Data Centers: Between a Threat to the Climate and a Solution for Climate Change
A study by Allianz Trade shows that data centers significantly increase electricity demand, but could reduce emissions more than they cause in the long term through AI applications
June 30, 2026
Source: E & M powernews
The expansion of data centers, including those for AI applications, is driving up electricity demand. However, researchers at Allianz Trade also see this development as an opportunity for climate protection.
The operation of data centers is far more harmful tothe climate than previously assumed. For 2025, the study “Code, Carbon, Kilowatts” by the credit insurer Allianz Trade estimates the resulting global emissions at 286 million metric tons of CO2. The authors note that this is 57 percent higher than the previous estimates used by the International Energy Agency (IEA). This is because, in addition to pure electricity consumption—which accounts for 79 percent of emissions—emissions from construction and hardware production must also be taken into account. According to the study, 23 percent of total emissions are attributable solely to the manufacture of servers, semiconductors, and infrastructure.
However, the actual level of emissions from a data center depends largely on its location. “Identical computing power can result in a multiple of emissions depending on the electricity mix,” says Patrick Hoffmann, senior climate economist at Allianz. “In countries such as India, Indonesia, or Malaysia, for example, over 600 grams of CO2 are emitted per kilowatt-hour (kWh). In Norway and Sweden, the figures are below 30 grams.” The comparatively low-emission electricity mix in Europe could therefore become a structural advantage for the expansion of climate-friendly AI.
However, the authors view Germany as lagging behind in this regard: “In this country, emissions from the German electricity mix stand at 329 grams of CO2 per kilowatt-hour. While this is below the levels in the U.S. and China, it is significantly higher than in Scandinavia, for example. The reason is the still-high share of coal-fired power generation (20.6 percent) and biomass use (10 percent).”
Structural Driver of Electricity Demand
Globally, companies invested approximately 580 billion U.S. dollars in data centers in 2025. According to the study, their electricity consumption stood at 515 billion kWh and could more than double to about 1,111 billion kWh by 2030.“Data centers are evolving from a marginal factor to a structural driver of electricity demand in many regions,” says Patrick Hoffmann.
Without further decarbonization of the electricity supply, global emissions from data centers would rise to 643 million metric tons of CO2 by 2030. With ambitious decarbonization efforts, however, they could be limited to 329 million metric tons of CO2 despite growing computing power. At the same time, emissions from the supply chain—particularly from the production of semiconductors and servers—would become more significant.
Water Consumption Is Rising
In addition to energy consumption, the study focuses on water demand. Data centers worldwide are projected to consume around 814 billion liters of water in 2025. By 2030, demand could rise to 1.3 to 1.8 trillion liters. About three-quarters of this is indirectly attributable to electricity generation, with the remainder coming from cooling and semiconductor manufacturing. The expansion of data centers in regions with severe water shortages is considered particularly critical.
AI Can More Than Offset the Additional Consumption
According to the study, AI applications currently account for about 15 to 20 percent of data centers’ electricity consumption. By the end of the decade, their share could rise to around 40 percent.
However, the authors also see this as a major opportunity: if widely adopted in industry, the energy sector, buildings, and transportation, AI could reduce global CO2 emissions by 1.4 billion metric tons per year by 2035, thereby more than offsetting its own emissions. This, however, would require the rapid expansion of renewable energy, more robust power grids, and greater transparency regarding data centers’ energy, emissions, and water consumption.
“AI has the potential to reduce net emissions—provided that its efficiency gains scale faster than the expansion of infrastructure. The transformation of energy systems is crucial here: The key to ‘Green AI’ lies in clean electricity,” concludes climate economist Hoffmann.
The full study,“Code, Carbon, Kilowatts,”is available on the Allianz Trade website.
Author: Katia Meyer-Tien