Microsoft’s carbon footprint is heading the wrong way, and AI is the reason. According to The Verge AI, the company’s 2026 sustainability report shows its carbon emissions rose 25 percent in 2025, hitting 34 million metric tons “without select interventions.” The Verge AI reports that Microsoft blames the spike “primarily by the expansion of our datacenter infrastructure,” the physical backbone of its AI ambitions.
This is significant because Microsoft has spent years positioning itself as a climate leader. Back in 2020, it pledged to be carbon negative by 2030, meaning it would pull more carbon out of the atmosphere than it emits. That target now looks harder to hit with every new server rack.
What happened
Two forces pushed the numbers up:
- Datacenter expansion. Training and running large AI models demands enormous compute. More compute means more data centers, more electricity, and more embodied carbon in the concrete, steel, and chips.
- A cleaner accounting choice. Last February, Microsoft stopped buying “non-additional, unbundled renewable energy certificates.” These are credits that let a company claim green energy without actually adding new clean power to the grid. Dropping them made Microsoft’s real emissions look higher on paper, but the reported figure is now more honest.
The report doesn’t sugarcoat the tension. It admits that “while AI infrastructure is driving demand for energy, water, land, and materials, sustainability solutions are not scaling fast enough to meet demand.”
Why this matters
What stands out here is that this isn’t a one-off stumble. Microsoft’s 2024 sustainability report showed a similar climb in climate pollution, so this is a trend, not a blip. The company set an aggressive 2030 goal before the generative AI boom reshaped its business. The math has changed underneath the promise.
And Microsoft isn’t alone. The Verge AI notes that Google reported a 25 percent jump in its supply chain emissions in its own 2026 report, while Amazon logged a 16 percent increase. The three biggest cloud providers are all watching their emissions rise at the exact moment they’re racing to build more AI capacity. That’s the core conflict of this AI era: the infrastructure that powers smarter models also burns more energy and water.
Water is becoming its own flashpoint. In June, Amazon said its data centers used 2.5 billion gallons of water in 2025, and claimed that’s less than Microsoft used. When hyperscalers start comparing water bills in public, you know the resource cost of AI has moved from a footnote to a headline.
The bigger picture
Here’s the honest read. The old status quo let tech giants buy their way to green-sounding claims through cheap, unbundled certificates. Microsoft walking away from that practice is a real step toward transparency, even though it makes the headline number worse. Cleaner accounting and rising demand collided at the same time.
But transparency doesn’t remove carbon from the sky. The gap between AI’s growth curve and the pace of clean energy buildout is widening. Renewable power, grid capacity, and carbon removal all scale slowly. AI compute is scaling fast. Something has to give.
What to watch next
For practitioners and anyone tracking the AI industry, a few things are worth keeping an eye on:
- Nuclear and grid deals. Expect the hyperscalers to keep signing power agreements, including nuclear, to feed data centers without wrecking their emissions targets.
- Water disclosure. After the Amazon-Microsoft water comparison, more detailed water reporting is likely coming from all the majors.
- The 2030 reckoning. Microsoft still says it wants to be carbon negative by 2030. Each yearly report now doubles as a progress check on whether that pledge survives the AI boom.
The uncomfortable truth is that the AI arms race and corporate climate goals are pulling in opposite directions right now. Microsoft’s report is one of the clearest admissions yet that the industry hasn’t solved that contradiction. You can read the full breakdown at the original report from The Verge AI.