Is the AI boom actually a bubble? Here’s what analysts are saying

Silicon Valley is pouring concrete and capital like it’s building a new interstate. Data-center blueprints are swallowing whole ZIP codes, utilities are rewriting load forecasts, and megacaps are spraying cash at chips the way old industries sprayed it at oil rigs. The scale feels historic — and that’s exactly what makes investors twitchy.

Today’s market feels both inevitable and unnervingly fragile. Every quarter brings a bigger data-center pledge, a nastier power bill, and a fresh set of claims about a productivity windfall just around the bend. And yet, in boardrooms and research calls, a conversation is happening: Is this the supercycle that re-rates capitalism — or is it a bubble assembling itself in real time?

One camp says today’s run is expensive but earned — grounded in earnings power, cash-funded capex, and use cases that will eventually catch revenue up to rhetoric. Hyperscalers are spending at a clip that would have sounded deranged three years ago; chipmakers are booked out years; entire utilities are being redesigned to feed “AI factories.” The other camp hears dot-com echoes: valuations are stretched, corporate adoption is lumpy, gains are clustered in a few AI-linked megacaps, returns are uneven, and the financing web powering AI has started to look rather circular as suppliers bankroll customers who then book orders back with the suppliers.

Read more: Is the AI boom actually a bubble? Here’s everything you need to know

The IMF says risk assets screen “well above fundamentals,” a posture that raises the odds of a sharp correction if something — growth, rates, politics — snaps. The Bank of England is worried about AI-centric tech in particular, calling valuations “stretched” and the market exposed if expectations cool. Under the hood, fresh microdata adoption complicates the victory lap. Ramp’s corporate-spend index shows the share of U.S. businesses paying for AI dipped to 43.8% in September — not a cliff, but a wobble that makes the timing of ROI matter more than the poetry of roadmaps. If the buildout keeps outrunning the billables, the narrative won’t hold its multiple.

And yet, positioning is still audacious. Bank of America’s latest global survey has cash at 3.8% — the most risk-on since February — while the same respondents call an “AI bubble” the market’s top tail risk. That contradiction is what this moment feels like: funding the boom, gaming the bust.

That split is visible in the analyst class. Some of the Street’s biggest names insist this isn’t 1999 — stretched, sure, but anchored to real earnings and cash. Others warn that the feedback loop between suppliers, customers, and financiers is turning capex into revenue on paper, and that a reckoning will be brutal if demand blinks. Here’s where the most influential voices have landed — and what, exactly, they’re saying.

Goldman Sachs (Peter Oppenheimer)

Goldman’s chief global equity strategist argues this isn’t 1999 on the numbers: Valuations are “becoming stretched, but not yet at levels consistent with historical bubbles.” The bigger risk, in his view, is concentration — a market that rises and falls on a handful of companies, which may just be another way of saying to mind the leadership and the plumbing, not just the headline P/E.

Bank of America (Michael Hartnett)

Positioning screams “party,” while the survey whispers “bubble.” And Hartnett’s October survey of global fund managers is the mood ring: Cash levels fell to 3.8%, the most bullish posture of the year, even as the same investors named an “AI bubble” their top tail risk. “Price action, valuation, concentration, speculation — all appear frothy,” Hartnett wrote, adding that “every bubble in history popped by central bank tightening.” With major central banks now on hold or easing, he argues, the rally could keep running — but that might only make the eventual correction sharper when — or if — it comes.

Read more: Tracking OpenAI's flurry of deals with Nvidia, Walmart, AMD, and more

Morgan Stanley Wealth Management (Lisa Shalett)

Right now, a handful of companies are doing nearly all of AI’s heavy lifting, and Shalett thinks that should make investors uneasy. Since the launch of ChatGPT, she noted recently, roughly 75% of the S&P’s gains and 80% of its profit growth have come from a tight circle of AI-linked giants. That narrow leadership, she says, risks creating a “Cisco moment” — a repeat of 2000, when investors discovered how much of the market was priced into a few names. She calls this the “seventh inning” of an AI-capex-powered bull market, and her read is that breadth that slim can levitate the index — until the revenue cadence slips, free-cash-flow growth cools, and the market remembers gravity.

