The AI Bubble Debate: How to Invest When Even Burry Is Nervous

The Nasdaq just posted its fifth straight losing session. The S&P 500 is down 2% on the week. Memory stocks are getting crushed — Western Digital down 13%, Seagate down 12%, Sandisk down 10% in a single day. South Korea's Kospi plummeted 10% in one session, dragged down by Samsung and SK Hynix. OpenAI is reportedly reconsidering its IPO timeline after SpaceX — barely two weeks public — sits more than 30% below its record high.

And Michael Burry — the man who saw the housing crash coming — has loaded up on bearish options against Nvidia and Palantir, urging investors to "reject greed" and "reduce positions almost entirely" in any stock going parabolic.

So... is this the top? Is the AI bubble finally popping? Or is this just another dip to buy?

What's Actually Happening

Let's separate the signal from the noise.

The numbers as of Friday's close (June 26, 2026):

The Nasdaq's weekly loss was its second-largest in the past year — eclipsed only by the 4.7% drop just three weeks earlier. This isn't a one-day blip. It's a meaningful shift out of the most crowded names in the market.

Meanwhile, the Dow actually gained ground for its third straight winning week. Translation: money isn't leaving the market — it's leaving tech.

The Bubble Case

Let's be honest about what the bears see, because they have a point.

Valuations are stretched. Palantir trades at a P/E of roughly 120-150 — down from over 250 in late 2025, but still at levels that demand extraordinary earnings growth to justify. Nvidia's P/E sits around 35-39, which sounds reasonable until you remember that it's priced for earnings growth that depends on an AI spending cycle continuing at full throttle.

Market concentration is at historic extremes. The top 10 stocks in the S&P 500 represent about 41% of the index's total weight — but generate only about 32% of its earnings. That gap between weighting and earnings is the kind of divergence you see when sentiment outruns fundamentals.

OpenAI's financials are, frankly, alarming. The company generated $13 billion in revenue in 2025 and reported over $20 billion in losses — though a significant portion stems from a one-time accounting adjustment tied to its restructuring from nonprofit to for-profit. Even adjusting for that, the burn rate is extraordinary. It's now valued above $850 billion on the private market, targeting up to $1 trillion in its IPO. That's a price-to-sales ratio north of 65 — for a company that isn't profitable. Even GMO, a respected asset manager, calls current AI valuations an "extreme bubble."

Burry's argument is simple. He's not saying AI won't change the world. He's saying that throughout history — 1999 dot-coms, 2006 housing, 2021 crypto — when valuations go parabolic and everyone is certain, that's precisely when the risk is highest. His prescription: "reduce exposure to stocks, to tech stocks in particular." And for anything going parabolic: "reduce positions almost entirely."

The "Not a Bubble" Case

But before you sell everything, hear the other side.

Earnings are actually growing. S&P 500 earnings per share are projected to grow 14.2% in 2026, driven largely by tech and financials. That's real profit growth, not just multiple expansion. Nvidia is generating enormous cash flow. Microsoft, Alphabet, and Meta are reporting genuine AI-driven revenue.

Companies are funding AI from cash, not debt. This is a critical distinction Fidelity highlights. During the dot-com bubble, companies burned through debt-financed cash with no revenue. Today's AI leaders are funding their capex almost entirely from earnings. That significantly lowers — though doesn't eliminate — the risk of systemic financial strain if the spending cycle slows.

Valuations are high, but not at dot-com extremes. The S&P 500's forward P/E is elevated relative to history, but still meaningfully below the peaks of 1999-2000. And today's elevated valuations are backed by companies with hundreds of billions in actual revenue — not Pets.com.

The technology is real. Unlike many dot-com companies that had no product, AI is demonstrably transforming productivity across industries. The debate isn't whether AI works — it's whether current stock prices already reflect decades of future growth.

So What Should Investors Do?

Here's the thing: you don't have to pick a side. The most resilient strategy right now isn't "all in" or "all out" — it's structuring your exposure so you profit if AI delivers, without getting wiped out if the bubble thesis is right.

1. The Pick-and-Shovel Approach

During the Gold Rush, the people who consistently made money weren't the miners — they were the ones selling picks and shovels. The same principle applies to AI.

Goldman Sachs projects that hyperscaler spending on AI and data centers will top $5 trillion by 2030. That money flows to companies that don't carry the sky-high valuations of the headline AI names:

• Infrastructure: Eaton Corporation (power management), Vertiv (cooling systems), GE Vernova (energy)

• Materials: Freeport-McMoRan (copper — essential for data center wiring), Caterpillar (construction equipment)

• Utilities: Power companies are becoming the unexpected AI plays — data centers consume enormous electricity, and utilities offer dividend yields while benefiting from surging demand

These companies participate in the AI buildout without trading at 120x earnings. If AI demand continues, they win. If the AI stocks crash, they're still essential businesses with reasonable valuations that provide downside protection.

2. Diversify Within AI — Don't Bet on One Name

If you want direct AI exposure, don't put it all on Nvidia. Spread across the ecosystem:

• Chips: Nvidia, AMD, Micron (memory is critical for AI inference)

• Cloud/Hyperscalers: Microsoft, Alphabet, Amazon — they have diversified revenue and actual AI products generating revenue

• Software: Companies embedding AI into existing profitable businesses

An AI-focused or broad market ETF gives you exposure without single-stock risk. And remember: owning an S&P 500 or Nasdaq-100 index fund already gives you significant AI exposure through the mega-caps.

3. Trim, Don't Liquidate

Burry says reduce exposure "almost entirely" to parabolic stocks. That's his style — he's a contrarian who makes concentrated bets. For most investors, a more practical approach:

• If any single AI stock exceeds 10-15% of your portfolio, trim back to a reasonable weight

• Take some profits off the table — you don't have to sell everything to de-risk

• Rebalance into the pick-and-shovel names or broader index funds

• Keep enough AI exposure that you participate in upside, but not so much that a 40% correction would be catastrophic

4. Watch the Signals

Fidelity identifies five factors to monitor that will tell you whether this is a healthy pullback or the start of something worse:

1. Earnings growth — Are companies delivering the AI-driven revenue they promised?

2. Earnings quality — Is growth coming from real operations or financial engineering?

3. Valuations — Are multiples expanding or contracting relative to earnings?

4. Capex sustainability — Can companies maintain AI spending levels, or are they overextending?

5. Interest rates — Higher rates compress valuations. Watch the Fed.

If earnings continue to grow at double digits and rates stay stable, this correction likely resolves as a healthy reset. If earnings disappoint and rates rise simultaneously, the bubble case gets much stronger.

The Bottom Line

Here's what I think is true: AI is a genuine technological revolution that will create enormous value over the next decade. It's also true that some of the most popular AI stocks are priced for perfection, and any disappointment will be punished severely.

The mistake isn't being bullish on AI. The mistake is confusing "AI will be transformative" with "any AI stock will go up forever." The dot-com bubble proved that you can be right about the internet and still lose money betting on the wrong companies at the wrong prices.

The investors who win the AI era won't necessarily be the ones who called the top or caught the bottom. They'll be the ones who managed risk, diversified their exposure, avoided paying 120x earnings for anything, and positioned themselves to benefit from the real, multi-year infrastructure buildout that's happening regardless of what happens to any single stock.

Burry might be right about the froth. That doesn't mean you should run from AI. It means you should be smart about how you own it.

Market data as of close, Friday June 26, 2026. This post is for informational purposes only and does not constitute investment advice. Always do your own research and consider consulting a financial advisor.