Microsoft's Emissions Rose 25% Building AI Infrastructure. The Company Says It's Also Water-Positive for the First Time.

Microsoft’s 2026 Environmental Sustainability Report, published July 9 and covering fiscal year 2025 against a 2020 baseline, discloses that company-wide emissions rose 25% year over year — driven, the report says, largely by continued datacenter expansion to support AI workloads. In the same report, Microsoft says it was water-positive globally for the first time, replenishing more than 14 million cubic meters more water than it withdrew, and matched 100% of its annual global electricity consumption with renewable energy. Vice Chair Brad Smith and Chief Sustainability Officer Melanie Nakagawa frame the tension directly: AI can deliver real societal and economic benefit, but the sustainability solutions needed to support it aren’t scaling as fast as the demand AI infrastructure is generating for energy, water, land, and materials.

That demand is regional as much as it is corporate. McKinsey estimates the global data center value chain needs roughly $6.7 trillion in cumulative capital investment between 2025 and 2030, with Asia-Pacific alone reaching about 34% of global demand by 2030 — a buildout scale that makes Microsoft’s 25% emissions jump one data point in a much larger curve, not an outlier. A separate whitepaper aimed at hoteliers makes the same point from the buyer’s side: the AI industry has entered an “engineering-reality phase” where infrastructure costs — chips, power, grid strain — show up in vendor pricing and system reliability months before they’re visible in a product release note.

For companies weighing AI infrastructure commitments or vendor contracts, Microsoft’s own numbers are a rare instance of a hyperscaler quantifying the environmental cost side of the buildout rather than only the capability gains — a disclosure worth reading alongside any vendor’s efficiency claims, not instead of them.