Let’s Talk about Data Centers
- Kimberly Worsham
- 12 minutes ago
- 10 min read
You've used a data center this week. Every chatbot query, streamed show, and cloud backup runs through one. Data centers have quietly powered our lives for decades. But AI changed the stakes. The race to build ever-larger models has turned these once-invisible buildings into some of the most contested infrastructure in the country, largely over how much water and energy they consume. There's a lot of noise out there right now, and the information is scattered, contradictory, and often confusing.
At FLUSH, we wanted to create a primer blog to help create a starting point for our readers to understand why data center conversations are complicated.
Plenty of articles will tell you data centers are thirsty. Fewer explain why the water piece is so hard to pin down — who's accountable, what's actually measured, and why the numbers you're reading rarely agree. That gap is where we work.
What even is a data center?
Data centers are buildings that hold technology that provides computing, storage, and digital services. In essence, it’s the "compute and storage" behind your cloud services and online activity. They can be massive campuses or just normal office buildings. And, perhaps most importantly, we’ve had data centers for decades.
Used well, data centers underpin real public goods — smarter city planning, faster access to health records, information shared across borders in minutes rather than months.
There are three main tiers of data centers we consider.
Hyperscalers (think Apple, Google, Meta, Microsoft): facilities at gigawatt scale; daily decisions here can move millions to billions of gallons of water a year
Middle-tier operators: run infrastructure for many clients, don't own the workloads, and are often less transparent about water data. These are a kind of blob of hundreds of faceless companies that share ownership and management of data centers for operational use.
Smaller/standalone centers: universities, utilities, local governments, even hospitals have long kept server rooms to hold their own data. These have been part of the landscape for years and quietly support daily life. Many are now being retired or folded into the cloud through mid-tier operators, but they still make up a large share of data centers.
All of these data center groups are major water and energy users, but their services are critical in the growing demand for digital services and data management. One colleague from Stanford University posited that data centers may eventually become a regulated public utility, much like water and electricity… but right now, we are in the 'wild wild west' of its growth before we regulate and better control its operations. Nonprofit Ceres estimates that data center growth could be exponential in the coming years.
However, as you may have noticed in your personal life, digital systems can get really hot and require a lot of power to run them. Data centers build much of their infrastructure to accommodate the power needed to keep them running smoothly and the cooling systems that prevent overheating. One effective way to cool things down is through liquid cooling – also known as water running through cooling towers.
So, why is this issue important now? While data centers are not a new concept, the demand for computing power driven by artificial intelligence is. Additionally, as concerns about climate change and the depletion of vital natural resources in our communities grow, the locations of these data centers are becoming increasingly problematic. The AI boom is directly driving rising demand for water and energy. However, we lack clarity on the long-term impacts of this demand.
The resource math
Reactions about data centers vary by town. Some welcome the tax base because data centers can improve economic outcomes, while others don't. Some note that data centers don’t really add much to communities, since they don’t create many jobs and consume more resources.
According to the Brockovich AI Data Center Reporting Center, the main concerns around data centers (and the AI boom’s influence) are:
Increasing and heavy use of energy that can break the grid
Increasing and high use of water that depletes resources for others
Risky placement of centers in environmentally sensitive areas
Growing noise pollution that disrupts people’s lives
Increased electronic waste because of frequent upgrades and replacements
Let’s focus on energy and water, since those are the most common complaints in much of the media.
Energy Consumption Strain
Energy-wise, demand is high. Some reports estimate that electricity demand for data centers will be 130 gigawatts by 2030; for context, that could power 97.5 million average homes in the US. Much of this data center energy boom is driven by AI, which accounts for about 20% of total data center energy use in 2025. Using this much energy could burn out grids and increase communities' carbon footprints by several million tons.
