When the first news about the arrival of data centers in Querétaro started almost a decade ago, most of us saw it as an opportunity. I work in infrastructure and energy, so I understood what it meant: foreign investment, employment, modernization. World-class technology companies were investing in our region, transforming it into a key digital node for Latin America.
Over time, I began to see the effects of this new scenario more closely. Local authorities reported that water restrictions would become more frequent (such restrictions already existed). At home, the cost of electricity also increased. My curiosity about the environmental impact of these data centers led me to investigate further, and I found an article from the Inter-American Development Bank (IDB) that posits a key question: Does LAC Have the Energy It Takes to Become a World Leader Host of Data Centers? The authors address the challenges related to the growing demand for energy and the need to improve energy infrastructure and regulatory policies to meet this demand. In my quest for answers to this question, I learned the following:
The cost of sustaining the digital world
Whenever we send a message, save a file in the cloud, or conduct an online transaction, a server somewhere in the world processes that data. According to the BBC Science Focus, a typical email generates 4 grams of CO₂ emissions and, if the email contains an attachment, the carbon footprint rises to 50 grams. These servers cannot be turned off, meaning they must be constantly cooled. The cooling methods vary:
- Industrial air conditioning: Works well in temperate climates but requires higher energy consumption in hotter regions.
- Liquid cooling: The method uses water instead of air to dissipate server heat, improving efficiency but increasing water consumption.
- Subsea data centers: This idea is under development, and intends to utilize ocean water for cooling without harming local resources. While promising, it has not yet produced scalable results. An example of this is the now-concluded Project Natick.
To illustrate the magnitude of the impact, Netflix’s total carbon footprint in 2023 was 896,871 metric tons of CO₂e, of which 44,844 metric tons of CO₂e (5%) were attributed to streaming emissions. However, internet streaming and end-user device emissions are not included in Netflix’s reported footprint. In comparison, Amazon’s total carbon footprint for the same year was 68.82 million metric tons of CO₂e. Considering these figures, the estimated global emissions from streaming alone reported by Netflix in 2023 would be equivalent to driving a gasoline-powered car for 5.27 billion kilometers. This distance is roughly equal to nearly 7,000 round-trips to the moon.
Water scarcity and the impact of data centers
So far, it appears that my community’s concerns about the increasing water restrictions in Querétaro are supported by studies such as “The environmental footprint of data centers in the United States.” This study finds that data centers consume substantial amounts of water for their operation, directly through liquid cooling and indirectly through electricity production.
Water consumption by data centers can create significant challenges, especially when these facilities are located in or dependent on water-stressed watersheds, leading to an ongoing debate about the sustainability of these facilities. Current data indicates that approximately 20% of data center servers draw water directly from watersheds with moderate to high water stress. Additionally, nearly half of these servers are partially or completely powered by power plants in water-stressed regions. In 2021, the average Google data center consumed about 450,000 gallons of water daily.
The impact of artificial intelligence on resource consumption
The expansion of artificial intelligence (AI) leads to increased data center resource consumption. The infrastructure that supports AI models, particularly large language models (LLMs) and deep neural networks, requires even more power and water due to the intensive computations that must be performed continuously. Thunder said energy estimates that energy consumption could double by 2030, largely due to AI, which could add 1% to global energy demand and 2.5% to global electricity demand.
According to the World Economic Forum, a 1-megawatt (MW) data center can use up to 25.5 million liters of water annually for cooling alone, equivalent to the daily water consumption of approximately 300,000 people. AI’s growing energy demands further drive water consumption in data centers. A single data center can consume up to 5 million gallons of potable water daily, enough to supply thousands of homes or farms.
The need for clear information
My family has always had to adapt to changes in water access since we live in a water-stressed region. However, in recent years, the differences have become more apparent: restrictions have become stricter, wells need to be dug deeper, and the cost of providing water to the city continues to rise.
The absence of accessible and transparent information regarding the actual impact of data centers on local resources has become a significant issue. In many areas, ESG reports fail to provide a clear breakdown of water consumption by specific location, hindering an accurate evaluation of their effects in water-stressed regions.
The previous study highlights that in some European countries, such as the Netherlands, a lack of transparency regarding water usage has led to temporary operational shutdowns when local communities discovered that actual water consumption greatly exceeded initial estimates. Additionally, the study indicates that focusing solely on energy consumption without considering water usage can have negative consequences. For instance, some data centers have managed to reduce their electricity use by improving cooling efficiency, but this has resulted in increased reliance on water. Without clear and detailed data, it is a challenge to make informed decisions about the sustainability of these operations. In Querétaro, where water is already a scarce resource, this issue could worsen if companies do not provide stakeholders with more transparent and comprehensible information.
The flip side of the coin
To address the community’s information needs, the company invited us to a discussion about data centers. They provided a different perspective, emphasizing that these centers are crucial infrastructure for everyone’s digital life. People stated, “If a data center goes down, the global impact is immediate”. The community also highlighted that businesses can lose revenue when online payment systems are interrupted or when e-commerce platforms experience downtime. Additionally, healthcare services may be disrupted, leaving hospitals without access to electronic medical records. The extent of the impact varies based on the size and importance of the affected data center, but there will always be consequences.
Despite the challenges, AI offers significant opportunities for optimizing resources. It can monitor energy and water consumption within data centers, helping companies reduce their carbon footprint.
I asked about water consumption in water-stressed regions and was informed that, although it poses a challenge, the issue is being addressed. Strategies have been implemented, such as water recirculation systems for cooling and dry cooling methods, which use heat exchangers to dissipate thermal energy without relying on water. These approaches can be particularly beneficial in arid climates or areas experiencing water scarcity.
Can we have it all?
I fully recognize the benefits of this industry. Querétaro has become a hub of innovation, creating more specialized jobs and opportunities within the technology sector. This trend is not limited to my community; it is happening throughout Latin America. However, these advancements come with costs, and, for development to be sustainable, we need greater transparency regarding these centers’ environmental, social, and economic impact. The issue lies not in resource consumption but in the lack of detailed and timely data required for informed decision-making, as some companies may even fabricate information or sign confidentiality agreements with governments.
This discussion is not about being for or against data centers; instead, it is about collectively reaching a consensus on how we want them to operate in our region. How can we ensure that access to essential resources is not compromised in the name of technology? In game theory terms, what is the payoff matrix of this situation? Can we demand the adoption of best practices by data centers? Do we really possess the bargaining power to influence their operations? How can we promote effective participation from the community and other key stakeholders in decision-making processes? Can we improve our behavior as users of technology? These are questions without clear answers. We must start with the understanding that progress and sustainability can, and must, coexist.
Leave a Reply