I was on a call when the slide changed to a skyline of humming data centers and a single line: “$41 billion this quarter.” You could feel the room tilt—engineers cheering, environmental advisers silent. I want you to hold that image while we unpack what it means for the planet.
Microsoft spent $41 billion (€38 billion) in one quarter on AI and data centers, and then quietly slashed carbon removals by about 80%.
That is the factual spine of a BloombergNEF finding: purchases of carbon removal credits are down nearly four-fifths year over year while spending on artificial intelligence infrastructure is soaring.
I say this not to scold but to orient you: corporate climate promises are often a portfolio of strategies, and when one pillar shrinks the balance sheet shifts the conversation. You’ve heard Microsoft’s 2030 goals—now ask which levers get pulled when Azure, OpenAI, and chip-hungry models call for capital.
Why did Microsoft cut its carbon removal purchases by 80%?
Because the company redirected billions to build AI capacity: Bloomberg reported a 25% rise in Microsoft’s emissions last year, and the latest quarter included that $41 billion (€38 billion) outlay, with the firm forecasting roughly $175 billion (€161 billion) on AI infrastructure for the year.
I’ll be blunt: buying fewer removal credits is not proof of a lost ambition, but it is evidence of shifting priorities. When you fund data centers and GPU farms at that scale, you are effectively choosing where your climate capital flows—into avoidance and efficiency, or into more raw compute that itself consumes energy.
On the ground, a new kind of buildout looks like a construction boom of humming warehouses and gas turbines.
Drive near planned data center sites and you’ll see cranes, substations, and often temporary power solutions that run on natural gas.
These gas turbines—internal combustion engines in all but name—are an expedient answer to enormous power swings. They vent nitrogen oxides, particulates, and formaldehyde into neighboring neighborhoods. Researchers at the Environmental Data & Governance Initiative found higher-than-average air pollution within a mile of some data centers. That’s not an abstract projection; that’s people breathing worse air.
How does AI increase emissions?
AI is two things at once: a tool for energy optimization and a voracious consumer of electricity. A Nature study last week argued that the climate gains from AI-driven efficiency are often outmatched by the emissions AI can enable—especially when the technology is used to make fossil-fuel extraction cheaper and faster.
I’ve spoken to oil and gas executives; they see models as a force multiplier—optimizing drilling, reducing costs, and extending the life of existing fields. In effect, the same algorithms that could shave waste from wind- and solar-integration are also being trained to squeeze more fossil output from the ground.
Local communities are already paying the price while global emissions climb 25% because of AI appetite.
That figure—AI-linked emissions rising by a quarter last year—comes from analyses tied to Microsoft’s own disclosures and industry estimates. You feel that growth in both the skyline of new data centers and in municipal permits for on-site generation.
New York’s recent moratorium on large-scale data centers is an example of pushback. Officials worried about water, grid strain, and air quality; residents worried about their health. I’m not saying the industry has no benefits—Azure and Google Cloud improve services—but benefits don’t erase the externalities when they’re concentrated in certain zip codes.
Policy, markets, and corporate choices will decide whether AI is a net climate help or harm.
Microsoft frames its reduced purchase of removal credits as part of a broader mix of actions. I’ve heard that line before: companies announcing a palette of measures while major capital flows tell a different story.
If you use the lexicon of investors, this is portfolio reallocation. If you use the language of neighbors, it’s a neighborhood getting louder. One metaphor: the corporate carbon plan is now less a balanced diet and more a heavy entrée with a small salad on the side.
Executives at oil, gas, and cloud firms are already plotting applications that favor extractive industries.
IBM, Shell, and drilling firms have all signaled interest in AI for reservoir modeling and predictive maintenance; Amazon Web Services and Google Cloud are selling similar services to energy clients.
Those market signals matter. AI vendors and cloud platforms can be the architects of efficiency—and the architects of cost cuts that prolong fossil projects. Think of AI as both a scalpel and an accelerant: one tool that can heal systems or make old habits burn faster.
Will Microsoft’s AI spending undermine its climate promises?
It can, unless policy and corporate governance reshape incentives. Carbon removal credits were never the only path to net-zero, but they were a fast way to offset emissions that are hard to avoid. Cutting them dramatically while pouring capital into new, power-hungry infrastructure raises legitimate questions about ambition versus activity.
You should watch three signals: where Microsoft locates centers (which grids and communities), whether it binds power purchases to verified renewables, and how it measures the lifecycle emissions of new AI services sold to oil and gas clients.
I can tell you this as someone who tracks both markets and municipal hearings: tech stacks don’t float above politics. If you want different outcomes, you will need different rules and different metrics—ones that put public health and system-wide emissions next to quarterly earnings in the same ledger.
So ask yourself: are we building an AI future that cools the planet, or are we funding an intelligence that makes fossil fuels cheaper and greener on paper but dirtier for the people who live next to the servers and the rigs?