Microsoft set one of the most ambitious climate targets in the tech industry: by 2030, it would remove more carbon from the atmosphere than it emits. Yet its latest sustainability report reveals the company is moving in the opposite direction. Greenhouse gas emissions rose 25.1% in the last financial year, climbing from 16.2 million metric tonnes to 20.3 million. That is roughly 58% above the 2020 baseline Microsoft established when it made the pledge. The news has sent shockwaves through the environmental community and raised serious questions about the feasibility of the tech giant's goals.
Why the jump is partly honesty
Not all of that increase represents fresh pollution. Microsoft has made a significant accounting change: it stopped buying short-term renewable energy certificates (RECs) that do not actually add clean power to the grid. Those certificates had artificially lowered last year's reported emissions. Without them, the true number climbs even higher. This is evident in one key metric: Scope 2 emissions, those from purchased electricity, soared from 1.6% of the total to 13.3% in a single year. That jump reflects accounting finally catching up with the reality of the data centres powering Microsoft's cloud and AI ambitions.
The rest of the rise is genuine growth. Microsoft is building AI and cloud infrastructure at breakneck speed, and each new data centre requires enormous amounts of energy, steel, and concrete. The company's diesel and crude oil usage rose 51%, even as it managed to cut natural gas and petrol consumption. This underscores a fundamental tension: the very technology that Microsoft is betting its future on is the same force that is dismantling its climate maths.
The progress it wants you to see
Microsoft is not solely pointing at the bad number. The company claims that its efficiency measures and clean-power deals would have pushed emissions close to 34 million tonnes had they not been implemented. It also matched 100% of its annual electricity consumption with renewable energy purchases, restored more water than it consumed (over 14 million cubic metres), and reused or recycled 92% of its retired servers. Chief sustainability officer Melanie Nakagawa stated, “Innovation at this scale must be matched by responsibility at the same scale.” These achievements are real and significant, but they also highlight the core problem: the headline target remains stubbornly out of reach. The pledge was a hard, time-bound goal, and the trend line continues to head in the wrong direction.
The company has poured billions into carbon removal technologies, mass timber building materials, and some of the largest corporate clean power deals in history. Yet none of these efforts are scaling as quickly as the data centres are rising. The deadline is now just four years away, and without a dramatic acceleration in emissions reductions, the goal appears increasingly unattainable.
The maths of the pledge
The tension between AI growth and climate responsibility is starkly reflected in one accounts line. Microsoft's investment in AI and cloud computing has become the single largest driver of its environmental footprint. Each AI model training cycle, each new data centre, demands massive amounts of electricity. A typical hyperscale data centre now consumes as much power as a small city. And as AI models become more complex, the energy requirements only grow. The company's own report acknowledges that its Scope 1 emissions (direct from operations) increased 11.7%, while Scope 2 (electricity) jumped 87% and Scope 3 (supply chain) rose 14.6%.
Meanwhile, the technological solutions Microsoft is betting on—like carbon capture, direct air capture, and advanced nuclear—are still in early stages of deployment. Carbon removal credits are expensive and supply is limited. Clean power contracts for data centres, while growing, often take years to come online. In the interim, many data centres rely on fossil fuel backup generation, especially during grid stress events. This has led to a resurgence of natural gas plant proposals in regions where data centre clusters are booming, including northern Virginia, Ohio, and the Pacific Northwest.
Everyone's number is moving
Microsoft is not alone in this predicament. Amazon's emissions rose 16% last year, and Google's climbed about 25%, with both companies citing AI and data centre expansion as primary drivers. Meta's massive Hyperion data centre campus recently ballooned past $50 billion in projected cost and sparked a political controversy in Louisiana over local zoning and environmental permits. The industry-wide build-out has triggered the largest gas plant construction wave in a decade, pushed residential electricity bills higher across the Rust Belt, and even led New York to become the first state to freeze new data centre connections due to grid capacity concerns.
In response, some firms are exploring radical solutions, such as floating data centres at sea to avoid land-based constraints. Others are investing heavily in advanced nuclear reactors and attempting to revive defunct nuclear sites. Yet all of these options remain expensive, risky, and years away from making a meaningful impact. The industry keeps promising that efficiency gains—better servers, liquid cooling, AI-optimized operations—will eventually bend the emissions curve downward. But for now, the curve is climbing, and the clock is ticking.
The broader implication is that the tech industry's climate goals are increasingly at odds with its growth ambitions. The very purpose of AI—to accelerate innovation—is also accelerating environmental harm. Microsoft's report serves as a stark reminder that even the wealthiest and most committed companies struggle to decouple growth from emissions. The coming years will test whether the industry can truly reconcile its two most important promises: to lead the AI revolution and to protect the planet.