The public’s data-center revolt spreads… even Texas taps the brakes… will it kill the AI trade?… why the move is “sort,” not “sell”
The public really hates data centers – but that could make you a lot of money.
Last month, I introduced an analytical framework I called “The Messy Middle” – the idea that AI won’t force us to choose between good and bad outcomes, but between two legitimate “goods” vying for priority. We won’t be able to have both in equal measure at the same time.
One example we profiled came from New York, where Gov. Kathy Hochul became the first governor to sign a law pausing the construction of new hyperscale AI data centers.
She gave up the “goods” of longer-term productivity gains, jobs, and the bigger tax base that those data centers would have delivered…in exchange for the “goods” of holding down electricity prices, protecting natural resources from potential pollution, and shielding communities from disruption.
Not right or wrong – simply a prioritization.
Now, this anti-data-center backlash isn’t limited to New York. It’s gathering momentum across the country at the very moment AI usage at home and at the office is accelerating, driving enormous demand for the very same data centers those communities are fighting.
That’s the Messy Middle setting up on a national scale. And while it could create real social tensions, it could also be incredibly lucrative for investors who know where to look.
Here’s our global macro investment expert, Eric Fry, to explain:
When new capacity becomes harder to build, existing capacity becomes more valuable.
And companies that already have the land, power and facilities needed for AI data centers could be sitting on valuable real estate – literally.
One company cashing in on the Messy Middle of AI’s infrastructure needs
Last week, Anthropic reportedly signed a 20-year, $9.1 billion deal with Riot Platforms, Inc. (RIOT) for 191 megawatts of data-center capacity at Riot’s Rockdale, Texas, campus.
Riot expects the deal to generate about $9.1 billion through 2048, with an option that could push the total value to $16.1 billion. The capacity is expected to be delivered in phases, beginning in 2026 and continuing through 2028.
Back to Eric:
The key here is that Riot doesn’t have to start from scratch.
The company, best known as a bitcoin miner, already has the land, power and infrastructure needed to support a data center.
So, instead of using all of that capacity for bitcoin mining, it can lease it to AI companies, like Anthropic.
As politicians and voters battle over the Messy Middle, companies like Riot that bypass these political and administrative bottlenecks stand to benefit.
Beyond RIOT, check out Core Scientific (CORZ). It holds massive, pre-existing gigawatt-scale power infrastructure and has been a pioneer in signing long-term, high-performance computing contracts with hyperscalers.
There’s also CleanSpark (CLSK), which commands a broad pipeline of powered industrial sites with approved grid connections. These sites can be retrofitted for AI data workloads or alternative compute.
The thread connecting all three stocks is the same: each one already owns something that’s suddenly become scarce – power, land, and grid access that would take years and a bruising permitting fight to assemble from scratch today.
But these infrastructure landlords are only the first rung on the AI opportunity ladder. There are plenty more rungs.
Think about everything it takes to turn a powered site into a working AI data center
You need the electricity itself, which points to power generators… You need to move that electricity, which points to transformers, substations, and transmission gear… You need to keep the chips from cooking themselves, which points to industrial-scale cooling… You need firms to design and build the facilities… And, of course, there’s everything that goes inside – the semiconductors and memory that do the actual computing.
Every one of those rungs is its own bottleneck. And every bottleneck is its own opportunity.
That’s a lot of ground for any single investor to cover, which is exactly why Eric, alongside Louis Navellier of Growth Investor and Luke Lango of Innovation Investor, have spent recent weeks covering it for you.
The three of them have rebuilt their collective AI Revolution Portfolio from the ground up – designing it to capture multiple rungs of that ladder rather than bet everything on a single stock or sector.
Our three experts first assembled this portfolio of elite AI stocks in 2023, then rebalanced it at the end of 2024 as the AI race shifted. And they see this moment – where a single infrastructure bottleneck is spilling opportunity across a dozen industries – as the next pivot point.
They’re pulling back the curtain tomorrow morning at 10:00 a.m. Eastern, when they’ll unveil their rebuilt portfolio and walk through exactly how they’re positioning for AI’s next phase.
Back to Eric:
The AI Revolution has created an enormous number of potential investment opportunities. But the more this technology spreads, the harder it becomes to know which companies deserve your attention – and, just as importantly, which don’t.
That’s what tomorrow is about – separating the best opportunities from the rest.
To join Eric, Louis, and Luke, just click here to register. We’ll see you tomorrow at 10:00 a.m. Eastern.
“But hold on, Jeff, if we see a wave of data center moratoriums across the nation, won’t that kill the AI trade?”
It’s tempting to see the backlash as pure upside: choke off new construction, and whoever already owns power gets richer. That’s the bull case, and I think it’s largely right. But a sharp reader should be asking the harder question – if the backlash keeps spreading, doesn’t it threaten the entire AI infrastructure trade?
Take an AI picks-and-shovels supplier like Vertiv Holdings Co. (VRT). Its order book has swelled past $12 billion on the power and cooling gear these facilities need. But those orders only convert to revenue if the data centers actually get built and powered. Slam the brakes, and a backlog is just a promise.
So, a widespread national moratorium on data centers isn’t something to brush off idly. Let’s walk through it.
Beyond memory, one of the most significant constraints on AI today is power. There simply isn’t enough of it, fast enough, to feed every planned facility.
When politicians layer permitting delays, grid audits, and ratepayer fights on top of an already power-starved buildout, they genuinely slow the pace at which that backlog turns to cash.
So yes, slower-than-expected revenue would ding AI infrastructure companies that have told Wall Street those revenues are coming. But a ding isn’t the same as a broad AI crash.
Here’s the part the doomsayers skip
Earlier this month, headlines trumpeted that even data-center leader Texas was slamming the brakes on its data-center rollout. It was positioned as a harbinger of doom for AI.
Yes and no.
Texas Gov. Greg Abbott ordered regulators to audit data centers waiting to plug into the state grid and deny those that didn’t measure up.
It was a screen, not a stop – and it came with a giant loophole: facilities that build their own on-site power can skip the grid queue entirely. Even his critics shrugged, with one Texas official calling it “all hat and no cattle” – Texas jargon for all talk and no action.
Sound familiar? On-site power is precisely what Riot, Core Scientific, and CleanSpark already have.
Meanwhile, all year, as the moratoriums piled up, the spending went up. The four biggest hyperscalers now plan roughly $725 billion in capital expenditures this year – the money they pour into building all this – up about 77% from 2025.
But how has this been happening even as the moratoriums have been growing? Because as we noted in our Messy Middle Digest last month, resistance in one state doesn’t kill AI demand – it relocates it.
A project blocked in New York or stalled in Texas moves to a friendlier corridor, another state, or offshore. The chips still get bought.
So, the backlash doesn’t break the trade, but it does redistribute the winners – and raises the cost of being on the wrong side.
Returning to our question then…
How worried should you be about your AI infrastructure stocks considering this growing data center backlash?
I’d call it a yellow flag, not a red one – but with one genuine red tail.
The yellow flag is timing. Delays and cost overruns can bruise these richly priced stocks even when demand remains intact.
Returning to Vertiv, we just watched it happen here in Q2. The company reported record demand and raised its full-year guidance, yet the stock still dropped about 12% as revenue slipped due to project timing and supply-chain snags. Nothing was wrong with the demand – only the pace of delivery – and the stock got slammed anyway.
The red tail is the one we flagged back in July: isolated state pushback just routes capital elsewhere, but a coordinated wall – a federal pause, or so many states acting at once that there’s nowhere friendly left to go – could truly break the trade. Today, that’s a risk to watch, but not the base case.
One thing to watch above all: the day that hyperscaler spending guidance stops rising. That’s the signal to take seriously. Until then, rising guidance is your green light to stay in the AI infrastructure trade.
Which is why the move today isn’t “sell,” it’s “sort”
This is the Messy Middle turned on your own portfolio. For every AI holding you own, ask a single critical question…
Which side of the productivity-versus-disruption tradeoff is it on?
The scarcity beneficiaries – the power, grid, cooling, and existing-capacity names – get more valuable as building gets harder. The chipmakers are largely insulated, because demand just relocates. The vulnerable ones are the opposite: names pinned to a single contested region, or priced as if the buildout will be cheap, fast, and unopposed.
All this brings us back to tomorrow’s AI Revolution Portfolio refresh, with updated positions spread deliberately across all the rungs of the AI ladder rather than concentrated on any one of them. Here’s that link again to reserve your spot.
We’ll keep tracking the backlash – and both sides of the trade it’s creating – in the months ahead.
Have a good evening,
Jeff Remsburg
(Disclosure: I own VRT.)
The post Don’t Sell Your AI Stocks – Sort Them appeared first on InvestorPlace.
