Google and Amazon emissions surge as AI infrastructure demand peaks
Google and Amazon reported significant year-over-year increases in carbon emissions, reaching 25% and 16% respectively. These figures are driven primarily by the massive energy and infrastructure requirements of data centers supporting AI and high-performance computing services.
Key Takeaways
- Google observed a 25% year-over-year emissions increase, with Scope 3 pollution double its 2019 baseline.
- Amazon emissions rose 16% to roughly 81 million metric tons of CO2 equivalent in 2025.
- Amazon added more than 1.2 gigawatts of data center capacity in Q4 2025 alone to meet AI demand.
- Both companies are increasingly turning to natural gas power plants to stabilize energy needs for AI facilities.
Why It Matters
The rapid expansion of AI compute is creating a fundamental conflict between cloud scaling and environmental mandates. For the streaming industry, which relies on these hyperscale providers for encoding and delivery, the rising carbon intensity of the underlying stack may impact corporate sustainability disclosures and green-tiering initiatives. As big tech shifts from renewable-only strategies to a mix including fossil fuels to maintain uptime, video engineers must weigh the cost of AI-driven optimization against their own ESG commitments. Watch for a divergence in cloud pricing as providers pass through the high costs of carbon removal credits needed to offset these infrastructure spikes.
Additional Context
The environmental strain is not limited to power consumption; it begins deeper in the supply chain with silicon manufacturing. According to 2026 reporting from TechInsights, semiconductor emissions are projected to rise 9% this year, driven by the energy-intensive fabrication of 2nm and 3nm logic chips. The manufacturing of High-Bandwidth Memory (HBM) used in AI accelerators is proving particularly problematic due to yield losses in complex 16-high stacks, which require more frequent etching and result in higher fluorinated gas usage. Per TechInsights, emissions from the memory sector are currently growing faster than silicon shipment volumes. Energy procurement has emerged as the primary bottleneck for hyperscale growth. Per The Tech Capital, Amazon added 4 gigawatts of capacity in 2025 but faces grid interconnection delays across Europe that now set an upper bound on regional expansion. To circumvent grid wait times—which can reach five years—providers are moving "behind the meter." Global Energy Monitor reported in January 2026 that proposed natural gas-burning facilities in the U.S. tripled in 2025, with more than one-third of this new capacity planned specifically to power on-site data centers. Google confirmed a partnership for a 933-megawatt natural gas plant in Texas to support its "Goodnight" data center campus, marking a significant strategic pivot from its prior grid-plus-renewables model. Simultaneously, Nvidia—the primary hardware provider for these builds—disclosed that its own total greenhouse gas footprint grew nearly 87% in fiscal year 2025. According to Greenpeace East Asia in April 2026, 99.8% of Nvidia's emissions fall under Scope 3, largely concentrated in its East Asian supply chain. While Nvidia's Blackwell architecture is marketed as 25 times more energy-efficient than previous generations per unit of work, the sheer volume of global deployments is resulting in a net increase in absolute emissions across the entire AI ecosystem.
Read full article at techcrunch.com
Get this in your inbox → Subscribe
Enjoy our coverage?
Add StreamingMeme as a preferred source on Google to see more of our streaming news at the top of your Search results.
Add as preferred source