Google this week released its 11th annual Environmental Report, disclosing that operational emissions fell 2% in 2025 even as electricity demand climbed 37% year over year, while supply chain emissions tied to its AI infrastructure buildout grew 25% over the same period.

The report, published on the company's Keyword blog and authored by Chief Sustainability Officer Kate Brandt, covers the 2025 calendar year and marks the clearest disclosure yet of the gap between Google's operational efficiency gains and the upstream emissions generated by the hardware, construction, and Asia-Pacific manufacturing that support its expanding data center footprint. Brandt frames the tension directly in the document, writing that the path to the company's climate ambitions "will not be linear" because AI infrastructure construction is accelerating faster than electrical grids are decarbonizing.

7f477eb723fe0c23d03f94b90a08882b9f28187d
7f477eb723fe0c23d03f94b90a08882b9f28187d.pdf
17 MB
download-circle

Clean energy contracts reach 12 gigawatts in a single year

According to the report, Google signed agreements for more than 12 gigawatts of net-new clean energy in 2025 alone, a figure the company says is roughly enough to power a country the size of Greece for a year once the projects become operational. That single-year total builds on a longer accumulation. From 2010 through 2025, Google signed more than 240 agreements covering nearly 35 gigawatts of net-new clean energy, according to the report. Combined, those contracts add enough generating capacity, the company states, to power more than 28 million American homes, a figure it compares to every household in New York, Texas, and Pennsylvania combined.

Those numbers sit inside a broader industry context that PPC Land has tracked closely. Google signed its first US nuclear deal for data centers in August 2025, a power purchase agreement with Kairos Power and the Tennessee Valley Authority tied to the Hermes 2 demonstration reactor in Oak Ridge, Tennessee. That deal, alongside the 12 gigawatts disclosed in the new report, illustrates how quickly the company has needed to diversify its clean energy sourcing beyond wind and solar contracts alone.

Google states in the report that its data center infrastructure uses 83% less overhead energy than the industry average, a figure the company presents alongside its clean energy procurement as evidence that new capacity is being used efficiently. The company also reports matching 100% of its electricity consumption with renewable energy purchases for the ninth consecutive year, even as electricity demand tied to data center operations rose 37% year over year, which the report describes as the company's largest load growth in its history.

Operational emissions fall despite record demand growth

Google reports that it reduced operational emissions by 2% year over year in 2025, a figure the company presents as evidence that operational growth can be decoupled from carbon output. The reduction covers Scope 1 and Scope 2 emissions from the company's own operations, distinct from the Scope 3 supply chain emissions addressed separately in the report.

The company attributes the reduction to a combination of factors. According to the report, advances in machine hardware efficiency, software and compute efficiency, and clean energy procurement collectively avoided more than 58 million metric tons of carbon dioxide equivalent in 2025. Google states that without those interventions, its ambition-based carbon footprint for 2025 would have been five times larger than what the company actually recorded.

The company frames this avoided-emissions figure against its own history. Google states that in 2025, it contracted for eight times more clean energy than it did in 2019, while the emissions the company says it avoided in 2025 were seven times larger than its 2019 ambition-based carbon footprint, which the report identifies as the baseline year for its climate moonshot targets.

Supply chain emissions climb 25% on Asia-Pacific grids

The report's most consequential disclosure for anyone tracking the environmental cost of AI infrastructure sits in a single sentence: supply chain emissions grew 25% year over year in 2025. Google attributes this increase to two compounding factors. The report states that the growth reflects both the scale of new AI infrastructure construction and the fact that Asia-Pacific supply chains, where much of that hardware is manufactured, continue to operate on electrical grids that remain undersupplied with carbon-free energy.

Google identifies specific structural causes behind that undersupply. According to the report, land constraints, high construction costs, and policy and regulatory hurdles in the region continue to slow the availability of clean power for manufacturing operations tied to the company's hardware supply chain. Unlike the company's owned and leased data center operations, where Google can select clean energy suppliers and structure long-term power purchase agreements, its upstream supply chain runs through third-party manufacturers operating on grids the company does not directly control.

This is a distinction that matters for how the 25% figure should be read. It is not a sign that Google's own facilities became dirtier; the report is explicit that operational emissions fell. It reflects, instead, the emissions embedded in the physical hardware, chips, servers, and construction materials required to build the infrastructure that supports the company's AI products, a category of emissions that has grown as the pace of AI infrastructure investment has accelerated across the industry.

