Google and Amazon published updated sustainability reports this week that show significant year-on-year rises in their greenhouse gas emissions. Google’s total emissions increased by 25% and Amazon’s by 16% compared with the prior year. The reports indicate that the spread of AI workloads and the infrastructure built to support them have made meeting net-zero commitments materially more difficult.
What is behind the rise?
Neither company explicitly says that AI alone is responsible for the higher emissions, but indirect evidence in both filings points toward AI-related activity. Both Google and Amazon acknowledge their energy use grew substantially over the past year, in parallel with rapid adoption of AI services and models.
The reports refer to carbon intensity—the emissions per dollar of revenue—as a useful metric to track how emissions scale with economic output.
Scope 3 emissions are the main driver
A large portion of the increase comes from Scope 3, or indirect, emissions. Scope 3 includes emissions tied to goods and services a company purchases and to emissions that occur when customers use sold products. Google reported that its Scope 3 emissions rose by 2.1 million tonnes last year and are now twice the level of 2019—2019 being the baseline year Google uses to evaluate performance.
Amazon’s Scope 3 growth is driven substantially by capital goods and by fuel and energy-related activities. Capital goods can include data centers and warehouses. In its report Amazon noted that in 2025 it added more new global data center capacity than any other company, including more than 1.2 gigawatts commissioned in the fourth quarter of 2025. Those investments help explain why Amazon’s Scope 3 emissions rose more sharply than Google’s, though data center expansion is likely a major factor for both companies.
Google groups some Scope 3 categories together and says emissions from use of its sold hardware are not material—most of its hardware are low-power devices—suggesting data centers are the principal source.
The practical and supply-chain challenges
On the energy procurement side, years of renewable energy purchases have so far helped limit operational emissions. However, meeting sharply higher electricity demand for AI has prompted some firms to invest in gas-fired generation or otherwise rely more on fossil fuels in certain contexts, complicating net-zero plans.
A harder-to-address problem lies in emissions embedded in building and equipping data centers. Steel and cement production are carbon-intensive industries; while lower-carbon processes are under development, they are not yet available at the industrial scale required by hyperscalers.
Chip demand—GPUs and memory used to train and run AI models—also adds emissions. Semiconductor fabrication is energy-intensive, and many advanced fabs are in Asia where grids still rely heavily on fossil generation. Some chemicals used in chip manufacturing are potent greenhouse gases with warming impacts many times that of CO2 on a per-tonne basis.
The surge in chip purchases is therefore likely to have materially increased the carbon footprints of both Amazon and Google.
Options to meet net-zero commitments
None of these challenges are insurmountable, but they require large-scale, coordinated action. To meet their net-zero targets the companies will likely need to:
- scale up actual supply of additional renewable electricity tied to their operations;
- invest in low-carbon steel and cement production technologies at industrial scale; and
- purchase very large amounts of carbon removal capacity or credits.
These steps will entail substantial cost, but are feasible. Rapid, broad adoption of AI has unquestionably made the path to net-zero more difficult and costly for major cloud providers.
Conclusion
The sustainability disclosures from Google and Amazon make clear that AI-driven growth and data center expansion are significant contributors to rising emissions. Both companies still commit to net-zero goals, but rising Scope 3 emissions from data centers, hardware supply chains and chip manufacturing mean they face sizable and potentially expensive steps to achieve those commitments.