A lot of energy is also generated by hydropower, which will tie into the next section, but there are systems that can cool data centers using less water but more energy. Also, many of the designs that raise electricity demand can, in turn, increase indirect water use, pushing the resource intensity down to other parts of the value chain. Water-cooled data centers consume about 10 percent less energy than air-cooled data centers. The trade-offs between increased energy and water use are frequent, and centers have to weigh decisions among effective, efficient systems and costs.
Water Use Complications
As the news has clearly noted over the past year, data centers are thirsty places, primarily for cooling the digital infrastructure. One expert blog wrote that large data centers can consume up to 1 million liters a day, equivalent to the use of a small town of about 5,000 people. And we know that there are currently about 11,000 data centers globally, though that number changes practically by the day.
Water use is across the entire chain of data centers, though. Ceres put it well in a recent report:
Data centers require large volumes of water to operate, and water use is mostly characterized as either “direct” or “indirect.” Direct water use refers to the water consumed on-site. Cooling systems account for the majority of a data center’s operational water withdrawals. Depending on size, a data center uses anywhere from 18,000 gallons (68,000 liters) to 550,000 gallons (2.1 million liters) — equivalent to the daily water use of approximately 60 to 1,833 U.S. households. In step with the growth of data centers, water consumed in data center operations in the U.S. increased from just over 5.5 billion gallons (21 billion liters) in 2014 to almost 17.5 billion gallons (66 billion liters) in 2023.
Cooling is part of it (about 75% of consumption), as well as hydropower used for electricity. There’s also concern about the high amounts of water used further upstream, where their chips are made (and more water is needed for newer chips that support AI). However, adoption of less water-intensive systems, such as cooling, is uneven across data centers, as regulations don’t currently require them to become more efficient… and smaller data centers don’t have the financial margins to prioritize efficiency upgrades. And there is no one-size-fits-all solution for data centers to maximize water efficiency. Global Water Intelligence and Xylem created a few useful visuals below that show increased water demand for data centers and where that water is used along the way.
In Newton County, Georgia, some proposed data centers have reportedly requested more water per day than the county uses in a day. Officials there now face tough choices: reject new projects, require alternative water-efficient cooling systems, invest in costly infrastructure upgrades, or risk imposing water rationing on residents.
What’s the problem with data center growth?
There are many reasons these data centers cause concern, as well as why their growth is almost unchecked right now.
Push for more digital systems, but not enough data to create guardrails
With increased AI use, the growing data center footprint is also unevenly distributed – many areas with the space and interest in expanding data center footprints are often increasingly water-insecure. And as climate change makes water shortages more acute, the fight for water access is likely to intensify.
A NY Times article in 2025 noted that data centers in Georgia had damaged local wells, heightening community tensions with Meta as conversations about water shortages became real, even though the data centers' water use was 10% of the county’s total. Waterloop also shared that, in Loudoun County, VA, even with over 200 data centers, the utility reports that this sector currently accounts for approximately 15% of its total water usage.
Part of the problem is how loosely government systems have historically managed water consumption. For example, in many US states, data center permits lack water-use limits, and some municipalities or data centers may not accurately measure their water consumption or even require consumption to be collected. A widely cited 2016 industry report found that fewer than 1 in 3 data center operators tracked their own water use.
It's not a priority. And that's a problem.
This isn't unique to data centers — loose water permitting lets most industries treat water as an afterthought. But data centers are forcing the issue, because their demand is too big and too fast to ignore.
Water used over here is put back over there
Additionally, water issues tend to be dependent on local context, but that’s not something global corporate strategies always navigate well. In personal experience, many data centers – particularly those that are medium or small in size - don't truly understand the intricate workings of their watersheds and the stakeholders in them.
What often happens is that corporates create "water-positive" pledges for their entire global footprint, but targets are hard to disaggregate the more local we get. These pledges don't always hold up because offsets often land in a different watershed than the one being drawn down.
And while some local government officials assure that data centers aren’t that draining on their water supplies, this isn’t always the case in all of the areas where data centers are popping up. Do we really need to keep pushing data centers into Phoenix, AZ, where growing water shortage concerns are well documented?
Are you using too much AI?
