The Growing Thirst of Technology: Water Needs of Data Centers and AI

Hello, Reader.

Lemons and blue jeans may seem unrelated, but they share a common need: a lot of water. From agriculture to technology, an incredible amount of water supports their production.

For instance, growing a single lemon requires about 13 gallons of water, while a 16-ounce steak demands a staggering 475 gallons. Producing just one pair of Levi Strauss & Co. (LEVI) iconic 501 jeans uses nearly 1,000 gallons of water.

In the realm of technology, the water consumption associated with artificial intelligence (AI) can be hard to assess directly, but it is massive and rapidly increasing. Although AI systems themselves do not consume water, the data centers that support them require significant amounts of water for cooling purposes. This water is crucial for preventing overheating in servers. The energy used for these operations also indirectly involves large amounts of water.

On average, an Alphabet Inc. (GOOGL) data center consumes 450,000 gallons of water daily. However, newer facilities and some semiconductor fabs can consume up to ten times that amount.

For example, Taiwan Semiconductor Manufacturing Co. Ltd. (TSM) is constructing a cutting-edge chip fabrication plant in Arizona that will use a whopping 4.75 million gallons of water every day—enough to meet the needs of about 1.5 million typical households.

The essential operations of “the internet” depend on millions of servers that run continuously in data centers around the globe. These servers not only require electricity but also consume vast amounts of water.

People often overlook how water-intensive advanced AI programs like ChatGPT can be. Research by Shaolei Ren, an associate professor at the University of California, Riverside, reveals surprising insights into this water usage. His research findings note:

GPT-3 needs to “drink” (i.e., consume) a 500ml bottle of water for roughly 10-50 responses, depending on when and where it is deployed. These numbers may increase for the newly-launched GPT-4 that reportedly has a substantially larger model size.

Although the tech industry’s water use hasn’t drawn as much attention as its energy consumption issues, the growing water demands of data centers will soon pose significant challenges and costs. Despite early indicators of potential water shortages, major tech companies are pressing forward with their expansion plans, leading to a surge in water consumption akin to “Old Faithful” erupting.

In fact, Google’s water usage rose by 17% in 2023 compared to the previous year.

According to Synergy Research Group, the global capacity of hyperscale data centers is expected to nearly triple over the next six years. While new facilities will contribute, the shift from CPUs (central processing units) to GPUs (graphics processing units)—especially those made by Nvidia Corp. (NVDA)—will account for most of this growth.

GPU servers have rapidly become the standard in data centers due to their superior performance in handling AI processes. Unfortunately, GPUs consume about four times the electricity of CPUs, resulting in higher heat production. As AI technologies gain traction, there will be a greater demand for data centers filled with powerful processors.

This situation necessitates more effective cooling systems for data centers, with water-cooled systems being the preferred option. Although air cooling can suffice, water has a thermal capacity 3,000 times greater than air, making it more efficient and economical in heat absorption.

While the data center industry’s water consumption is not at crisis levels yet, it is critical to recognize that as global water supplies become more stressed, the costs for water usage by data centers are likely to rise.

This growing concern is beginning to create investment opportunities. Recently, I recommended an emerging water management company that is poised to benefit from this trend to my subscribers of Fry’s Investment Report.

This company stands in a strong position to seize this mounting opportunity. To learn more about this investment, join me at Fry’s Investment Report today.

Now, let’s look at what we covered here at Smart Money this past week…

Smart Money Roundup

My Urgent Election Debrief

Following last week’s election results, I’ve recorded a message for investors. It’s essential to not get too wrapped up in the historical ties between politics and financial markets. While political events can affect markets, focusing on the dominant market trends of the moment is crucial.

Trump’s Presidency: Impacts on U.S. Chip Manufacturing

Intel is struggling to keep up with competitors, much like the tortoise and the hare. The Biden administration introduced the CHIPS and Science Act in 2022 to support America’s last high-end chipmaker. However, with Donald Trump’s return to the White House, the future of that plan appears uncertain.

Why This Industry is a Solid Investment for Coming Years

With Donald Trump back in the White House and Republicans regaining control of the Senate, he will likely face little opposition in pushing through his economic agenda over the next two years. This shift raises questions about which sectors will flourish. I suspect strong growth lies ahead for this industry.

Estimating the “Trump Bump”: Potential Stock Gains for 2025

According to InvestorPlace analyst Luke Lango, we’ve seen two main effects of the “Trump Bump” during his first term. First, more pro-growth policies and tax cuts led to increased corporate earnings. Second, higher investor confidence in the economy drove stock valuations up. In the latest guest issue, Luke details these factors and outlines how an election-driven rally might elevate stocks in 2025.

Looking Forward

In the next edition of Smart Money, I will focus on water consumption trends in data centers globally and highlight a U.S.-listed company addressing this growing opportunity.

Best Regards,

Eric Fry, Smart Money