UBS Global Wealth Management (Mark Haefele)

UBS’ calculations assume scale can still justify steep expenditure. The firm now expects global AI capital spending to hit $375 billion in 2025 (a 67% increase over 2024) and to rise a further 33% to $500 billion in 2026. The firm also forecasts that global AI revenues will grow at a 41% compound annual rate through 2030, reaching about $2.6 trillion over that period. UBS argues that, relative to the dot-com era, today’s technology companies have stronger cash flows, better margins, and more sustainable business models — giving valuations more room to maneuver. Still, the firm cautions that excitement has overshot in some names and that investors should balance exposure across software, semiconductors, and infrastructure segments in the AI ecosystem.

Read more: Is the AI boom actually a bubble? Here’s what the biggest names in business are saying

Barclays Research (Ajay Rajadhyaksha)

Barclays’ research chief is in the “frothy, not bubble” camp. He thinks the bears will keep arguing that use cases lag spend, but his team’s outlook leans constructive on growth and earnings with AI as a legitimate driver — so long as the market doesn’t run into power, data, or financing bottlenecks that upset the capex math. Their models “stress-test” capex assumptions to probe downside, yet they conclude that AI investment remains on "solid fundamental footing," supported by cash generation from hyperscalers. They warn that bottlenecks in power, memory, and infrastructure could stress the cycle, but argue that today’s funding is less dependent on leverage and more anchored by real projects.

Wedbush (Dan Ives)

Ives views the current tech market as still in its infancy. In a recent interview, he called this a “golden age” of technology, predicting a strong second half as the AI investment cycle unlocks productivity. He sees recent pullbacks as opportunities rather than warnings — a chance to accumulate names he believes will benefit as infrastructure spending transitions to enterprise adoption. Ives has written that this is a “1996 Moment... and NOT a 1999 Moment.” He treats pullbacks as opportunities to add core AI winners, not as warnings of a top.

Citi Research

Citi is sounding the alarm on how capex is evolving. The firm’s updated forecast for next year’s $490 billion hyperscaler spending is a stretch upward, and crucially, analysts note that an increasing share of that buildout may be financed via debt rather than cash flow. That shift introduces another risk layer: rising interest costs, refinancing danger, and weaker operating margins on leveraged infrastructure. Citi still expects AI demand to be robust but cautions that financial structure is becoming an essential line of sight.

Evercore ISI (Julian Emanuel)

Emanuel sees an AI bull run with a built-in warning light. His team raised the odds that the S&P 500 could reach 9,000 by 2026 under a “big, beautiful bubble” scenario, while anchoring their base case around 7,750, allowing room for extra upside before downside. Their internal client survey backs the tension: 67% of Evercore’s clients believe a bubble is already inflating. Emanuel frames volatility as opportunity — “don't fear a bull market drawdown” is his line — but he also notes that once momentum stops, the return path can be vicious. “Near-term volatility ahead is a buying opportunity, not the start of a bear market.”

Morgan Stanley, Accounting and Valuation (Todd Castagno)

The increasingly complex relationships between AI players worry Castagno — suppliers funding customers, revenue sharing, and cross-ownership make it harder to see what’s real and what’s structural artifice. He has raised red flags around disclosures and financial structure more than demand curves; his concern is that vendor financing, related-party deals, and revenue-sharing contracts can distort the apparent performance of AI customers and suppliers. Without that, he warns, booked contracts and backlogs could mislead rather than inform. Until the Street sees more clarity on customer concentration, margin assumptions, and cross-ownership, he says it’s premature to award full credit to booked contracts and backlogs.

Bernstein (Stacy Rasgon)

When Nvidia revealed it would invest in OpenAI while supplying its chips, Rasgon didn’t dismiss the spectacle. Instead, he warned that the structure “will clearly fuel ‘circular’ concerns.” That dual role of supplier and investor, he argues, blurs the line between demand and engineering. He accepts that compute scarcity is real (hence the booked orders), but insists investors look past press releases to how capacity is used, how margins hold, and whether bookings translate into consistent use.