Nvidia’s supply chain emissions surged 725% since 2020, now rival a Russian coal producer

Nvidia’s meteoric rise as the engine behind the AI revolution comes with a carbon receipt that keeps getting longer. A coalition of environmental and civil society organizations, including Greenpeace International, has published a report showing the chipmaker’s supply chain emissions have climbed 725% since fiscal year 2020.

The numbers are hard to ignore. Nvidia’s Scope 3 emissions, which cover the broader supply chain rather than just its own offices and factories, grew from 1.3 million tonnes of CO2 equivalent in fiscal 2020 to approximately 10.7 million tonnes in fiscal 2026. That’s an additional 9.4 million tonnes of carbon over six years, driven almost entirely by the insatiable global appetite for GPUs and AI training infrastructure.

The emissions Nvidia doesn’t report

The supply chain figures are only part of the story. Climate analyst Ketan Joshi, who led the research, estimates that downstream emissions from the actual use of Nvidia’s sold chips could reach as high as 21 million tonnes of CO2 equivalent by 2025. That’s nearly double the supply chain emissions themselves.

To put that in perspective, the report compares that figure to the reported emissions of Russian Coal, one of Russia’s major coal producers.

Nvidia has been quick to highlight its own operational achievements. The company says it has used 100% renewable electricity for its operations for two consecutive years, which translates to zero market-based Scope 2 emissions. But critics in the report argue this framing amounts to greenwashing, because the vast majority of Nvidia’s environmental impact doesn’t come from its own buildings. It comes from the sprawling network of energy-hungry data centers that its hardware powers.

Why the gap keeps widening

Nvidia’s own sustainability disclosures show the acceleration in real time. Scope 3 emissions jumped from 3.64 million tonnes in fiscal 2024 to 10.7 million tonnes in fiscal 2026. That’s nearly a tripling in just two years, coinciding with the explosive demand for AI training and inference compute that followed the launch of ChatGPT and the broader generative AI arms race.

The manufacturing process itself is carbon-intensive. Cutting-edge semiconductor fabrication, predominantly handled by TSMC in Taiwan, requires enormous amounts of energy, ultrapure water, and specialty chemicals. As Nvidia’s order volumes have scaled to meet AI demand, the emissions embedded in that manufacturing have scaled proportionally.

What this means for investors and the broader market

The report also complicates Nvidia’s narrative that AI will be a net positive for climate action. CEO Jensen Huang has repeatedly argued that AI-driven efficiency gains across industries, from energy grid optimization to materials science, will more than offset the energy required to run the models. The environmental groups behind this report are essentially calling that claim into question, arguing that you can’t credibly position yourself as a climate solution provider when your products’ emissions rival those of a coal company.

The International Energy Agency has flagged AI-driven data center growth as a significant factor in its energy forecasts.

Nvidia has not yet issued a detailed public response to the report’s specific claims.

Disclosure: This article was edited by Editorial Team. For more information on how we create and review content, see our Editorial Policy.