In the race to dominate artificial intelligence, some of the world’s biggest technology companies are betting on vast new data centers. But those concrete-and-server monoliths come with an enormous hidden price tag: carbon emissions that threaten to undermine years of progress on climate change.
According to a new analysis highlighted by the Financial Times and picked up by CNET, the 60 largest data centers currently under construction by America’s leading tech firms will pump an estimated 101.5 million tons of carbon dioxide into the atmosphere every year once they are fully operational. That is roughly equivalent to putting an additional 24 million gasoline-powered cars on the road, a stark illustration of AI’s growing environmental footprint.
A Carbon Shock Equivalent to Millions of Cars
The numbers are staggering. The 101.5 million tons of annual CO2 emissions would be comparable to the yearly output of a small industrialized nation. To visualize it: if all those emissions came from cars, they would form a line of vehicles stretching across the globe many times over.
Data centers are energy hogs. They require round-the-clock electricity to keep servers humming and cooling systems operational. Even as renewable energy expands, the sheer volume of new compute capacity is creating a surge in fossil-fuel-based power demand. The report notes that while solar power has boomed in the U.S. and internationally, the gains are at risk of being wiped out by the power-hungry systems that run AI models—models that, as the original article wryly notes, can hallucinate entire motorcycles that don’t exist.
Big Tech’s Emissions Are Climbing Fast
Perhaps the most troubling indicator is that emissions from the biggest cloud providers are rising at an alarming clip. According to the Financial Times analysis, Amazon’s emissions increased by 16% between 2024 and 2025. Alphabet, the parent company of Google, saw an 18% rise. Microsoft’s emissions jumped a staggering 25% over the same period.
These increases are not accidental. They are the direct result of a surge in data center construction and an insatiable demand for AI compute. The percentages reflect a broader industry-wide trend that runs counter to the net-zero commitments many of these companies have loudly proclaimed. In short, the cloud has a carbon cloud of its own.
- Amazon: +16% emissions (2024–2025)
- Alphabet/Google: +18% emissions
- Microsoft: +25% emissions
The Grid Is Already Feeling the Strain
Those massive electrical demands are not just a climate issue; they’re also stressing the existing power grid. Utility companies in several states have warned that the rapid build-out of data centers could lead to brownouts or delays in connecting new facilities to the grid. The problem is compounded by the fact that solar energy, while cheap and clean, only works when the sun is shining. Batteries are helping to bridge the gap, but they cannot yet provide the 24/7 reliability that data centers demand.
The situation is made worse by policy headwinds. The report notes that the Trump administration has been slashing clean energy tax credits, making it more expensive—and therefore more difficult—to build new renewable capacity to offset the data center surge. The result is a perfect storm: rising energy demand, a grid stretched thin, and fewer incentives to decarbonize.
Outlandish Solutions: Orbital Data Centers and Space Mirrors
Faced with these obstacles, tech giants are exploring ideas that sound less like engineering and more like science fiction.
Beaming Sunlight From Space
Meta has proposed launching giant mirrors into orbit to reflect sunlight down onto solar panels on Earth around the clock. This would, in theory, solve the intermittency problem of solar power by providing constant illumination. Critics have noted that the plan sounds like a supervillain’s scheme to melt the ice caps, but Meta insists it is a serious avenue of research.
Data Centers in Orbit
Google, meanwhile, has floated plans for a formation-flying fleet of AI data centers in orbit. Not to be outdone, SpaceX has filed for permission to place a million such satellites in orbit. The appeal is obvious: in space, solar panels receive unfiltered sunlight 24/7, and no emissions are produced on Earth. The logistical hurdles—launch costs, maintenance, and space debris—remain towering.
Nuclear Power: The Molten-Salt Elephant in the Room
If space is too far-fetched, many tech companies are turning to a more terrestrial option: nuclear energy. Several firms have signed agreements with nuclear startups, betting that a new generation of smaller, modular reactors can provide clean, constant power without the radioactive baggage of traditional plants.
Nuclear power has the advantage of running around the clock with zero CO2 emissions. But it also carries risks. The radioactive waste problem remains unresolved, and public opposition to new plants has stymied the industry for decades. The startups promise that their new, smaller reactors are safer and more adaptable to changing energy needs—a promise that remains unproven at commercial scale.
The Bottom Line
Artificial intelligence may be transforming the world, but its physical infrastructure is taking a measurable toll on the planet. The 60 data centers now rising across the U.S. represent a CO2 burden that will persist for decades, and the companies responsible are still struggling to align their growth with their climate pledges.
Until cleaner energy sources can scale quickly enough—whether from the sun, the atom, or some invention not yet imagined—every AI query will carry a heavier environmental cost. And as the analysis shows, that cost is no longer theoretical. It’s as real as the 24 million cars that those data centers will, in effect, add to the road.
This article is based on reporting by Jalopnik. Read the original article.
Originally published on jalopnik.com
