The Shiller CAPE ratio — a measure of stock prices against long-term earnings — sits at roughly 41 for U.S. equities right now. On Black Tuesday in 1929, it was 32.5. That comparison alone should give any investor pause.
Bank of America dropped a research note in late June that reads less like Wall Street analysis and more like a weather warning. Their proprietary Bubble Risk Indicator scores the semiconductor sector at 0.91 out of 1.0 and the broader tech sector at 0.82 — levels the bank describes as dangerously close to “extreme bubble-like price action,” according to Reuters. BofA’s private bank research puts an even finer point on it: Big Tech has mobilized more than $300 billion in planned AI-related capital spending in 2025 alone via initiatives like the Stargate Project, and these companies have “yet to see a return on their investments.”
The Money Is Flowing. The Returns Aren’t.
Massive AI infrastructure spending is outpacing actual profits by a margin Wall Street may be systematically underpricing.
The pattern is familiar. Streaming platforms spent a decade burning cash to grab subscribers before anyone seriously asked about margins. AI infrastructure is running the same playbook at ten times the scale. UBS estimates that debt tied to AI build-outs is climbing roughly $100 billion per quarter. Morgan Stanley projects tech firms could lean on as much as $800 billion in off-balance-sheet financing for AI infrastructure by 2028. BofA calls AI a “double-edged sword” that could “cannibalize profits” rather than reliably boost them, according to Yahoo Finance.
The valuation dashboard looks like a check-engine light festival:
- The Buffett Indicator — which compares total U.S. market capitalization to GDP — sits at 218%, just below its all-time record of 219%, signaling the market is strongly overvalued relative to the underlying economy.
- The S&P 500 price-to-sales ratio stands at 3.22, nearly double the long-term average of 1.84.
- The Nasdaq 100 trades at roughly 37x earnings versus 22x for the broader S&P 500, showing how concentrated the most aggressive valuations are in tech and AI-heavy names.
“All measures of asymmetry and risk are flashing amber.” — Oliver Shale, Ruffer
Not everyone is sounding the alarm. Oppenheimer has set a 2026 S&P 500 target around 8,100. Goldman Sachs projects a 12% total return this year. But even BofA notes that hyperscalers — the large cloud providers most exposed to AI infrastructure — have already underperformed the broader market by nearly 15% since January. Meanwhile, open-source AI Chips and models are gaining traction, threatening the premium pricing power baked into current valuations and increasing margin pressure as the market matures.
“Doubts around the AI revolution are emerging.” — Bank of America research note
BofA’s Verdict: “Air Pocket,” Not Full Crash
BofA stops short of calling this a full bubble, but their suggested playbook still reflects serious caution about pure-play AI exposure.
BofA prefers the term “air pocket” — a sharp, temporary repricing rather than structural collapse. Their advice: shift toward AI-adjacent sectors anchored in supply-chain and geopolitical fundamentals rather than pure-play high-multiple names. Capital Economics calls recent volatility “evidence of excessive froth,” according to Futurism.
The dot-com era proved that transformative technology and financial reckoning are not mutually exclusive. The internet changed everything — and still erased trillions. Right now, the story surrounding AI sounds spectacular. The spreadsheet tells a different tale.
