AI, the challenge of the future is played on energy and water

Artificial intelligence promises to accelerate the energy transition and offer new tools to tackle climate change. However, it requires ever-increasing amounts of electricity and water to power the data centers on which it is based. It is the paradox that emerges from the eleventh Environmental Report published by Google, in which the group claims record investments in renewable energy and a reduction in operational emissions, while recognizing that the rapid expansion of AI makes achieving climate objectives increasingly complex.

In 2025, Google has signed contracts for more than 12 gigawatts of new capacity clean energythe largest volume in the company’s history and, according to the group, enough to feed a nation the size of Greece once the plants are operational. However, electricity demand grew by 37%, the largest annual increase the company has ever recorded, largely due to the expansion of its AI data centers. The operational emissions they decreased by 2% compared to the previous year and Google claims to have offset 100% of its electricity consumption with purchases of renewable energy for the ninth consecutive year.

@Google cuts operational emissions, but says moonshot climate goals “getting harder”https://t.co/RQomF8Pm4K#netzero #sustainability #cleanenergy #AI #climategoals

— ESG Today (@EsgToday) July 1, 2026

AI emissions

The picture changes, however, if we look at the overall emissions considered for the purposes of the company’s climate objectives (ambition-based emissions), which also include a large part of the supply chain. These increased by 18%, reaching approximately 14.5 million tonnes of CO2 equivalent, driven approximately 80% by emissions linked mainly to the production of hardware and the construction of new infrastructures.

Since 2010, Google has signed over 240 agreements for nearly 35 gigawatts of new renewable capacity and continues to invest in technologies considered strategic for decarbonization, such as advanced nuclear, enhanced geothermal and fusion. According to the report, advances in the efficiency of hardware, software and infrastructure, together with investments in clean energy, have avoided more than 58 million tonnes of CO2 equivalent in 2025 alone compared to a reference scenario developed by the company itself.

The use of water

The group also claims results on the management front water resources. The picture, however, is more complex than just the data on reintegration. Total water consumption rose 34%, reaching 10.9 billion gallons in 2025. That same year, Google-funded projects would return about 7.7 billion gallons to the environment, or 78% of annual freshwater consumption, as part of the company’s goal of replenishing more water than it consumes by 2030.

The contribution of AI to sustainability

A substantial part of the report is dedicated to the contribution that artificial intelligence can offer to sustainability. Google claims that nine of its products, including Google Maps, Nest, Google Earth, Solar API e Waymo, have made it possible to avoid a total of around 41 million tonnes of CO2 equivalent in 2025. The so-called “avoided emissions“However, they remain the subject of debate among researchers and experts, since they are based on scenarios and methodologies that do not always allow independent verification.

The supply chain

In 2025 the emissions of supply chain increased by 25%, mainly due to the construction of new data centers and hardware production in Asia-Pacific countries, where electricity grids still rely heavily on fossil fuels. Google also admits that the expansion of AI infrastructure is progressing faster than the decarbonization of electricity grids, held back by connection times, permitting, bottlenecks in supply chains and limited availability of new low-emission capacity.

The debate on water consumption

The report comes as debate grows over the environmental impact of artificial intelligence. An analysis published by Wall Street Journal draws attention to an aspect that is often less visible than energy consumption: the use of water by data centers. In addition to the water used directly to cool the servers, the newspaper highlights the indirect consumption linked to the power plants that produce the energy necessary for the operation of digital infrastructures, a component which, according to a study cited by the newspaper, for Google could be approximately three times higher than the direct water consumption declared.

The analysis also highlights how water replenishment programs, through which companies finance interventions to restore water resources, do not always compensate for the effects in the areas where the data centers are located. If the water is replenished in basins other than those affected by the withdrawals, observe some experts quoted by the newspaper, the overall balance can be positive without alleviating the water stress of local communities.

The growing demand for electricity and water needed to support the expansion of artificial intelligence is thus opening a new front in the debate on the sustainability of the technology sector. It is a challenge that, according to analysts and researchers, risks putting pressure on climate objectivesof the entire technology sector.