Chip manufacturing at a sustainability crossroads

Stephen Russell, Senior Technical Fellow, Sustainability, TechInsights, explains how the semiconductor industry is facing a critical sustainability inflection point with the growing demand and complexity of integrated circuits (ICs). Success will require making sustainability a core business and manufacturing priority rather than a constraint.

Artificial intelligence (AI) is driving one of the biggest technology booms in decades. But behind the rapid growth of AI systems lies a difficult reality: the semiconductor devices powering this revolution are becoming increasingly resource-intensive to manufacture, creating a sustainability challenge that the industry can no longer treat as secondary.

From advanced logic chips to high-bandwidth memory (HBM), the newest generation of semiconductors requires more process steps, more materials, and significantly more energy than previous generations. As the industry pushes toward smaller process nodes and more complex packaging architectures, manufacturing emissions are rising alongside performance gains.

This challenge is explored in depth in a recent sustainability e-book from TechInsights that examines how semiconductor manufacturing complexity is reshaping the industry’s environmental footprint, and what companies can do about it.

Chip complexity drives up carbon consequences
One of the clearest examples is the transition from older manufacturing nodes, such as 90nm technology, to leading-edge 2nm production. Under comparable assumptions for wafer size, fab location, capacity, utilization, and abatement, TechInsights analysis indicates that as nodes approach 2nm, total manufacturing emissions can be roughly three times those at 90nm, while holding wafer size, location, capacity, utilization, and abatement constant. That gap is driven by process complexity and the energy demands of the leading-edge tool chain. While modern chips deliver dramatically improved computing power and energy efficiency during operation, the manufacturing process itself becomes substantially more carbon intensive. Advanced nodes demand far more lithography, deposition, etch, and inspection steps. Each additional process consumes energy, chemicals, ultrapure water, and specialized materials (Figure 1).

AI and the HBM boom
AI is not just creating demand for more chips; it is changing the definition of what a chip is. Modern AI accelerators are multi-die systems that combine reticle-sized logic dies, HBM stacks, and advanced packaging. Each additional die, stack, and interconnect layer multiplies embodied emissions. In fact, leading-edge accelerators are expected to integrate roughly 250 HBM dies by 2030, making memory the dominant source of embodied carbon in many AI hardware systems. The industry is facing a paradox: chips are becoming more efficient in use, yet more environmentally demanding to produce.

At the same time, demand is growing faster than efficiency gains can offset. AI infrastructure expansion, data center growth, and increasing global semiconductor consumption continue to drive overall emissions upward. According to TechInsights modeling cited in the e-book, by 2030, manufacturing emissions from AI GPU production are projected to rise more than twelvefold, from about 1.8 million MT CO2e in 2025 to 21.6 million MT CO2e (Figure 2).

Policy and reporting expectations are tightening
The growing complexity is changing how companies approach sustainability reporting. Historically, environmental reporting in semiconductors focused on broad facility-level metrics. Today, regulators, customers, and supply-chain partners increasingly require far more granular data. Companies are now being asked to understand emissions at the product, process-node, and supplier level. This is especially the case in Europe, where reporting requirements are raising the bar for climate disclosure and supply-chain transparency. In the United States, California climate disclosure rules are increasing pressure on large companies to account more rigorously for emissions and climate-related risk.

This is particularly important because semiconductor supply chains are global. A chip designed in the United States may be fabricated in Taiwan, packaged in Malaysia, and integrated into products sold in Europe. As environmental regulations tighten, especially within the European Union, companies throughout the entire supply chain are being pushed to provide more detailed emissions data, even if they are not directly located within those regulatory regions.

Scope 3 emissions, which are indirect greenhouse gas emissions produced throughout a company’s supply chain, spanning from raw material extraction to the end-of-life treatment, have become especially challenging. Tracking the carbon footprint of materials, manufacturing steps, and outsourced suppliers across multiple countries requires far greater visibility than many companies previously maintained. Teams now need credible Scope 3 answers that can support design choices, procurement conversations, customer requests and emerging regulatory expectations.

Figure 1: Sustainability matrix for 3nm logic illustrating processing carbon hotspots. Scope 2 is concentrated in lithography.

Better data is the foundation for lower-carbon manufacturing
To make those decisions, companies need more than facility-level averages. TechInsights’ EcoInsights platform was developed to help semiconductor companies better understand the environmental impact of semiconductor manufacturing. The platform provides detailed analysis of manufacturing carbon, process complexity, and supply chain considerations, helping organizations make more informed sustainability decisions.

A key focus is identifying “high leverage” opportunities where companies can reduce emissions most effectively. That may include process optimization, material choices, manufacturing location decisions, or supplier selection. By improving visibility into the environmental impact of specific manufacturing flows, companies can move beyond generalized sustainability goals toward more actionable strategies.

Figure 2: AI GPU Manufacture Emissions Forecast, emissions plotted to left y-axis, shipments to right y-axis. Manufacturing emissions rise significantly faster than shipments.

Figure 3: Comparative Analysis: SK hynix HN8T374ZJKX141, Samsung KLUGGARHHD-B0G1, and KIOXIA K1C5233. Source: Screenshots from TechInsights’ Carbon Analyzer Module, 2024.

The EcoInsights modeling suite (Figure 3), for example, quantifies embodied emissions and water use across wafer fabrication, grounded in process flows, equipment sets, and region-specific electricity factors. It helps teams answer questions like:

• Which tool families dominate emissions for a given node and device type?
• How much of the footprint is electricity (Scope 2) versus process gases (Scope 1) versus materials and chemicals (Scope 3)?
• How sensitive is the outcome to fab location, utilization, and abatement efficiency?
• Where will the hotspots move as the industry advances from 3nm to Angstrom-class nodes, or from 200-layer to 800-layer 3D NAND?

AI itself can be part of the solution
While AI is currently a major driver of semiconductor demand and energy consumption, many industry experts believe AI technologies could eventually help improve sustainability practices across manufacturing and infrastructure management. AI is already being explored for applications such as weather prediction, infrastructure optimization, and industrial efficiency improvements, areas that could have meaningful long-term sustainability benefits.

AI-assisted tools can also help companies manage the complexity of the data problem. Helping to identify proxy data, flag emissions hotspots, compare scenarios, and support faster analysis across complex manufacturing and supply-chain datasets. TechInsights is developing AI agents designed to help customers navigate the enormous volume of semiconductor manufacturing and sustainability data available today. These tools aim to help users identify proxy data, uncover patterns, and better analyze environmental impacts across complex supply chains.

Figure 4: Turning pain points into measurable actions.

Conclusion
The semiconductor industry’s sustainability challenge is unlikely to disappear anytime soon. Manufacturing complexity will continue increasing and AI demand shows little sign of slowing. But improved data visibility, more precise reporting, and emerging AI-assisted optimization tools may help the industry begin addressing a problem that is becoming impossible to ignore.

The first step is turning pain points into measurable action (Figure 4). Identify the hotspot, compare how outcomes change with location, yield, configuration, and sourcing choices, and then take action. Prioritize the decisions where carbon concentration and decision control overlap. The companies that make the most progress will be those that connect carbon data to real manufacturing and sourcing decisions. That means identifying emissions hotspots, testing how outcomes change by location, yield, configuration, and supplier, and prioritizing areas where teams have both high carbon impact and real decision control.

For a deeper look at the data, trends, and solutions shaping semiconductor sustainability, readers can explore the full e-book from TechInsights here: https://www.techinsights.com/carbon-age-ai-chips-earth-day-ebook.