Full Breakdown
AI-Driven Emissions Surge Among the World’s Largest Cloud Providers
7/12/2026, 11:25:55 AM
Core Event
In the 2025-2026 reporting period, Microsoft, Amazon and Google collectively emitted 119 million metric tonnes of carbon-dioxide-equivalent (mTCO2e), roughly one-third of France’s annual emissions. Their emissions rose from about 101 mTCO2e the prior year, marking increases of 25 % for Microsoft, 18 % for Google and 16 % for Amazon. The growth is attributed primarily to the rapid expansion of AI-focused data-centre infrastructure and associated supply-chain activities.
Background & Context
The surge follows a multi-year boom in demand for cloud services that support generative-AI model training and inference. Since 2020, the three firms have pledged net-zero or carbon-negative targets—Microsoft and Google by 2030, Amazon by 2040—but the AI data-centre build-out has reversed earlier emissions-reduction trends.
Data & Statistics
- Total emissions: 119 mTCO2e (2025-2026) vs 101 mTCO2e (2024-2025).
- Company-specific jumps: Microsoft + 25 % (to 20 mTCO2e), Google + 18 % (to 21 mTCO2e), Amazon + 16 % overall (supply-chain emissions + 20 %).
- Scope breakdown: For Microsoft, Scope 2 electricity-related emissions accounted for 13 % of total; over 40 % of Microsoft’s and Google’s emissions stem from “capital goods” (new construction and fixed assets). Amazon’s largest scope-3 driver is “Purchased Goods and Services.”
- Energy use: Microsoft now consumes roughly four times the pre-pandemic electricity; Google’s electricity use has more than doubled. Estimated Amazon data-centre consumption in 2025 was 78,000 GWh, comparable to Microsoft and Google combined.
Official Statements & Responses
Microsoft’s sustainability report attributes the rise to “the expansion of our datacenter infrastructure.” Google cites “increases in supply chain activities that supported the rapid expansion of our business.” Amazon frames its 16 % rise as “making progress” toward its 2040 net-zero pledge. All three companies reaffirm their long-term climate goals and note ongoing investments in renewable energy, carbon-free electricity and carbon-removal programs.
Criticism & Opposition
Economics professor Cecilia Rikap (University College London) argues that the firms’ “claims … about their clouds being ecologically friendly and sustainable are a marketing strategy,” warning that cloud migration enables other corporations to “obscure their environmental footprint.” Electrical-engineering professor Shaolei Ren (University of California, Riverside) emphasizes that “the increases in total carbon emissions are strongly correlated with [the companies’] AI investment,” suggesting that current pledges may be untenable without deeper systemic changes.
Verbatim Quotes
- “Cecilia Rikap, an economics professor at University College London, said: “Claims by Microsoft, Amazon and Google about their clouds being ecologically friendly and sustainable are a marketing strategy.” — Cecilia Rikap, economics professor, University College London
- “The increases in total carbon emissions are strongly correlated with [the companies’] AI investment,” — Shaolei Ren, professor of electrical engineering, University of California, Riverside
- “driven primarily by the expansion of our datacentre infrastructure” — Brad Smith, vice chair and president, Microsoft
- “The search company says its AI systems have come up with solutions that have already helped to reduce emissions elsewhere by 41m tonnes of CO 2 last year.” — Google sustainability report
- “making progress” — Amazon sustainability report
Conflicting Reports & Gaps
Microsoft’s figures cover its fiscal year ending 30 June 2025, whereas Google and Amazon report calendar-year data. This timing mismatch complicates direct year-over-year comparisons. Additionally, Amazon does not publish recent electricity-usage data, leaving a reliance on third-party estimates for its power consumption.
What’s Next
All three firms have outlined continued investment in renewable-energy contracts, battery storage and carbon-removal technologies to meet their 2030-2040 climate targets, though specific timelines for scaling AI-efficient hardware or reducing scope-3 emissions were not detailed in the released reports.
