5 Warning Signs Your AI Investments Are in a Bubble, According to Graham Stephan
Graham Stephan isn't predicting a crash. But in a recent video, he laid out the data on artificial intelligence (AI) and stock market valuations with enough honesty to make even optimistic investors pause.
He said the warning signs he identified aren't speculation. They're measurable, current and comparable to conditions that preceded some of the worst market collapses in modern history.
Check Out: Money Expert's Guide for Beginners on How To Invest $1,000 in 2026
For You: 8 Low-Effort Ways to Make Passive Income (You Can Start This Week)
Here's what he said to watch for.
AI Concentration Is Approaching Dot-Com Territory
The top 10 stocks now make up approximately 40% of the entire S&P 500, which is a level of concentration that actually exceeds what existed at the peak of the 2001 tech bubble. When that much market weight is carried by that few companies, the index becomes far more exposed to a single sector's fate than the diversification it implies.
Stephan noted that AI investment concentration specifically is tracking close to the same point as 2001 tech stocks and the Japanese everything bubble of the late 1980s. That comparison doesn't require the outcome to be the same, but it does mean the structural setup rhymes in ways worth understanding before dismissing.
Valuations Are Disconnected From Historical Norms
According to Stephan at the time of the filming of his video, the S&P 500's price-to-earnings (P/E) ratio exceeded 28, which is roughly two-thirds higher than the 100-year historical average of around 17. The Shiller Cyclically Adjusted Price-to-Earnings (CAPE) ratio, which smooths earnings over a ten-year period to reduce distortion from short-term swings, has reached its second highest level ever, surpassed only by the dot-com peak.
Stephan also shared the counterargument: these companies are genuinely profitable in ways their 2001 predecessors were not. According to him, the average P/E ratio at the dot-com peak was 65 versus 28 today, and Fidelity data confirms that current earnings are strong, cash flows exceed income and net profits are above their five-year averages. The valuations are stretched, but they're stretched on top of real earnings rather than pure speculation.
Companies Are Increasingly Invested in Each Other
Stephan flagged what he called the circular nature of the current market: major companies are each holding stakes in other major companies, creating a web of interdependence that looks stable until it isn't. When companies are this interconnected, a major decline in one can trigger cascading selling across the others in ways that don't show up in individual company analysis.
Learn More: Enter for a Chance To Win $500 in MoneyLion's Summer Break Giveaway (No pur. nec. Ends 7/4/26. See official rules at mlion.info/summerbreakofficialrules)
The Buffett Indicator Is at an Extreme
The Buffett indicator (which compares total stock market capitalization to GDP) sits 2.4 standard deviations above its historical trend. Warren Buffett has described this ratio as the single best measure of whether markets are overvalued at any given time.
Stephan cited Fidelity's specific warning signs for bubble conditions: companies burning through cash while spending more than they make, debt levels growing faster than profits and borrowing costs that could compress margins if they rise. Today's tech companies largely don't exhibit those patterns, but the market-level valuation signal is harder to dismiss.
The Japan Comparison Has a Very High Bar — But It's Not Zero
To see a repeat of Japan's 40-year economic stagnation, Stephan said earnings would have to stall completely and the S&P 500 would need to rise above 14,000 before a comparable collapse. That's not where we are. But the structural parallels he identified — cheap money flooding the system, land and equity values reinforcing each other upward, a sense that the new normal is permanent — are worth holding in mind.
Japan's market rose 22% annually for nearly two decades before falling 50% in a single year. Nobody called it a bubble while it was happening because the fundamentals kept supporting it — until they didn't.
What Would Need To Be True for AI Not To Be a Bubble
Stephan offered the other side explicitly. For current valuations to be justified rather than inflated, AI adoption would need to spread meaningfully to smaller companies across healthcare, professional services, industrial firms and utilities; boosting their productivity and earnings enough to justify broader market multiples. Fidelity argues this is already beginning to happen, and if it continues at scale, today's stretched valuations might look reasonable against future earnings.
The other condition: no unforeseen shock. Markets don't crash simply because they're expensive. They crash when something unexpected arrives that triggers a cascade of selling nobody was positioned for. Absent that, expensive markets can stay expensive (or get more expensive!) for longer than any rational analysis would predict.
Stephan's own position: stay invested, stay diversified, keep roughly 20% in treasuries as a buffer and dollar-cost average regardless of where the market is. Not because a crash isn't possible, but because the alternative — sitting in cash losing 1.8% annually in real purchasing power while waiting for clarity that never comes — has consistently produced worse outcomes than staying the course.
To help Americans navigate the added cost of summer, MoneyLion is giving away $1,000 every day through July 4. Enter the Summer Break Giveaway here (No pur. nec. Ends 7/4/26. See official rules at mlion.info/summerbreakofficialrules)
This article was provided by MoneyLion.com for informational purposes only and should not be construed as financial, legal or tax advice.
More From MoneyLion:
