Every boom attracts doomsayers—and AI is no exception. A record number of global fund managers now say AI stocks are in a bubble, per Bank of America’s latest survey.
It’s an understandable concern. AI-related stocks have exploded higher this year. Companies are spending tens of billions building data centers and buying chips. And the headlines are full of big promises about how AI will change everything.
But that doesn’t necessarily mean we’re in bubble territory.
To figure that out, it’s worth understanding what a real bubble looks like—and how today’s AI landscape stacks up.
What a bubble actually is
Financial bubbles tend to follow a familiar pattern:
Investors become obsessed with a powerful new trend… Capital floods in… Excitement pushes valuations beyond what the fundamentals can support… And eventually, the market becomes oversupplied.
Think about the dot-com boom of the late 1990s, when hundreds of “internet companies” went public with no revenue… or the housing frenzy in the mid-2000s, when lenders handed out mortgages to anyone and everyone.
Once demand cooled, those markets collapsed.
But that isn’t happening in AI—at least not yet.
Right now, demand is still outpacing supply. Tech giants like Microsoft, Amazon, Meta, and Google are pouring record amounts into AI infrastructure.
For instance, Meta’s latest project with private equity partner Blue Owl Capital involves a massive new data center complex in Louisiana valued at $27 billion. And Taiwan Semiconductor (TSMC) continues to run at full capacity, making chips for Nvidia, AMD, and others.
In short, the supply side hasn’t caught up to the demand side—a key distinction between today’s AI boom and previous manias.
Both things can be true
Yes, some AI-labeled stocks are trading on vibes, not cash flows. There are pockets of froth (there always are in a bull market). Some companies will spend a fortune on AI experiments and end up with nothing to show for it.
But that doesn’t make AI itself a bubble. It means you must separate mini-bubbles from the megatrend.
Leaders like Zuckerberg and Altman have been blunt: overspending is a risk… but underinvesting is a bigger one.
It’s also important to remember that being right about a bubble doesn’t mean you’ll make money calling it too soon.
Consider that back in the 1990s, then-Federal Reserve Chair Alan Greenspan famously warned about “irrational exuberance” in stocks. He made that speech in 1996—four years before the tech bubble finally burst in 2000. In the meantime, the Nasdaq more than doubled.
Where to focus now
Instead of obsessing over whether AI is a bubble, investors should focus on who stands to benefit from the spending surge already underway. Here are some of the sectors primed for the most upside:
1) Grid and transmission upgrades
AI is exposing years of underinvestment in the power network. That means more high-voltage lines, substations, transformers, and interconnects. Winners include utilities that are allowed to earn regulated returns on new projects, companies that design and build long-distance lines, and manufacturers of the heavy gear that keeps the grid stable.
2) Power producers with expandable capacity
Data centers need firm, scalable electricity. Natural gas, nuclear (including SMR-adjacent services), and hybrid renewable and storage near load centers all stand to win big from AI’s thirst for power.
Watch for developers with land, permits, and interconnects already in hand—time-to-power is a major moat.
3) Solar and storage (utility-scale)
The solar math has changed. With storage, it’s increasingly the fastest path to incremental megawatts the AI ecosystem can actually use.
Look for strong backlogs, cash generation, and projects explicitly tied to data center loads.
4) Pick-and-shovel plays inside the data center
Not every dollar flows to GPUs. High-bandwidth networking, fiber links, advanced liquid cooling, dense rack power gear, and backup systems are essential—and tend to offer steadier margins and service revenue than the chip arms race. These vendors benefit as capacity scales, regardless of which model wins.
5) Software that prints ROI
The best AI software pays for itself fast—think months, not years—by cutting fraud, speeding customer support, or tightening supply chains. If a platform can show hard-dollar savings in 90–120 days, budgets can be secured even in tough environments. Look for clear case studies and usage that sticks after pilots.
While many chase the shiniest GPU narrative, the boring plumbing of the AI economy—power, lines, land, permits, cooling, interconnect—offers cleaner, longer runways with less headline risk.
What would change the thesis?
That doesn’t mean the risk of an AI bubble is zero. If we start seeing widespread budget cuts, falling chip utilization, or mass cancellations of data center projects, it would be a warning sign that demand is rolling over. But there’s no evidence of that yet.
Until spending slows meaningfully—or valuations detach completely from cash flow—it’s safe to assume we’re in an early-stage buildout, not the end of a speculative mania.
The bottom line
Artificial intelligence isn’t a bubble. It’s a massive secular trend still in its early innings.
There will be hype, winners, losers, and plenty of volatility along the way. But the core of this story—growing adoption, massive productivity gains, and trillion-dollar capital flows—is real.
The smarter play isn’t trying to guess when it ends. It’s positioning in the companies and sectors that make AI possible: the energy suppliers, the infrastructure builders, and the hardware innovators powering the next wave of global growth.
Editor’s note:
In last week’s issue of Curzio AI, Frank recommended a solar stock with groundbreaking tech that increases energy output by 25% vs. traditional solar panels.
Discover why this stock could easily triple from current levels. Join Curzio AI today and get 70% off!
