Sure, Cigarettes Will Cure Your Cold, and Data Centers Will Rebuild Your Rural Economy

It’s been a bad few weeks for big tech billionaires.

“Campaigns or party committees can not fix the toxic brand of an entire segment of the economy.”

That was the conclusion of a private memo from the National Republican Senatorial Committee sent to AI tech donors this month.

Ouch.

The memo goes on to state that Sherrod Brown has made data centers the centerpiece of the campaign against Sen. Jon Husted in Ohio, and it is working. A Fox News poll last week had Brown up 8.

Watch: Sherrod Brown blasts Sen. Jon Husted on data center development as the two prepare to face off this November

The problem data centers are having is that the more voters learn about them, especially the hyperscale campuses built for generative AI, the more they hate them. Gallup now finds 7 in 10 Americans oppose a data center in their own community.

In Florida this week, David Jolly won the Democratic gubernatorial primary. During the campaign, he pledged to halt new data center construction. His Republican opponent, Byron Donalds, is now running on a platform to make data centers leave the public grid. It’s a complete shift from his previous position.

This spring, the pro-AI super PAC Leading the Future pledged $5 million to Donalds’ campaign, saying in part, “Byron Donalds understands we’re on the cusp of a magical era of technological advancement that will unleash enormous economic benefits and protect Florida’s future.”

As with the Senate race in Ohio, some are now wondering whether data center backlash is putting the Florida governor’s race in play.

Democrats are rethinking their positions too. On Tuesday, Gov. Josh Shapiro signed what he described as the strictest data center guardrails in the country, and he is the same governor who spent years courting the buildout.

As I wrote previously in this series, the conventional wisdom in Washington last year was that data centers were a digital gold rush for rural America. This summer, we found out that assumption was a career-ending miscalculation for a number of politicians.

But it’s clear that the tech giants are not going down without a fight. And with the industry on the ropes, they’re now calling in the experts from big tobacco.

In 1993, the EPA classified secondhand smoke as a carcinogen, and Philip Morris hired a public affairs firm, APCO, to build the Advancement of Sound Science Coalition, an innocuous-sounding nonprofit whose job was to popularize the phrase “junk science.”

Today APCO sells data center advisory services. An investigation found that they paid for some of the ads run by Elevate New Mexico in support of Project Jupiter, the OpenAI and Oracle hyperscale campus in Doña Ana County, New Mexico.

APCO is not the only piece of tobacco infrastructure finding new life in this fight. The Consumer Choice Center, the group that built the tobacco industry’s third-party advocacy operation, has issued reports asking policymakers to “reduce regulatory friction” around grid upgrades for data centers. In other words, Consumer Choice Center wants to make it harder for local people to fight development and eminent domain in their own communities.

The “data center front group” strategy is everywhere now. Virginia Connects spent roughly $700,000 in one year on mailers and texts, and its tax filing lists the head of the Data Center Coalition, a group that actually calls itself “The Voice of the Data Center Industry,” as its principal officer. And nearly a dozen state-based front groups use their same website template and share a single Google Ads ID.

Meta put $34 million into American Edge, which now publishes briefs telling tax-starved counties that a data center is the biggest opportunity in a generation.

Right. And more doctors smoke Camels than any other cigarette.

Industries hire front groups to manufacture support and write bogus reports when they cannot find any organically, and AI products this year have struggled this year to demonstrate any real value to the American public.

The New York Times recently reported that Spotify, LinkedIn, YouTube, Meta, and Google are scrambling to remove low-quality AI-generated content from their platforms. Spotify removed 75 million spam tracks. LinkedIn built a button so users can flag posts as slop. Google researchers found 50,000 clusters of coordinated accounts and removed them. TikTok says it has labeled more than three billion AI-generated videos.

Amazon, Microsoft, Google, Meta, and OpenAI are building most of the hyperscale AI capacity in this country, and they’re now drowning in slop generated by their own data centers.

Last October, during an earnings call, Mark Zuckerberg told shareholders that social media had entered a third phase. The first was content from friends, family, and accounts you followed. The second phase was creator content. And the third phase was AI-generated content. “We’re going to add yet another huge corpus of content,” he said.

Well, he got what he asked for, and users hate it.

For the last five years, we were told AI would change everything. Sure, the first few years of the 2020s were a real step up in advanced computing, and there is demand for advanced computing. But for everyday users (who also happen to be the voters), we have yet to see any real benefit.

I bet I use AI more than the average American, and if you made me choose between Claude and Wikipedia, I’d pick Wikipedia every time, and I am not alone. Only a small fraction of ChatGPT users actually pay for the service. Turns out an app that just repackages Reddit posts and invents facts out of thin air isn’t really worth a monthly subscription.

The paying customers have yet to show up, but even the ones who have don’t really see the value. MIT examined what companies actually got out of their AI projects and found that 95 percent produced no measurable results. It’s an undeniable trend that’s starting to shake confidence in the sector.

The lack of demand means AI companies are spending far more to build the technology than they collect from users. That’s why they want to leave taxpayers holding the bill.

Bain estimates that by 2030, the industry will come up about $800 billion short of what it needs each year. To fill a hole that size, JPMorgan figures you would have to charge about $35 per month to every person on Earth who owns an iPhone to turn a profit. That’s not just the iPhone owners who want AI, it’s all of them.

Even Dario Amodei, who runs Anthropic, has conceded the point on X, writing that “the most accurate criticism of AI companies, including Anthropic, is that we haven’t yet delivered on our big promises to benefit the world.”

This is a big reason why Rural America has been so skeptical of the claim that data centers will rebuild local communities. We have been asked to make bad industry bets before, and you can still see them across our landscapes. The abandoned mines, empty distribution warehouses, and brownfields that nobody locally has the money to clean up are all a constant reminder of the snake oil and silver bullets that were sold to communities in decline.

My family is from the Prairie Pothole region. I grew up a mile from the Big Sioux River outside Brookings, South Dakota. My mom is from outside Emmetsburg, Iowa, and my dad is from near Fort Dodge. All of these places got sold on industrial-scale cellulosic ethanol when I was a kid, and none of the prosperity we were promised ever materialized.

In these private-public partnerships, the money flowed in one direction, and the risk flowed the other. Taxpayers paid for the job training and the infrastructure. Farmers signed the contracts and bought the equipment. When it falls apart, the farmer still owes the note, but the executives have already been paid. And eventually everyone ultimately somehow forgets about all the jobs that were promised.

And AI data centers might be the riskiest bet for Rural America yet.

Let’s assume the best-case scenario for a community that hosts a hyperscale data center: the industry solves its math problem and finds enough paying customers to cover the cost of the technology and power. A massive campus gets built in your county, stays full, runs for thirty years, and generates a couple million dollars in local property taxes. Sure. It’s plausible. But your power bill still goes up. Your neighbors still get the substation or the transmission line.

That is the best case. But it’s not as likely as the industry suggests.

The more likely scenario, in today’s economy, is that the bubble bursts partway through construction, the buildings sit empty, and the taxpayers who took all the risk are left holding the bag. And ultimately, if the space is filled, the new company that purchased the abandoned lot fails to follow through on the promises made by the now-bankrupt initial developer.

Maybe that is why the only people still telling us data centers will rebuild our rural economies are the same people who told us cigarettes would cure our cough?

Next up in our Data Center Summer series: Rural Americans hate data centers and top-down, billionaire-driven development. So what are the other options for building a strong local economy in a small town? What do voters actually want? And who at our taxpayer-funded institutions is already building the playbook for homegrown rural prosperity? We’ll cover that next.