Brasil
Anais da Academia Brasileira de Ciências
Tabla de contenido Anterior Actual Siguiente
Tabla de contenido
ECOSYSTEMS • An. Acad. Bras. Ciênc. 98 (4) • 2026 • https://doi.org/10.1590/0001-3765202620260340 linkcopy

Open-access Data centers and Artificial Intelligence: a double threat to worldwide sustainability and a worsening of the recently declared “water bankruptcy”

SCIMAGO INSTITUTIONS RANKINGS

Abstract

The fast growth of open AI together with the growing demand for new data centers to support AI are clear demonstrations of advancements of the uses of new technologies and very often this positive vision masks the impacts these new technologies play upon society and the environment. The high energy and water consumption are two major examples to be considered in a fair and detailed evaluation of their role and costs to society and the environment. What is proposed is a detailed analysis of the impacts and consequences of these new technologies to society and the environment, particularly in relation to the conservation and sustainable use of water and energy. What is expected is that this analysis brings a leveling and indication of limits to the growth and uses of these technological infrastructures together with further clarification of the existing regulations thus guaranteeing they don’t push further the already surpassed planetary limits.

Key words
Data Centers; Artificial Intelligence; Water; Energy; Water Bankruptcy; Sustainability

INTRODUCTION

The growth of AI is spectacular and brings a disguised view of what shall happen with water and energy consumption, two vital elements for the vertiginous growing of Big Techs worldwide. Moreover, and despite what is known on this matter, further discussions and actions are needed to provide alternatives to attenuate the indiscriminate use of these resources, disregarding the fact that both are two limited worldwide resources. The concern on this high use of water and energy is even reinforced when we consider the Big Tech’s growing influence on and integration with Governmental bodies, as pointed out by Goudarzi (2025) which demonstrated how Big Techs redefined governance and economy in 2025. Having all this power and influence will not prevent them of using as much water and energy as they need and this shall result in weakening the concern with a sustainable use of these resources.

Data centers, in order to maintain ideal operating temperatures (10oC to 27oC), consume significant volumes of water and often cause disruptions to local communities and ecosystems (Hogan 2015). As reported by Barbosa et al. (2025) and according the Google’s 2023 Environmental Report the company used 21 billion liters of water, a 20% increase from 2021. Li et al. (2023) estimated that the training of open AI’s GPT-3 model (Generative AI chatbot by open AI) would have consumed 3.5 million liters of water if conducted on servers of the US and 4.9 million liters if hosted in more congested datacenters in Asia. The new model GPT-4 probably has consumed grater volumes of water and energy, consuming no less than 0.5 L of water for every 20-50 question-answer sequences. Barbosa et al. (2025) called attention for the fact that the highwater consumption by these infrastructures raises significant concerns about sustainability and water security and has even sparked protests worldwide as the ones in the Netherlands (van Kessel 2021) and Uruguay (Viana 2023) fearing that the volume of water consumed by data centers would threaten countries water supply.

Water has a centrality in human actions and natural resources equilibria, a centrality captured by the United Nations Sustainable Development Goal 6 (SDG-6) which ultimately is the resource that guarantee food and energy security, biodiversity, environmental resilience, public health, climate stability, and peace (Madani 2026). Moreover, the world is still far from meeting SDG-6, and any other use or demand from society that makes this objective to be achieved must be considered carefully, having in mind the natural constraints for water availability such as the facts that 2.2 billion people still lack managed safely drinking water, 3.5 billion lack safely managed sanitation, and approximately 4 billion people experience severe water scarcity at least one month per year (Madani 2026). To complete this picture, not less than 75% of the world’s population live in countries classified as water insecure or critically water insecure with progress toward SDG-6 being far off track for 2030, indicating that water related risks are now systemic. The central message of the “United Nations University Institute for Water, Environment and Health’s Report is “the world has entered the era of global water bankruptcy” (Madani 2026).

The following are worldwide major consequences of this water bankruptcy: rivers that no longer reach the sea (e.g., Colorado River, Yellow River, Amu Daria River); shrinking or disappearance of lakes, swamps and glaciers (e.g., Huyana Potosi Glacier, in Bolivia; Sperry Glacier in Montana-USA; Aral Sea in Kazakistan/Uzbeckistan; Lake Chad, in Africa; Mesopotamian Marsh, in Iraq); land subsidies and salt intrusions due to excessive pumping of aquifers (e.g., Mexico City, Jakarta, Indonesia, Sonora, in Mexico, Chennai, Tamil Nadu, in India); drying and burning of forests and peatlands (e.g., Indonesia, Sumatra, and Borneo); expanding of deserts (e.g., Gobi and Sahara Deserts, Southwestern US) and dust storms (Mongolian-Chinese in 2023 and Australian dust storm in 2020). Moreover, according to the laws of nature described by thermodynamics, any transformation of energy from one type to another (1st. law) generates disorder in nature (2nd. law).