That acceleration has been extensively documented elsewhere. Alphabet's advertising revenues climbed 14% as its Gemini app reached 750 million users, and the company disclosed capital expenditure guidance of between $175 billion and $185 billion for 2026, with roughly 60% of that spending directed toward machines including servers. In December 2025, Alphabet agreed to acquire Intersect Power for $4.75 billion in cash, a deal explicitly aimed at accelerating the company's ability to bring new energy and data center capacity online in the United States. That acquisition closed in March 2026, and by June 2026, Google and Intersect had jointly announced construction of two co-located data center and clean energy sites in Texas, including the Meitner Energy Center in Gray and Roberts Counties, which pairs a new facility with more than a gigawatt of wind, solar, and battery storage.

The pattern of pairing new data centers directly with dedicated clean energy generation extends beyond Texas. Google broke ground on its first self-developed, owned, and operated data center in Horndal, Sweden on June 2, 2026, a facility designed to be air-cooled rather than reliant on water from a nearby lake, addressing a concern local residents had raised years earlier. These projects represent the operational side of Google's clean energy strategy, the side where the 2025 report shows measurable progress. The supply chain side, where hardware is fabricated before it ever reaches a data center, is where the report shows the opposite trend.

Water replenishment reaches 78% of freshwater consumption

Beyond energy and emissions, the report discloses progress toward Google's water stewardship commitments. According to the report, water stewardship projects replenished approximately 7.7 billion gallons of water in 2025, representing roughly 78% of the company's 2025 freshwater consumption. Google states this marks progress toward a 2030 ambition to replenish more water than it consumes.

The figure invites comparison with disclosures from other hyperscale operators navigating similar water-intensity questions. Amazon reported reaching a water efficiency figure of 0.12 liters per kilowatt-hour in 2025, a number the company said was seven times better than the 0.84 liter-per-kilowatt-hour industry average it cited, and 75% of the way toward its own water-positive target for 2030. Both companies report progress toward similar water-positive commitments using different metrics, water replenished as a share of consumption for Google, and liters consumed per unit of energy for Amazon, making direct comparison difficult without a shared measurement standard.

AI tools deployed for emissions reduction and disaster response

The report also details how Google is applying artificial intelligence to environmental problems beyond its own operations. According to the report, nine of the company's AI-powered solutions collectively enabled individuals, cities, and partners to reduce an estimated 41 million metric tons of carbon dioxide equivalent in 2025, a figure Google describes as roughly three times the size of its own operational emissions.

Specific examples cited in the report include Google Earth, which the company states optimizes site layouts for solar and wind developers to accelerate project planning; Nest thermostats, which use machine learning to automate home energy savings; and fuel-efficient routing within Google Maps, which the report says analyzes traffic and terrain data to suggest lower-emission routes for drivers.

The report additionally describes AI applications directed at natural disaster response and biodiversity monitoring. Google states it has advanced AI systems for forecasting and early detection of wildfires, floods, earthquakes, and extreme weather events. On biodiversity, the report names two specific tools: Perch, described as a bioacoustic embedding model for analyzing large bioacoustic datasets, and SpeciesNet, an Earth AI model the company states can recognize more than 2,000 animal species in images captured by motion-triggered wildlife cameras with over 94% accuracy.

Internal AI tools shaped the report's own production

Separately from the environmental content itself, Google has previously disclosed that artificial intelligence played a role in producing its sustainability reporting process. Google shared an internal AI playbook in December 2025 documenting two years of experimentation with NotebookLM and custom Gemini configurations for tasks including claims validation, where a custom tool cross-references draft environmental statements against internal reporting guidelines before human review. Luke Elder, Senior Lead for Sustainability Reporting at Google, described the broader philosophy behind the effort at the time, stating that the company views sustainability reporting as "a collaborative endeavor, not a competitive one." That playbook did not address the 2026 report specifically, but it documents the workflow infrastructure Google has built for processing exactly the kind of claims validation and data classification that annual environmental disclosures require.

Moonshot targets set in 2020 face mounting pressure

Google set net-zero and 24/7 carbon-free energy targets, which the report refers to as moonshots, at the beginning of the current decade. The report states these targets were intentionally aspirational from the outset and that the company acknowledged achieving them would require significant innovation in both technology and policy.

According to the report, the rapid rise of AI over the past several years has reshaped global infrastructure demands and placed new pressure on electrical grids, a development the company says has expanded the scale of what its moonshot targets now require. Google states plainly in the report that reaching its climate moonshots "is getting harder," citing long wait times to connect new projects to electrical grids, fragmented energy markets, supply chain delays, and regulatory bottlenecks as factors slowing the pace at which new carbon-free energy can come online.