One of the key cruxes around the data center conversation that is often overlooked, though, is our individual role in the push for more. Ultimately, it's about meeting consumer demand through usage, alongside industry advances. We need people to understand their consumption habits so they can use their resources more responsibly.
Data center growth is driven by increased demand from all of us to digitize our lives – for better life management, to minimize our material “stuff”, and for entertainment. And yet, while we are rightfully concerned about data center resource consumption, often we divorce ourselves from the growth of an industry.
Do we really need AI to tell us our horoscopes, give us therapy, and plan our trips?
What do we actually need in data centers, personally… and how much of it is waste or frivolous?
What is our tolerance in the balance between convenience and sustainable personal life practices?
An MIT article said, “An everyday user doesn’t think too much about [this]. The ease of use of generative AI interfaces and the lack of information about the environmental impacts of my actions mean that, as a user, I don’t have much incentive to cut back on my use of generative AI.”
Our advisor Manohar Patole is leading an initiative at NYU to better understand how this all comes together in our personal lives – beyond AI, including cloud use, streaming, and all things that require data center resources. "We rely on this metaphor of 'the cloud,' but there's nothing ethereal about it," his team explains. "These are physical facilities with real impacts on real communities, and with the AI boom, those impacts are growing exponentially."
The AI boom is pushing our understanding of data centers’ water-intensive
Another issue, as the UN has warned countries in recent news, is that AI has accelerated our need to address the environmental costs of applying AI to every possible thing. They noted that “When a technology scales this quickly, unintended social, economic, geopolitical and environmental consequences can accumulate quietly, and then become complex and inequitable to correct once systems, investments and dependencies lock in.”
People are rightfully concerned about this growth. Protests about data center plans are happening worldwide, and some governments are suing data centers that are not being transparent about their water use. News articles aren’t always helpful, as they tend to rely on big headlines without providing enough context about data centers, raising more alarm even when we don’t have the data to really understand what we are upset about.
How the tide is shifting
Fortunately, this AI conversation has prompted society to begin interrogating our resource consumption and to initiate changes. And people are starting to inquire: where are these data centers living, anyway?
Communities are pushing back against data center growth, prompting governments to pause and reflect on how to better engage with and regulate data centers when they do come around. It’s, in many ways, a forcing mechanism to get government agencies to update their permitting requirements and to require transparent use of water data from their industrial users.
And when data centers are inevitable in an area, the Lincoln Institute for Land Policy shared at a recent conference that it’s key to get decision-makers to start asking now, “If the negative impacts can’t be completely eliminated over time, what can communities do to manage them?” Actual community impacts vary and depend on a complex set of environmental, technological, and policy factors, and we need not blanket our responses to this question to make it easier to answer.
Amid all the media conversations about water use and data centers, St. Louis, which already has 12 data centers, recently approved a conditional permit for a new data center in April 2026. This permit included a laundry list of conditions for the data center to address, such as reducing noise, ensuring sustainable resource use, implementing closed-loop cooling systems, paying new water rates based on the large loads, contributing to local environmental activities run by the city, and investing in public infrastructural improvements.
Tech, finance, real estate, local government, sustainability — data centers now touch all of it, as an asset class, a permitting fight, a climate liability, or a line item in a town budget. So what needs to happen? Two things, aimed at two groups.
Governments have to regulate. Not because companies are villains, but because transparency doesn't happen voluntarily — firms disclose water and energy use when they're required to, not before. Nearly every honest number in this piece exists because someone was compelled to report it, or a journalist pried it loose. Voluntary pledges and "water-positive" commitments are not a substitute for permits with real limits and mandatory, verifiable reporting. If we want honest conversations, we have to make disclosure a condition of operating.
The rest of us have to get more literate about our own resources. Where does your water come from? What is your watershed already under stress from? Data center demand is, at root, our demand — for storage, streaming, and AI we often don't need. Understanding that is the first step to asking better questions of the companies and officials making these decisions on our behalf.
