Microsoft’s carbon footprint is swelling.

A BloombergNEF report from Thursday found that the tech giant bought about 80% fewer carbon removal credits so far this year than it did in the same time period last year, marking its first retreat in three years in the global carbon removal market that it dominates.

In response, the tech giant told Bloomberg that carbon removal is not the only strategy they are implementing towards previously proclaimed decarbonization goals, and that the changes did not imply a shift in that “ambition.” But that retreat from removing carbon emissions went hand in hand with Microsoft’s accelerating investments in climate-polluting artificial intelligence infrastructure.

According to Bloomberg, Microsoft’s AI bet resulted in a 25% increase in emissions last year. Since then, that bet has increased dramatically.

In just the three months from April to June, Microsoft said that it spent a whopping $41 billion, with a majority of that money going towards artificial intelligence initiatives and data center construction. The company said in its latest earnings report that it expects to spend roughly $175 billion on A.I. infrastructure this entire year.

Artificial intelligence is often touted as a productivity hack that could drive wins in the fight to reduce climate emissions by helping industries optimize energy efficiency. But a study published last week in Nature found that the climate benefits that can be gained from AI’s ability to optimize energy across renewables or the grid are dwarfed by the amount of benefits that the technology can similarly provide to boost the fossil fuel industry.

In effect, artificial intelligence can also be used to make fossil fuel extraction cheaper, and oil and gas industry executives say they are eager to adopt the technology for just that purpose.

“Absent policy steering, AIâs modeled effects increase the carbon intensity of the global economy and reinforce fossil fuel incumbencyâoutcomes that current analytical and governance frameworks do not fully capture,” the researchers wrote.

The prevalence of AI has also increased the need for power-hungry data centers and kickstarted an unprecedented AI infrastructure buildout that has taken over pretty much the entire world. More often than not, these data centers rely on fossil fuels to address their huge energy demand. A particularly popular option is gas turbines, aka internal combustion engines that rely on natural gas to spin the turbine and generate energy, emitting smog-forming pollution, fine particulate matter, carbon monoxide, and formaldehyde in the process. A study from the Environmental Data & Governance Initiative found that the communities living within a mile radius of a data center were breathing above-average levels of air pollution.

Although data centers account for a relatively small amount of carbon emissions, they are one of the handful of polluting industries that continue to grow exponentially in this more climate-conscious world. According to the International Energy Agency, some scenarios see data centers as boasting “the largest emissions growth among all sectors.”

Studies have also claimed data centers can further harm the environment by creating heat islands in their local neighborhoods. These numerous negative environmental impacts tied to data center construction have gained increasingly more attention in the past few months, and were a key reason why New York state recently announced a moratorium on major new data centers, along with many other jurisdictions.