It's the question hanging over every trading screen in 2026: is the artificial intelligence boom a once-in-a-generation wealth-creation event, or the largest stock market bubble since the dot-com mania of 1999? The stakes are enormous. AI-linked names like Nvidia, Palantir, and Broadcom have driven the lion's share of the S&P 500's gains, and the index has clawed back roughly 16% year-to-date after April's tariff-driven crash. JPMorgan now sees the S&P reaching 7,500 by year-end — and possibly 8,000 if the Fed starts cutting rates. But the market is also sitting at one of the most expensive valuations in history.

This is not a "the world is ending" article, and it's not a cheerleading piece either. It's a clear-eyed breakdown of the bull case, the bear case, the specific warning signs that matter, and — most importantly — how to position as a trader when nobody can reliably call the top. (Nothing here is financial advice; it's market education.)

First, the Bull Case: Why the Boom Might Have Years Left

The strongest argument for the AI rally is simple: the spending is real, it's contracted, and it's enormous. The biggest technology companies on earth — Amazon, Microsoft, Alphabet, Meta, and Oracle — are collectively committing close to $405 billion in AI infrastructure investment, and hyperscaler capital expenditure guidance for 2026 has been revised up toward $660 billion. That is not speculative retail money chasing a meme; it's the balance sheets of the most profitable companies in history.

Surveys back up the conviction. According to one widely cited 2026 outlook, roughly 60% of AI investors believe AI-focused companies will deliver strong long-term results, and 9 in 10 plan to hold or buy more. Bulls argue that even if individual stocks get ahead of themselves, the underlying demand for compute, data centers, and AI software is a multi-year — possibly multi-decade — secular trend. In this view, dips are buying opportunities, not exit signals.

The Bear Case: Michael Burry Bets Over $1 Billion Against AI

Then there's the other side — and it has a famous face. Michael Burry, the investor made famous by The Big Short for predicting the 2008 housing collapse, has reportedly built put-option positions worth more than $1 billion against Nvidia and Palantir through his fund, Scion Asset Management. The largest single position is in Palantir puts, with a nominal value reported around $912 million, and he has doubled down rather than backing off.

Crucially, Burry's warning isn't "AI doesn't work." His thesis is that we're in a specific phase of the cycle — the overexuberance phase — that rhymes with the late-1990s internet bubble. On Palantir specifically, he has argued the stock is worth somewhere from "single digits to low double digits," and that he's "shorting the inherent unsustainability of its business model," not merely a high multiple. When the person who shorted the housing bubble starts shorting your favorite AI stock, traders should at least understand why.

The Circular Financing Problem

This is the part that genuinely worries even some bulls. A growing share of AI "demand" is increasingly circular — the same dollars moving between a small group of interdependent companies. Consider the loop: Oracle backs the massive Stargate project and a roughly $300 billion data-center buildout for OpenAI; Nvidia is the primary chip supplier into that buildout; Nvidia has also committed to invest as much as $100 billion in OpenAI; and Nvidia holds a stake in CoreWeave, which in turn has a cloud deal with Oracle.

Capital keeps recycling inside a closed circle, and every loop inflates headline "growth" numbers. The respected asset manager GMO has described this arrangement as "reminiscent of the circular financing of the internet bubble." For traders, the lesson is that some reported demand may be less independent — and less durable — than it looks. If one link in the chain blinks, the others feel it fast.

The Capex-vs-Revenue Gap

Here is the single number that best captures the bear case. OpenAI's 2025 revenue was roughly $13 billion, while its capital-expenditure commitments over the next eight years have been reported around $1.4 trillion. That is a staggering gap between money coming in and money committed to going out. The entire AI infrastructure thesis rests on the assumption that end-user demand and monetization will eventually catch up to the spending.

If that monetization arrives, today's capex looks visionary. If it's slow, or if enterprises pull back on AI budgets during an economic slowdown, then hundreds of billions in data-center commitments — including nearly $1 trillion in undisclosed future lease obligations across the big hyperscalers — start to look like a serious overhang. The bubble doesn't necessarily pop on bad AI; it can deflate on a demand air pocket that makes the spending look premature.

How Stretched Are the Valuations?

Valuation is where the dot-com comparison gets concrete:

• Nvidia: A trailing P/E around 43 and a forward P/E near 24 — not insane for a hyper-grower — but a price-to-sales ratio around 23, which is in the neighborhood of peak dot-com-era multiples. The forward P/E only stays reasonable if earnings keep compounding at their recent pace.
• Palantir: Priced at more than 50 times earnings expected two years out. For the multiple to make sense, near-flawless execution has to continue for years.
• Broadcom: Less extreme, but it carries its own twist — its custom AI accelerators could actually eat into Nvidia's long-term dominance, a reminder that even within the AI trade, today's leader isn't guaranteed to be tomorrow's.

The market overall is entering this period as one of the priciest on record. High valuations don't cause crashes, but they reduce the margin for error: when a richly priced stock merely meets expectations instead of crushing them, it can still fall hard.

Warning Signs Traders Should Actually Watch

Forget trying to predict the exact top. Instead, monitor the signals that historically precede a regime change:

• Capex guidance cuts. The moment a major hyperscaler trims AI spending guidance, the "demand is infinite" narrative cracks. Watch earnings calls closely.
• Deteriorating breadth. If the index keeps rising but fewer and fewer stocks participate — leadership narrowing to a handful of AI names — that's classic late-cycle behavior.
• Credit and financing stress. Because so much of the buildout is debt- and lease-financed, rising borrowing costs or a financing hiccup at any node in the circular chain matters.
• The Fed. With rates held at 3.50%–3.75% amid sticky inflation from the energy and Iran-conflict shock, "higher for longer" pressures the exact high-multiple growth stocks that lead the AI trade. A surprise hike would hit them hardest.
• Insider selling and momentum breaks. A close below key moving averages (the 50-day, then the 200-day) on heavy volume tells you sentiment has shifted, regardless of the story.

How to Trade It: Position Without Predicting

The trap in any bubble debate is feeling forced to pick a side and bet the account on it. Professionals do the opposite — they build a plan that survives either outcome:

• Size for the volatility, not the conviction. If you're trading AI names, risk 1% or less of your account per trade. High-multiple stocks move violently in both directions; position sizing is what keeps you in the game.
• Respect your stop-losses. The fastest way to turn a bubble debate into a blown account is holding a loser because you "know" it'll come back. Define your exit before you enter.
• Don't short on valuation alone. Burry can hold a billion-dollar losing position for months; you probably can't. "Expensive" is not a timing signal — bubbles stay irrational longer than most accounts stay solvent. If you must be bearish, define risk with options rather than naked shorts.
• Diversify your exposure. If your portfolio's fate is 100% tied to three AI stocks, you don't have a portfolio — you have a concentrated bet. The 2026 rally has rewarded that bet, but concentration cuts both ways.
• Keep some dry powder. If a real correction comes, cash becomes the most valuable position. The traders who profit from a burst aren't the ones who called it — they're the ones who had capital ready to deploy.