The chip landscape

2. Data

This paper presents the first analysis based on the OECD Semiconductor Production Database. The database is being developed based on the principles elaborated by the Network (OECD, 2024[1]) that are reproduced in Annex A. This database builds on semiconductor production capacity information available from the industry and commercial data providers, augmented by information on other relevant dimensions that could help meet the goals described in section 1. Data are aggregated at the economy level according to different dimensions (e.g., by chip types and/or specific technologies). Data aggregation is required to ensure co-operation with industry and commercial data providers.

Available data sources

This paper focuses on the wafer fabrication stage of the semiconductor value chain, also referred to as front-end manufacturing, which involves the deposition, patterning, and etching of materials to build an integrated circuit on a wafer, based on in the semiconductor taxonomy developed by the OECD (2024[1]). Thus, only front-end fabs have been included in the analyses presented in this paper. The reason for this initial focus is that front-end manufacturing is the most capital-intensive process step of semiconductor production, and it has received a lot of attention from government policies in recent years. The OECD plans to expand the scope of its Semiconductor Production Database to potentially include back-end manufacturing (assembly, test and packaging) and semiconductor manufacturing equipment.

Definitions

The production capacity of a fab is measured in wafer starts per month (WSPM) in 8-inch (8’’) equivalents (EQ). WSPM denotes the maximum number of silicon wafers that can be processed in a fab per month. WSPM is independent of the type of chip (e.g., memory, logic, power, or analog), the yield (share of functioning chips per wafer) or the die size. Since fabs process a variety of different wafer diameters, such as 6” (150mm), 8” (200mm) and 12” (300mm), the capacity is typically normalised across wafer diameters by using 8” equivalents. Unless explicitly stated otherwise, all wafer capacity calculations in this paper are given in WSPM in 8’’ equivalents.

The database also tracks the status of a fab. Following is the definition of the different status categories for all fabs in the database, in chronological order of the timeline until a fab reaches full production:

• Planned: includes fabs whose planning has been mentioned by the company or that have been officially announced. Importantly, this category may include fabs whose construction is planned to start in five or even more years and fabs whose construction will start the next quarter. In the OECD Semiconductor Production Database, 54 out of a total of 1 433 fabs fall under this category.

• Under construction: Fabs are categorised as “under construction” if they (facilities, cleanrooms, etc.) are currently being built or if they have been built and are currently being equipped with machines. In this vintage of the OECD Semiconductor Production Database, 53 fabs are currently under construction.

• Upcoming: This paper refers to fabs as “upcoming” if they are either “planned” or “under construction”.

• In-production: Fabs that are currently producing chips, including the “first silicon” run1, fall under this category. 1 326 fabs out of 1 433 fabs are considered to be in-production for the purposes of analyses presented in this paper.

The analysis includes R&D and pilot lines (145 out of 1 433 fabs), unless explicitly stated otherwise, but excludes fabs involved solely in epitaxy or back-end manufacturing. Micro Electro-Mechanical Systems (MEMS) fabs have been included in the database. While MEMS themselves are not semiconductors many fabs that are capable to produce MEMS are also producing semiconductors. Furthermore, MEMS devices are increasingly integrated with semiconductor devices already during wafer fabrication (Samudrapom Dam, 2023[4]). Analyses on other stages of the semiconductor value chain (design; assembly, testing and packaging) are beyond the scope of this paper.

The analyses in this paper build on highly disaggregated data at the level of the semiconductor fab sourced from two commercial data providers, SEMI (World Fab Watch third quarter 2025) and TechInsights Inc. (300mm Watch September 2024 and Semico Fab Database Summer 2023), and complemented with desk research, described in this section. For further discussion of all data sources please see Annex B.

Data limitations

Currently commercially available datasets about wafer fabrication come with several substantial limitations, requiring data to be interpreted with caution. While addressing the current shortcomings will take considerable time and effort, the findings described in this paper are a first step in an iterative process of constant improvement. Until recently, quantitative data about the global semiconductor industry was collected and prepared by the industry for the industry, with little consideration for aspects relevant to policymakers, such as geographical diversity of manufacturing. Increased interest by the policy making community in semiconductor data will also support the efforts of the Network towards developing a database with the purpose of addressing policy questions.

Assessing and meaningfully comparing the geographic distribution of wafer fabrication capacities and modes of production faces several challenges. These stem largely from the limitations of currently available data sources as well as general technology trends. As mentioned before, wafer starts per month (WSPM) serves as the primary metric for assessing manufacturing capacity in commercial datasets. Given the reliance on imputed data and estimates from data suppliers, uncertainty analysis (i.e. confidence intervals) is not meaningful as the data reflects industry-informed assumptions rather than direct observations. These educated guesses are standard in industry practices but do not allow for confidence intervals or variability analysis tied to a true underlying value.

The level of aggregation of economies presented in the paper is based on relevant focus on the most concentrated region. Further disaggregation is only possible in the case that the data agreements with the commercial data providers is not violated which is reliant on any disclosure of firm level information.

At the time of writing, WFW and World Fab Forecast (WFF) from SEMI are the most comprehensive, commercially available datasets of front-end fabs (SEMI, 2024[5]). WFW currently lists more than 1 500 fabs with several data points each. Given their comprehensiveness, these commercial databases are currently the primary data sources used for the analysis in this paper. As such, the findings described in this paper rely largely on the accuracy, completeness and correctness of these databases. While they are valuable sources, like any established database evolved over many years, they may have certain limitations.

The following is a brief overview of some of the key constraints of the analysed data and how they affect conclusions drawn from the figures.

When interpreting the figures and analyses presented in this paper, several important data limitations should be taken into account. These limitations arise from inconsistencies across commercial datasets, missing detail on process technologies and materials, incomplete geographic coverage, and uncertainty surrounding future developments.

• Process node density: Inconsistent and incomplete data on process node density (feature size) per fab limit the ability to accurately assess the geographic distribution and substitutability of specific manufacturing capabilities.

• Process technologies: Commercial datasets do not consistently capture or disaggregate process technologies (e.g. BCD, RF, analog) per process node and fab, limiting meaningful analysis of technological specialisation and substitutability across fabs.

• Wafer materials: While commercial data sources often identify whether fabs can use different wafer materials beyond silicon, such as silicon-carbide (SiC) and gallium-nitride (GaN), they lack information on the production share by material type, hindering assessment of material-specific capacity and capabilities.

• Geographic coverage: Data coverage is uneven across economies, especially for China, where limited transparency and delayed reporting result in likely underestimation of actual wafer fabrication capacity.

• Developments in semiconductor technology: The growing relevance of advanced packaging and the resulting integration of front- and back-end manufacturing processes blurs traditional capacity metrics and renders assessments solely focused on wafer fabrication capacity increasingly incomplete.

• Uncertainty of upcoming fabs: Reported projections for future wafer capacity are uncertain, as planned facilities face delays, downsizing, or cancellation, meaning current datasets represent only a temporary snapshot of an evolving landscape.

Data limitations are discussed in greater detail in Annex B.

Note

← 1.  “First silicon" refers to the initial wafers processed through the full production flow to validate equipment, processes, and integration, also commonly called a "fab qualification run" or "pilot run".