The recent declared water bankruptcy, along with the impacts mentioned above, prompted us to highlight the urgent need to address the growing demands of data centers and artificial intelligence, while seeking effective mitigation strategies.

The impacts of data centers

According to a recent ABI Research (2026) (accessed on 16 March 2026) market analysis by the end of 2026 there will be 8,821 data centers worldwide and the projections are more than 10,000 will be operating by 2030 and reaching 12,000 facilities by 2025. This number is increasing rapidly driven by global demand by AI, cloud computing and 5G networks. Particularly for Brazil, a promising area considering mainly its water reserves, abundant energy supply and its own needs to use AI, the number of these infrastructures is estimated in 195 facilities (Statista as of November 2025). Since data centers are vital for AI, these figures shall grow fast worldwide and particularly in Brazil which has been seen as an attractive area to hold new data centers mainly driven by large local demand and clean energy, as judged by the large investments announced by Microsoft’s (US $14.7 billions) and Amazon’s (US$ 1.8 billion) (IEA 2025). Examples of new data centers installed in Brazil are the data center linked to Tik Tok in the state of Ceará (https://www.intercept.com.br/2025/07/03/data-center-tiktok-energia-estudo-interno/) and the ones installed in the state of Sergipe (Optimus Technology Group; Smart data center) and in the city of Uberlândia, in Minas Gerais (RT-One and Algar Telecom-Ascenty Partnership).

Brazil has approved in July 2024 the “Brazilian Plan” for AI to be implemented along 2024-2028 entitled “AI for the good of all”, as the national strategy to transform Brazil into a global leader in the development and ethical use of technology. As objectives and differences of this Plan are cited: i) Sovereignty - to develop advanced language models (LLMs) in Portuguese to reduce its dependence on foreign technologies; ii) Sustainability - Leveraging Brazil’s renewable energy matrix to create data centers with low environmental impact; iii) Funding - Resources come from a combination of public funds (BNDES and FINEP) and partnerships with the private sector Brazil Ministry of Science, Technology and Innovation (2024).

There are however difficulties and oppositions to the worldwide “explosion” of data centers as reported by the Governor of Florida — Mr. Ron DeSantis calling attention for the impacts of these projects on the electricity bill and the use of subsidies and the need to review the present regulations. Furthermore, there are also local communities that do not want having these technological facilities within their backyards and that blocked or delayed US$100 billions in projects in the second quarter of 2025 Data Center Watch Lab – 10 Labs (2025).

The rapid data center expansion has sparked global debate over resource use, particularly energy, water and land. Energy and water consumption are now the central environmental and regulatory concerns — global data centers electricity use is projected to 1,000 to 1,050 TWh by the end of 2026 what is more than double the levels of 2022. Globally and only in 2024 Data Centers consumed approximately 1.5% of electricity consumption (IEA 2025). The growth of AI shows a direct dependence of data centers and this increases the impacts on the environment. AI-focused servers consume 4-7 times more power than traditional hardware. Moreover, data centers emissions of CO2 have increase up to 300% since 2018. Roughly 56% of data centers energy still comes from fossil fuels globally. Data centers energy needs are highly localized and projected to account for 30% to 79% of total electricity consumption by 2028 as are the cases of Northern Virginia in USA (25%) and Dublin in Ireland (20%) (Bresnihan & Brodie 2025).

Concerning water consumption, a large data center can consume up to 5 million gallons (c. 18,925,000 L) of water daily, roughly the same as a town of 50,000 people. This is further worsening when we consider that approximately 2/3 of the new data centers built since 2022 are located in water-stressed regions and by 2030 the global data center water use is projected to reach 450 million gallons per day (1,703,250,000 L). And no less than 80% of the water used by these facilities is lost to evaporation (Yañez-Barnuevo 2025).

