Capital & Power Short: CFRA’s Angelo Zino Says the AI Bubble Narrative Is Changing—But Not for the Reasons You Think

The AI trade has moved from euphoria to anxiety faster than any other market theme in recent memory. Stocks tied to artificial intelligence have whipsawed violently, Nvidia’s historic earnings pop evaporated in hours, and traders are suddenly questioning whether the AI buildout has outrun its fundamentals. According to Angelo Zino, Senior Vice President and Technology Equity Analyst at CFRA, the conversation has undeniably shifted. But the core question—“Is this really a bubble?”—is far more nuanced than the day-to-day market action suggests.

“The narrative as far as the AI story is concerned definitely appears to be changing at this time,” Zino says, identifying a structural evolution in who is driving the spending. For the last two years, hyperscalers—companies with “essentially unlimited amount of capital to spend”—have powered nearly all incremental demand. Now, he argues, we’re entering a new phase “going into 2026 and 27, where it’s not just the hyperscalers, it’s kind of broadening out to more and more players.”

That broadening, he warns, introduces new fragility. “Some of those players don’t necessarily have the financial power that these hyperscalers do,” he says, adding that the market is increasingly reliant on second-tier operators that may lack stable financing, diversified customers, or the balance sheet heft to keep pace with Nvidia’s product cadence. In other words, the risk isn’t the AI story ending—it’s that the cast of characters now includes actors who can stumble.

So, is it a bubble? Not necessarily.

Zino draws a clear line between valuation excess and execution risk. “When we think about an AI bubble, there’s really two things that can cause a bubble or pop a bubble,” he explains. “One is valuation. And we actually don’t think there’s a valuation bubble for the biggest names out there in the market.” He notes that when you analyze the seven largest tech companies—substituting Broadcom for Tesla—“you’re talking about 22, 23 times on average on our 2027 estimates.” Far from the absurd multiples of past tech manias.

Instead, the real question is whether growth expectations are grounded. “Where the concern is, is whether or not the growth expectations are reasonable, right? Will this spending continue here going into 2026 and 2027?”

So far, Zino still sees strength. Nvidia’s backlog and pipeline, highlighted by Jensen Huang, provide “visibility over the next 12 months,” reinforcing that “at least here over the next year or two, I think things look pretty good.” Hyperscaler commentary from Microsoft and Alphabet has also been supportive.

But beyond 2025, the picture gets murkier. “As you go into 2026 specifically, there’s more potential hurdles and bumps in the road… whether it be tied to financing, whether it be tied to energy bottlenecks, whether it be tied to the macro picture.” That stack of uncertainties is what Zino believes has begun to “unnerve the market.”

Part of the volatility stems from investor expectations that Nvidia’s blowout earnings alone should magically fix sentiment. That’s never been the right framework, he argues. “The bear case out there doesn’t necessarily change based on great Nvidia results,” he notes. “We knew Nvidia’s numbers were gonna be good…the question wasn’t if they were going to be, it was going to be to what magnitude.”

To change the narrative, he says, investors should shift their focus to the next tier of AI infrastructure providers—companies like Oracle and CoreWeave, which sit “in the eye of the storm.” Oracle, with its concentrated OpenAI exposure and recent need to secure additional financing, is a prime test case. “We do want to see what they have to say in terms of diversification of that customer base… their ability to achieve financing.” If Oracle can demonstrate enough liquidity—“coupled with the cash flow from operations”—to avoid tapping markets “over the next two to three years,” the entire AI complex could stabilize.

The other unresolved issue is the payoff. “The payoff is definitely an important question out there,” Zino says, noting that depreciation cycles, GPU monetization, and customer economics all factor into sustainability. Much of the current demand has come from consumers. “Consumers are using it a lot more,” he explains. “You can look at any metric… there is definitely significantly more tokens generated from these AI workloads.” But this enthusiasm has not yet translated into broad enterprise revenue. The enterprise curve “has been fairly slow in nature,” even though historically consumers tend to adopt first and businesses follow later.

That second wave—enterprise adoption—is what ultimately determines whether AI is simply a capital-intensive boom or a durable economic transformation.

For now, Zino does not subscribe to the idea that the AI bubble has already burst. Instead, he sees a narrative shift driven by legitimate questions about who funds the next trillion dollars of infrastructure, how diversified demand is, and when enterprise adoption meaningfully accelerates. The market may be skittish, but the core of the AI story—hyperscaler demand, Nvidia’s dominance, and consumer engagement—remains firmly intact.