As AI hysteria reaches a fever pitch, it's time to look at the evidence

Newer data-center facilities increasingly use closed-loop liquid cooling, wasting less water.

Over the past few weeks, an Anthropic researcher resigned after warning that the people racing to build advanced artificial intelligence could wipe out humanity by the end of the decade. Soon after, Anthropic's Dario Amodei, OpenAI's Sam Altman and SpaceXAI's $(SPCX)$ Elon Musk all endorsed slowing frontier-model development.

To me, the incentives here look suspiciously like regulatory capture: frontier labs advocating rules they can afford, while smaller and open-source competitors struggle with the compliance burden.

But regardless, this sentiment is spilling into the real world, as a recent Gallup poll showed 71% of Americans opposed construction of the physical manifestation of AI: data centers.

Data-center skepticism has also become unusually bipartisan. Sen. Bernie Sanders, the Vermont independent who caucuses with Democrats, and Rep. Alexandria Ocasio-Cortez, the New York Democrat, introduced legislation to temporarily halt construction of new AI data centers. Republicans such as Sen. Josh Hawley of Missouri have pushed bills stripping tax breaks and shifting infrastructure costs back onto data-center operators. Sen. John Kennedy, a Louisiana Republican, has taken a different route, proposing mandatory emergency "kill switches" for advanced AI systems.

Whether you hear from Sanders or Hawley, data centers are now blamed for higher electricity bills, water shortages, environmental damage and relatively few jobs.

But much of this narrative is misguided. Thousands of data centers already operate across America. They are the physical backbone of the internet. Your Zoom (ZM) meeting, Netflix $(NFLX)$ stream, online brokerage account, cloud storage and DoorDash (DASH) order all ultimately depend on servers sitting inside these buildings. When you pulled up this article, you used a data center.

AI simply requires a lot more of them. But that's a good thing, for multiple reasons.

The water story is already outdated

The "AI is taking our water" claim is the most outlandish.

Older data centers did rely heavily on evaporative cooling, much like an outdated car relied on a carburetor. But the new facilities increasingly use closed-loop liquid cooling, which continually recirculates coolant instead.

Microsoft (MSFT) says roughly 90% of its owned data-center fleet now uses low- or zero-water cooling, and its newest AI designs can use no water evaporation for cooling at all. Its massive Wisconsin facility uses closed-loop cooling across more than 90% of the site.

Nationally, U.S. golf courses applied about 531 billion gallons of water in 2024. Direct on-site water consumption by data centers has been estimated at roughly 17 billion gallons annually. That comparison narrows significantly if you include water used to generate their electricity, which is why location and power source matter.

At the national level, direct data-center water consumption is roughly 3% of what golf courses use. So if golf courses aren't causing a national water panic, I'm not sure why data centers should.

The power problem is real - but solvable

This is the biggest political issue right now. The House of Representatives just passed legislation directing utility regulators to consider whether large power users should pay the infrastructure costs created by their demand. It passed 417-3.

That principle makes sense. If a hyperscaler needs a new substation, transmission line or power plant, existing homeowners shouldn't get stuck with the bill.

But there's surprisingly little evidence that data centers have broadly driven residential electricity inflation so far. A new 50-state study from University of Colorado Denver finance professor Yosef Bonaparte, covering 2021 through 2024, found no statistically significant relationship between data-center presence, capacity growth and residential electricity-price inflation. And in aggregate, inflation-adjusted residential electricity prices are still only about 4% to 5% higher than in 2013, despite a roughly 50% nominal increase.

So if data centers are increasing energy costs, it's not showing up in the data.

Local communities benefit tremendously

The AI boom is quietly creating some of the fastest-growing blue-collar jobs in America.

Data-center job postings have more than doubled in two years, even as overall U.S. job postings fell roughly 12%. Hourly installation and maintenance jobs tied to data centers advertise pay about 42% higher than comparable roles elsewhere.

A May 2026 PwC study estimated the industry directly employed more than 1 million Americans in 2024 and supported 5.5 million jobs across the broader economy, up 17% in a single year.

For all the talk about bringing industrial jobs back to America, this is what reindustrialization actually looks like.

The investment opportunity

None of this means every proposed data center will get built.

Power, permitting, financing, transmission and community opposition will kill plenty of projects. And it will be a major political football this coming election cycle.

That's why I'm less interested in simply betting on "more data centers" than identifying what the successful ones cannot operate without.

They include cooling systems from Vertiv Holdings (VRT), switch gear from Eaton (ETN), gas turbines from GE Vernova (GEV), and nuclear and renewable generation from companies like Constellation Energy (CEG) and NextEra Energy $(NEE)$.

I'm adding to an AI data-center stock in my personal portfolio next week. It sits directly on one of the industry's biggest bottlenecks: skilled labor. In the next free issue of my Let's Analyze newsletter on Substack, I'll reveal the company, explain why I'm buying more and share the price I'm targeting.

While everyone argues about whether AI will eventually bring about the apocalypse, hundreds of billions of dollars are being spent building the infrastructure underneath it.

Investors don't need to predict the end of humanity. They need to follow the money.

Robert Ross is the founder of TikStocks and author of "A Beginner's Guide to High-Risk, High-Reward Investing." A former chief equity analyst at Mauldin Economics, Ross writes the investment newsletter Let's Analyze on Substack and hosts the weekly "Room to Run" podcast.

-Robert Ross