FTFuture Technology
COMPUTING

Microsoft's Carbon Emissions Jumped 25 Percent Last Year. AI Is Why.

· 3 min read · By Nath Connell

Key takeaways

  • Microsoft's carbon emissions rose 25 percent in the past year, according to its 2026 sustainability report
  • The increase is driven primarily by AI data centre expansion
  • Microsoft pledged in 2020 to be carbon negative by 2030
  • Google reported a 48 percent emissions increase in its 2024 sustainability report, showing this is an industry-wide pattern
  • Microsoft has signed deals including purchasing power from the reopened Three Mile Island nuclear plant

Microsoft has published its 2026 sustainability report, and the headline number is uncomfortable: the company's carbon emissions rose 25 percent in the past year. For a company that made a high-profile pledge in 2020 to be carbon negative by 2030, meaning it would remove more carbon than it emits, that trajectory is going in exactly the wrong direction.

The culprit is not a mystery. Microsoft has been on an extraordinary data centre building spree, driven almost entirely by demand for AI compute. Training and running large language models consumes enormous amounts of electricity. The chips involved, primarily NVIDIA GPUs, draw hundreds of watts each, and data centres run thousands of them simultaneously. Cooling those facilities adds further energy demand. The result is that every meaningful advance in AI capability comes with a significant carbon cost attached.

The Scale of the Problem

To understand why a 25 percent jump is so significant, it helps to remember that Microsoft was not a small emitter to begin with. The company's global operations span data centres on multiple continents, office buildings for hundreds of thousands of employees, and an enormous supply chain. A 25 percent increase on top of that baseline represents a very large absolute quantity of additional emissions.

The 2030 carbon negative pledge was ambitious when it was made, and Microsoft has invested genuinely in renewable energy procurement, sustainable aviation fuel for business travel, and internal carbon pricing mechanisms. The problem is that AI demand has grown faster than any of those mitigation efforts can keep up with. Clean energy procurement takes years to negotiate and build. AI investment decisions can be made in weeks.

Microsoft is not alone in this bind. Google's emissions have also risen sharply in recent years, with the company reporting a 48 percent increase in its 2024 sustainability report. Both companies are discovering that the theoretical compatibility of clean energy and AI in practice requires a pace of renewable deployment that the energy sector simply cannot match right now.

The future, in 3 minutes a day. The biggest tech story explained every morning, free. Get the briefing →

What Microsoft Is Doing About It

The company has pointed to several initiatives it says will help bend the curve. These include investments in nuclear energy, particularly small modular reactors, long-term power purchase agreements for renewable energy, and research into more energy-efficient AI architectures. Microsoft has also made investments in carbon removal technologies, including direct air capture and enhanced weathering.

The nuclear angle is particularly interesting. Microsoft signed a deal in 2023 to purchase power from the reopened Three Mile Island nuclear plant in Pennsylvania, a genuinely significant commitment that signalled how seriously it views nuclear as part of its energy strategy. Small modular reactors, if they reach commercial scale by the late 2020s as some developers project, could provide the always-on clean power that intermittent renewables like wind and solar cannot reliably deliver.

But all of that is a longer-term play. In the near term, the honest answer is that Microsoft's emissions are going up because it is building more data centres than it can currently power cleanly, and it is prioritising AI infrastructure investment over staying on the 2030 emissions trajectory.

The Broader Question

This is the tension that the tech industry has been dancing around for two years now. AI is broadly presented as a tool that will help solve climate change, through better climate modelling, smarter grid management, and optimised industrial processes. That may well be true in the long run. But the short-term cost of building the AI infrastructure to get there is a meaningful increase in the emissions those tools are supposed to help reduce.

There is a reasonable argument that this is a worthwhile trade-off. There is also a reasonable argument that companies should be more honest about the environmental cost of their AI ambitions rather than emphasising the long-term potential benefits. Microsoft's 2026 sustainability report, to its credit, does not hide the numbers. A 25 percent increase is right there in the headline. Whether that transparency is enough, when the 2030 pledge is looking increasingly unlikely to be met, is a different question.

Sources

Get the briefing, free

The biggest tech story, explained in 3 minutes every weekday. Choose your briefings →

Free. No spam. Unsubscribe in one click.