In the case of Brazil and for 2024 data centers used 1.7% of national electricity consumption (~8.2 TWh out of 650.4 TWh consumed in the country) and this shall more than double by 2029 (Brasscom 2025). The estimated water footprint of Brazilian data centers for 2025 is of 0.003% of the national total consumption (64.5 trillion of liters) what is approximately 2 billion liters, an amount consumed by approximately 34,900 people. This water footprint consumed by data centers in Brazil shall be of 0.006% in 2029 (30.4 billion liters) (Brasscom 2025). Brazil is now debating a national data centers policy in response to public protests concerning the potential competition of these facilities with local communities for resources during droughts, despite the companies inform that they are moving toward closed-loop cooling systems.

Worldwide and due to the high water consume of data centers rises significant concerns about sustainability and water security that has sparked water protests in various regions globally, of which are examples the public demonstrations in the Netherland (van Kessel 2021), in Santiago, Chile impeding Google to install a new data center Lais Martins – Intercept Brasil (Martins 2025) and in Montevideo, Uruguay where public opposition emerged against Google’s plans to install a data server on a 29 ha site (Viana 2023).

Mitigation measures and the sustainable use of water and energy

The boost of data centers and use of AI will continue to grow to attend the fast-growing demand worldwide. The projected growth of AI demand shall result in 4,200 L to 6,400 L of water by 2027 and the global data centers water consumption is projected to reach 1,200 billion L annually by 2030 Nicoletti, Ma & Bass, for Bloomberg Technology + Green, 8 May (2025). To face this water consumption constraint, major tech firms have introduced new policies in January 2026, in response to backlashes and US$64 billion in cancelled projects due to resource concerns. Examples are i) Microsoft’s “Community-first initiative” which has committed in 40% improvement in water-use intensity by 2030 besides pledging to replenishing more water than it consumes; ii) zero-water cooling, when new data centers designs are shifting toward closed-loop liquid cooling and direct-to-chip cooling, methods that can reduce water consumption by up to 90%–95% compared to traditional evaporative methods; iii) water positivity (https://www.theguardian.com/environment/2021/oct/14/water-positive-pledge-corporations) where Google and Amazon confirm their compromise to be water positive targets by 2030 and are investing in local watersheds restoration and using recycled sewage for cooling Spindler, Fisher and Hans-Petersen – Centre for Nature and Climate, World Economic Forum (2025) (https://googlediscovery.com/2023/08/03/consumo-de-agua-nos-data-centers-do-google-aumenta-20/).

What needs to become clear is that the expansion of AI is creating a “water-energy-nexus” crises, particularly in arid regions, namely southwestern USA, parts of India and Latin America Nicoletti, Ma & Bass, for Bloomberg Technology + Green, 8 May (2025). Moreover, in the particular case of southeast Brazil, where 195 data centers are located, Vinhedo, a town distant 75 km from São Paulo declared hydric emergency in May 2025 and had to adopt water rationing schedule. Despite this serious water constraint, Vinhedo holds two data centers (Ascenty Enterprize) who is considered the largest data centers campus of Latin America, demanding not less than 61 MW of energy CEMADEN, Climainfo, 14/01/2026. (https://climainfo.org.br/2026/01/14/reservatorios-de-agua-do-sudeste-nao-se-recuperarao-neste-ano-preve-o-cemaden).

Data Centers, AI and the Planetary Boundaries: A Nexus to be Recognized

The present regulations on these new technologies very likely are weak, not comprehensive enough, and not fully understood or implemented, since the demands for these facilities are huge and fast growing. Moreover, being new areas there must be tremendous competitions which may delay regulations and implementation actions.

Seven of the nine planetary boundaries — the thresholds that keep life on Earth within a safe operating zone or safe boundaries (PBScience 2025) — are now breached and all of them show worsening trends. Freshwater change is one of them, together with ocean acidification, land system change, climate change, modification of biogeochemical flows, introduction of novel entities, and change in Biosphere integrity (PBScience 2025). According to the latest assessment of the planetary boundaries conducted by the Potsdam Institute for Climate Impact Research, human impacts on rivers and soil moisture is growing and pushing water systems further from stability and heightening drought and flood risk, for what the main drivers are irrigation and agriculture, industrial water use (e.g., data centers), household water use, climate change, increase in atmospheric aerosol loading, and land system change.

An obvious conclusion is that societies are not paying due attention to the meaning and consequences of these transgressions. As demonstrated above, among the industrial water uses, data centers and AI demands are now major players in water use and will grow further to attend the continuous increasing demands of modern society. To contribute to safeguard Earth resilience and stability these “new water uses” must contribute to bring the planet back to its planetary boundaries what in practical terms means to invest in saving water and energy. Will they manage through the implemented initiatives listed above?

