Artificial intelligence (AI) has been a critical driving force behind the current bull market and, given the veritable torrent of artificial intelligence talk by tech companies, one would think that Wall Street would be beating down the doors of any AI investment worth its salt.
And while that’s largely true, there is one AI infrastructure investment that the market has been seemingly ignoring the whole time.
Water.
[text_ad]
There are credible estimates that a single interaction with the various AI chat models (5-50 queries; so perhaps a single question and a few follow-ups or a brief “conversation”) consumes 500ml of (primarily) potable water, the equivalent of a typical bottle of water.
In their 2024 Environmental Report, Google alone disclosed that their water consumption in 2023 was 6.4 billion gallons, up from 5.6 billion gallons the year before.
And they’re not alone. Microsoft saw similar spikes in its own water consumption, which rose from 1.69 billion gallons in 2022 to 2.07 billion gallons in 2023 (a 22.5% increase) according to its Environmental Sustainability Report.
The water consumption in question is owed largely to the proliferation of large (hyperscale) data centers, which consume huge quantities of water for cooling and humidity management, to say nothing of water consumed by chipmakers in the production of those data center chips. (Global chipmaker water consumption is estimated to rival that of the city of Hong Kong and its 7.5 million residents.)
Water use (and thus its role in AI Infrastructure) is ubiquitous amongst large tech companies, with Amazon (for their web arm, AWS), Meta Platforms and Apple also using billions of gallons annually.
Now, to give the (thirsty) devils their due, four of the five companies listed above (excluding Apple) have committed to becoming either water-neutral or water-positive (returning more water to the ground than they consume) by 2030.
But with data center growth (and water consumption) showing no signs of slowing down, there’s a case to be made that parched server farms are turning water into an AI infrastructure investment in its own right.
The downside of taking the water angle on AI is that you’re unlikely to experience the massive speculative upside of a pure-play AI investment such as a software company or chipmaker.
The upside, however, is that it’s a more conservative angle that’s less likely to get caught up in the market’s AI-driven hype cycle.
Plus, it’s more straightforward. It doesn’t matter which platform is the better AI image generator, nor does it matter whether Google’s Gemini (nee Bard) gobbles up market share from OpenAI’s ChatGPT or Microsoft’s Copilot. All of them will be slurping up groundwater for years to come.
Investors looking to put a little water weight into their portfolios would be well-served by any number of ETFs. The First Trust Water ETF (FIW) is a fund we’ve written about before. With $1.8 billion (give or take) in assets under management (AUM), it also happens to be one of the larger water-focused funds. It’s up 13.6% in the last year with an expense ratio of about 0.5% and a modest yield of 0.7%. This fund broadly invests in companies that derive the majority of their revenues from potable and wastewater services.
The Invesco Water Resources ETF (PHO) is an equally large ($2.2 billion AUM) fund with a similar yield and expense ratio (both right around 0.6%) but is a more concentrated portfolio that prioritizes water conservation and purification. It’s also modestly outperforming FIW, up 14.0% in the last year.
The Invesco S&P Global Water Index ETF (CGW) is a slightly smaller fund ($900 million AUM) that has comparable expenses, but a higher yield (1.5%) and lower one-year returns (up 6.6%). CGW invests globally in the largest companies in the water business and thus offers more exposure to utilities (hence the higher yield and lower performance).
Global water management is likely to have long-term tailwinds as the climate changes and, for now at least, it also happens to be the one AI infrastructure investment that the market is ignoring.
[author_ad]