This acknowledgment arrives as broader research has flagged environmental harm from AI infrastructure as a persistent risk category. A study cited by PPC Land found that environmental harm from AI development retained a catastrophic probability above 10% even under pragmatic mitigation scenarios, a threshold the research treated as structurally resistant to improvement because the economic incentives driving compute buildout continue to outpace the incentives for reducing its environmental footprint. The report from Google does not reference this research directly, but its own account of accelerating AI infrastructure construction outpacing grid decarbonization describes a similar dynamic from the inside.

Why this matters for the marketing and advertising community

For advertising professionals, the report's supply chain emissions disclosure connects directly to infrastructure questions that shape the industry's daily tools. The compute capacity that runs programmatic auctions, real-time bidding systems, and AI-powered campaign automation is physically housed in the same data centers this report describes. Scope 3 emissions accounting, the category that captures indirect environmental impact from purchased services and supply chains, is becoming a procurement consideration for advertising technology buyers and agencies evaluating vendor sustainability claims, a trend PPC Land has tracked as cloud vendors increasingly disclose site-level energy and water data.

The AI Max and Performance Max campaign types that have driven substantial changes in how advertisers interact with Google's platforms in 2025 and 2026 depend on the same underlying compute infrastructure documented in this report. Google's own data cited at its Marketing Live event in May 2026 stated that Gemini reduces irrelevant ads by 40%, based on internal data spanning 2024 to 2025, a claim that depends on the AI infrastructure whose environmental footprint is detailed in this report. As AI Mode surpassed one billion monthly users by Google I/O in May 2026, the compute demand behind that growth compounds the same infrastructure pressures the environmental report describes.

The report's transparency about rising supply chain emissions also offers advertisers and publishers a more complete picture than isolated marketing claims typically provide. Where earlier disclosures from large technology platforms have sometimes emphasized headline clean energy figures without addressing offsetting increases elsewhere in their footprint, this report explicitly states that supply chain emissions rose by 25% in the same year that operational emissions fell. That level of disclosure gives marketing organizations building their own environmental, social, and governance reporting a clearer basis for assessing the actual footprint of the cloud and AI infrastructure providers they rely on, rather than a single favorable metric presented in isolation.

Timeline

  • 2010: Google begins the multi-year clean energy contracting program referenced in the report, which by 2025 totals more than 240 agreements.
  • 2020: Google sets its net-zero and 24/7 carbon-free energy moonshot targets, later identified in the report as aspirational goals requiring significant technological and policy innovation.
  • 2019: The report identifies this year as the baseline for its moonshot-related carbon footprint comparisons.
  • 2025: Google signs agreements for more than 12 gigawatts of net-new clean energy; operational emissions fall 2% year over year; electricity demand rises 37% year over year; supply chain emissions increase 25% year over year; water stewardship projects replenish approximately 7.7 billion gallons, or 78% of freshwater consumption; nine AI solutions enable an estimated 41 million metric tons of CO2 equivalent in avoided emissions for external partners and cities.
  • June 30, 2026: Google publishes its 11th annual Environmental Report on the Keyword blog, authored by Chief Sustainability Officer Kate Brandt.

Summary

Who: Google, through Chief Sustainability Officer Kate Brandt, published the disclosures. The report also references work from teams across the company's data center energy, AI research, and water stewardship programs.

What: Google's 11th annual Environmental Report discloses a 2% reduction in operational emissions alongside a 25% increase in supply chain emissions, both measured year over year for 2025. The report also details more than 12 gigawatts of new clean energy agreements signed in 2025, a 37% increase in electricity demand, water replenishment reaching 78% of freshwater consumption, and AI tools the company states enabled an estimated 41 million metric tons of avoided emissions for external partners.

When: The report covers the 2025 calendar year and was published on June 30, 2026, on Google's Keyword blog.

Where: The disclosures span Google's global operations, with specific reference to Asia-Pacific supply chains as the primary source of the reported increase in supply chain emissions.

Why: The report matters for the marketing and advertising community because the compute infrastructure it describes, including data centers, clean energy contracts, and AI hardware supply chains, underpins the programmatic advertising, AI-powered campaign tools, and search products that advertisers and publishers depend on daily. The disclosure of rising supply chain emissions alongside falling operational emissions gives procurement and sustainability teams a more complete basis for assessing the environmental footprint of the AI infrastructure providers they rely on.