Note: We were able to collect useful data from only 29 companies, covering about XX% of the total global manufacturing capacity in 2024 (SEMI, 2025). Thus, the total emissions and energy consumption of the global semiconductor industry are most likely much higher than displayed in this analysis.
This interactive chart allows users to explore the share of individual companies across four key categories:
By allowing users to browse figures reported by companies with the largest footprint, we intend to showcase the influence these manufacturers have on the overall trends in emissions stemming from semiconductor manufacturing.
Semiconductor GHG emissions and energy consumption are highly concentrated among a hand full of manufacturers, which can be attributed to their large manufacturing capacities and activities in advanced wafer fabrication, where process complexity, fluorinated gas use and energy intensity scale disproportionately with capacity.
Our analysis shows that the three biggest memory producers alone account for XX% of all Scope 1 emissions in 2024, reflecting a significant role of high-volume DRAM and NAND fabrication in driving direct emissions. Memory manufacturing relies on extensive plasma etching and chamber cleaning cycles where process gases with global warming potentials many times higher than CO2 cannot be fully abated using current technologies (Ranganathan et al., 2004). As a result, the highest-throughput fabs contribute disproportionately to industry-wide emissions.
Furthermore, advanced lithography tools, especially EUV, require more process steps and a significantly higher electricity consumption than previous lithography machines (Boakes et al., 2023). At the same time, abatement systems - while essential for reducing emissions - are themselves highly energy-intensive. Roughly 56% of fab electricity is consumed by production equipment, with the remaining 44% used by facility utilities such as chillers and abatement tools (Hess, 2025).
These structural factors explain why just five companies - Samsung, TSMC, Micron, Intel and SK Hynix - account for nearly 69% of all reported energy consumption. Their dominance in mature node and leading-edge logic as well as memory production, likely leads to Scope 1 emissions being sharply clustered among the largest players.
This interactive chart lets users explore how the values reported by semiconductor manufacturers have evolved over the years 2015-2024. Companies are grouped based on similar reported values, enabling more meaningful comparisons among peers. Users can browse data by zooming into each group, and hover over individual points to reveal company-specific information. A menu in the top-left corner lets users switch between key metrics: energy consumption (GWh), Scope 1 emissions (MMTCO₂e), Scope 2 market-based emissions (MMTCO₂e), Scope 2 location-based emissions (MMTCO₂e) and Scope 3 emissions (MMTCO₂e).
Indirect emissions (Scope 2) provide a clear example of how methodological choices can distort reported climate performance in the semiconductor sector. By zooming into the difference between location-based and market-based accounting, it becomes evident why companies' emissions figures often diverge based on/due to their use of RECs, even when their underlying energy use is similar.
Scope 2 reporting in the semiconductor industry remains heavily influenced by methodological ambiguity, particularly when it comes to the distinction between location-based and market-based values. Although most large manufacturers purchase significant amounts of electricity, the lack of standardized reporting framework means that companies apply the GHG Protocol's dual-reporting requirement unevenly or partially, resulting in datasets that are difficult to compare across companies and years (GHG Protocol, 2024).
Location-based emissions, which reflect the real-world carbon intensity of the local grid, are often omitted in favor of the typically lower market-based values influenced by contractual instruments such as renewable energy certificates (RECs). Market-based reporting, however, does not necessarily represent physical electricity generation and can mask continued dependence on carbon-intensive grids. While energy consumption in semiconductor manufacturing increased by nearly 148% from 2015 to 2024, market-based emissions rose by only XX%, a discrepancy that cannot be interpreted without the accompanying location-based figures. Companies purchase large volumes of RECs may appear to perform better than peers operating in the same region with similar energy mixed, while firms that report more transparency – by publishing both metrics or limiting renewable energy claims to verifiable local sources – often show higher Scope 2 emissions on paper.
This discrepancy is particularly evident in data reported by several companies in 2024. Intel had the highest discrepancy, as its location-based emissions were more than 30 times higher than their market-based emissions (Intel CSR Report 2025). This was followed by Infineon and STMicroelectronics reporting location-based emissions that were more than five times higher than their market-based emissions (Infineon CSR Report 2025) (STMicroelectronics CSR Report 2025).
This interactive heatmap aims to shed light on the importance of missing data. By selecting one of the available metrics, users can track the frequency of CSR reporting by each individual semiconductor manufacturer. The chart also highlights sharp increases in reported values (whenever the increase is above 50%), aiming to draw attention to unusual jumps that could indicate changes in reporting practices, data quality issues, or significant operational shifts. The heatmap serves two purposes: it shows where information is missing, ensuring transparency, and it helps track each company's progress towards fully-fledged and accurate tracking of emissions and energy consumption stemming from their operations.
The overall quality of CSR reporting in the semiconductor sector remains highly inconsistent, fragmented and often insufficient for meaningful comparison. A central issue is the lack of a standardised reporting framework, which leads to opacity, inconsistent data and missing metrics across companies and years. Out of 59 major manufacturers we identified globally for the purpose of our spring 2025 publication, usable CSR data could be collected from only 29 companies, covering just XX% of global wafer capacity in 2024 (SEMI, 2025). It is important to note that many companies not always differentiate Scopes 1, 2 and 3 or publish numbers for all dimensions, rely on internal calculation methods, or normalise emissions using their own metrics, making reliable evaluation nearly impossible (Hess, 2025).
Even where data is disclosed, granularity varies widely. For example, Scope 3 data remains largely incomplete, especially for downstream emissions: from a total of 15 reporting categories in Scope 3 emissions – structuring up and downstream activities – a company can decide which of them to report on. They can also choose to report only accumulated data or not report Scope 3 figures at all. One resulting phenomenon is that companies that report more transparently and extensively on Scope 3 have – at least on paper – much higher total GHG emissions.
This becomes evident by considering Samsung's upstream and downstream emissions. On paper, the company has one of the highest reported Scope 3 emissions, exceeding 105 MMTCO2E in 2024 (CSR Samsung Report 2025), which is more than 4 times higher than that of Intel, the next highest emitter with 25 MMTCO2E in 2024 (CSR Report Intel 2025). Samsung reports much higher Scope 3 emissions largely because it includes use-phase emissions – those generated during the lifetime use of its products – within its Scope 3 totals. Intel, on the other hand, does not report those.
Our data brief, TITLE HERE, explores how the global AI boom driving up chip demand may be causing an increase in direct emissions from semiconductor manufacturing.