The rapid buildout of US datacenters has led to planet-heating emissions equivalent to Denmark's annual carbon output, according to a new study. Specifically, 403 hyperscale datacenters emitted about 25.7 million metric tons of CO2 over a year, matching the entire annual carbon pollution of Denmark, a country of 6 million people.
The research, published in Nature Sustainability, highlights the environmental cost of the AI boom. These facilities consumed about 1.8% of total US electricity, with emissions two to three times higher than previous estimates for 2018. The study's authors warn that the true impact is likely underestimated.
What the Study Found
The study, led by Gianluca Guidi and co-authored by Francesca Dominici of Harvard University, used federal government data to track emissions from hyperscale datacenters from May 2024 to April 2025. Hyperscalers are the largest datacenters, primarily used to handle the growth of artificial intelligence.
The 25.7 million metric tons of CO2 is equivalent to Denmark's annual emissions. This is two to three times more than a previous estimate of hyperscale emissions in 2018, indicating a sharp increase in the sector's carbon footprint.
Dominici noted, "There has been an explosion in these datacenters and the CO2 emissions are pretty substantial." She added that since 2025, many more datacenters have come online, suggesting the study's figures are conservative.
Why This Matters
The findings underscore the environmental trade-off of the AI revolution. As tech companies race to build more datacenters to support AI models, energy demand and emissions are surging. This growth is concentrated in states like Virginia, Oregon, Ohio, and Iowa, where many hyperscale facilities are located.
The study's estimate is likely conservative because it only includes data up to April 2025. Since then, the expansion has accelerated, meaning the actual emissions could be higher. This raises questions about the sustainability of the AI industry and its impact on climate goals.
Moreover, the emissions from datacenters contribute to air pollution and climate change, affecting public health and the environment. As Dominici pointed out, "we need to think of this" in the context of broader energy and environmental policies.
Key Takeaways
- US hyperscale datacenters emitted 25.7 million metric tons of CO2, equal to Denmark's annual emissions.
- These facilities consumed about 1.8% of total US electricity.
- Emissions are two to three times higher than 2018 estimates, and likely underestimated due to rapid growth.
- The buildout is concentrated in Virginia, Oregon, Ohio, and Iowa.
- The study calls for greater scrutiny of the AI industry's environmental impact.
Comparing Datacenter Emissions
To put the emissions into perspective, here's how US hyperscale datacenters compare to other entities:
| Entity | Annual CO2 Emissions (million metric tons) |
|---|---|
| US hyperscale datacenters (2024-2025) | 25.7 |
| Denmark (entire country) | 25.7 |
| US hyperscale datacenters (2018 estimate) | ~8.6-12.9 |
The table shows that the emissions from just 403 datacenters equal those of a whole nation. This scale highlights the need for sustainable practices in the tech industry.
What to Watch Next
As AI continues to grow, so will the demand for datacenters. Companies and policymakers must address the energy sources powering these facilities. Renewable energy adoption and efficiency improvements could mitigate emissions, but the pace of construction may outstrip these efforts.
Regulators might also scrutinize the environmental impact of datacenters more closely. The study's authors emphasize the need for accurate reporting and transparency. Without action, the carbon footprint of the tech sector could undermine climate progress.
For those following energy and utility trends, the intersection of technology and sustainability is crucial. The debate over gas stoves and asthma is one example of how energy choices affect health, as discussed in replacing gas stove benefits asthmatics equivalent to medications. Similarly, the environmental impact of datacenters is a growing concern. The background to this story is set out in London Datacentre Carbon Footprint: 27,000 Flights to New York. A related thread runs through London Datacentre Carbon Footprint Equals 27,000 Flights.
Water usage is another issue. Datacenters require significant water for cooling, leading to pollution risks. A case study on US datacenters face water pollution backlash: Meta case study illustrates these challenges. As the industry expands, such environmental trade-offs will need careful management. Earlier coverage in Drax Datacentre Emissions: Nearly Double Gatwick Flights Annually shows how the argument developed. A closer look at the same subject appears in Energy Utilities: latest articles.
The energy demand from datacenters also strains utilities. In regions with high concentrations, grid operators must balance supply and sustainability. The study's findings should prompt discussions on energy policy and corporate responsibility. Anyone following the wider picture can start with EV Owners Overpaying: The Real Cost of Charging Your Electric Car. The same theme surfaces in Data Center Heat Recovery Powers 11,000 Homes in Denmark, which takes a different angle.
FAQ
What is a hyperscale datacenter?
A hyperscale datacenter is a large-scale facility designed to handle massive amounts of data and computing power, primarily for artificial intelligence and cloud services. These facilities are operated by major tech companies and are known for their high energy consumption.
How do datacenter emissions compare to other industries?
The 25.7 million metric tons from hyperscale datacenters is equivalent to Denmark's annual emissions. This is significant but smaller than emissions from sectors like transportation or manufacturing. However, the rapid growth of datacenters makes them a growing concern.
Why are emissions likely underestimated?
The study used data up to April 2025, but since then, many more datacenters have been built. The authors note that the expansion has accelerated, so actual emissions are probably higher than reported.
What can be done to reduce datacenter emissions?
Solutions include powering datacenters with renewable energy, improving energy efficiency, and implementing stricter regulations. Companies can also invest in carbon offsets and innovative cooling technologies to lessen their environmental impact.