguishes product types, fabrication capabilities, installed capacity, location, ownership and operational status. Using that structure would allow officials to specify what a number represents before debating its policy significance.
The evidence brief should then make the comparison explicit. If the claim concerns analog competition, an all-semiconductor capacity total is background rather than the measure of that market. If it concerns production expanding ahead of demand, both sides of the comparison need compatible product definitions and periods. A broad label such as foundational semiconductors cannot by itself supply those definitions.
Operating status needs similar discipline. The OECD's 2025 study treats future fabs as uncertain and describes its data as a September 2025 snapshot. It also warns that incomplete Chinese coverage can understate capacity. These limitations do not establish the direction of the overall error in a particular market. They require the brief to distinguish measured production, estimated capacity and announced expansion, with dates attached to each.
That distinction would preserve two different questions for negotiators. Existing output can be compared with current demand. Planned investment can be assessed against an explicit future-demand assumption. Combining them into one present-tense surplus would conceal the timing assumption driving the result.
The government practice still has to be identified
Under Section 301(b), USTR considers whether the foreign practice is unreasonable or discriminatory, burdens or restricts U.S. commerce, and warrants action. Product detail can strengthen the explanation of competitive effects. The legal question remains distinct from the platform's data design, just as coalition membership remains distinct from resolution of a country's investigation.
The OECD's separate industrial-subsidy study leaves excess-capacity estimation outside its empirical analysis, citing difficult measurement and limited comparable utilization and price data. It also recognizes that some spare capacity is normal. An evidence brief therefore needs a causal account alongside its capacity exhibit. It should identify the government measure, explain how it affects production or investment, and state the evidence connecting that effect to the claimed commercial burden. Uncertainty in one link should remain visible rather than being filled by a country ranking.
Share the method without overstating the evidence
The October commitment to nonconfidential exchange makes disclosure design part of the technical task. USTR's investigation already has procedures for business-confidential submissions and accompanying public versions. The joint statement does not establish an equivalent platform procedure or provide for sharing confidential docket submissions. The investigation's original comment deadline has passed. The platform announcement does not reopen it.
A company can prepare a public explanation of its definitions, aggregation choices and missing fields without claiming that the resulting table proves more than it does. Officials may hold additional confidential evidence. The public analytical boundary is narrower, and the terms of reference should make clear which conclusions shared data can support.
A shared method could help participants assess a capacity claim even when they disagree about tariffs. It cannot decide which partners face U.S. action or how the costs of cooperation should be divided. For a semiconductor company, the useful contribution is narrower. Its brief should show which capacity actually competes in the market at issue, identify any government practice sustaining it, and leave room for ordinary commercial expansion. Without those distinctions, a joint response risks protecting producers from competition that the capacity figures never showed to be distorted.
Daily Trade Brief
Get the next one in your inbox.
New U.S. trade actions and the analysis that changes an import, customs, or counsel file, on weekday mornings after you confirm your address.
From reading to review
Run the numbers on your lane.
The duty calculator runs the current stack for any HTS code and origin. A free account opens full tool output, AD/CVD detail, Chapter 98 processing, and available exports.