We believe that more is needed in terms of regulations and investments in saving energy and water considering their fast-growing demands to feed the even faster augment of data centers and use of artificial intelligence. However, the discussion about investments runs a serious risk of becoming anchored in techno-solutionism since often, saying that investment is necessary is just an excuse to generate new businesses (more profit) and solutions that create new problems and demand new technological solutions, which in turn will generate more profit for the “green entrepreneurs”. Moreover, this can be a negative social result beside the new business what can result in augment inequality and thus augment poverty.

We believe that these new technologies need further governance and regulation, and they must accept that in planning their growth they have to operate within the planetary limits. Particularly concerning the use of water, it is important to stress that more than a fifth of global land area now faces significant dry or wet deviations in stream flow (22.6%) and soil moisture (22%) roughly twice the preindustrial-like baseline state and far beyond the safe levels (12.9 % and 12.4%). Moreover, both indicators are rising and putting freshwater in the Zone of Increasing Risk (PBScience 2025). This use cannot growth indefinitely and other water uses must be considered, among which the high consume by irrigation, agroindustry, mining, industrial and household deserve attention and further investigation in order to save water and energy.

FINAL REMARKS AND CONCLUSIONS

When we started writing this manuscript our understanding was that it could act as an alert disclosing the potential risks and impacts the ever-increasing demand for energy and water by data centers has already become. Moreover, it was also our intention to demonstrate that these new technological infrastructures together with the growth of AI need to be better evaluated and framed within reasonable limits, namely the safe demands of water and energy, compatible with the planetary limits. As time passed and new information was gathered, we conclude that adaptation measures offered by this sector will not be sufficient to restrain the continuous growth of more data centers and demands by artificial intelligence. This conclusion became clear when we found the report prepared by a team of experts from the United Nations University Institute for water, environment and health – UNU-INWEH lead by Kaveh Madani (Madani 2026). Quoting this report “according to regions and levels of development, water systems are under unprecedented pressure. Rivers, lakes and wetlands are degrading, groundwater resources are being depleted beyond sustainable limits, and glaciers are retreating at accelerating rates. These trends signal not only stress but, in many contexts, a structural imbalance between water demand and available resources.” Madani (2026) called this condition water bankruptcy and call for effective action to protect water-related natural capital before damages become irreversible, a condition that according to him, the planet has entered the Water Bankruptcy era.

Corroborating this, we conclude that the huge growth of data centers and artificial intelligence demands will continue to grow taking as an example the recently approved bill project in Brazil – PL 278/2026 – which suspended federal tributes to attract big data centers to the country mainly to feed cloud computation and AI through the creation of a Special Tax Regime for Data Center Services (Redata). This bill project is now to be approved by the Senate and its text does not include any important socio-environmental safeguarding as reported by Priscila Pacheco on the Observatório do Clima (Pacheco 2026).

There is an urgent need to implement a global governance and regulation to data centers and the growing demands of AI. This will guide the initiative at country’s level but similarly as what has been suggested previously for the water — a global polycentric and networked governance Barbosa et al. (2025) — will have to be discussed and hopefully agreed globally, considering the regional differences and demands which still have to be defined having in mind the planetary limits, particularly those for water.

It seems appropriate to quote James Bridle in his instigating book of 2022 Ways of Being “our own survival depends of our capacity to establish a new pact with the more than human world in which the intelligence, the innate essence of all things — animals, plants, machines — be seeing not as other insinuations of our superiority but as indications of our total interdependency and an urgent call for us to be more humbles and caring.”

Data availability

The data set used in this manuscript were published in this article.

References

Correspondence

Correspondence to: Francisco Antônio Rodrigues Barbosa barbosa.ufmg@gmail.com

Handling editor

Patrik Viana

Publication Dates

  • Publication in this collection
    23 Oct 2026
  • Date of issue
    2026

History

  • Received
    30 Apr 2026
  • Accepted
    18 Aug 2026
Creative Common - by 4.0
This is an open-access article distributed under the terms of the Creative Commons Attribution License
location_on
Academia Brasileira de Ciências Rua Anfilófio de Carvalho, 29, 3º andar, 20030-060 Rio de Janeiro RJ Brasil, Tel: +55 (21) 2391-7901 - Rio de Janeiro - RJ - Brazil
E-mail: aabc@abc.org.br
rss_feed Acompañe los números de esta revista en su lector de RSS