What to know
- The Bank of England says AI stock valuations now rival the dot-com bubble's peak.
- Cisco took 25 years to recover its bubble-era stock price — even great companies can get crushed by overpaying.
- If AI deflates even partially, boring value stocks could be the biggest winners for years.
You probably don't remember Pets.com. It was an online pet supply store with a sock puppet mascot, an $82 million IPO, and a Super Bowl ad. Nine months after going public, it was dead.
Pets.com became the punchline of the dot-com bubble — the symbol of what happens when investors throw money at a story instead of a business. For years, it was a history lesson. Something that happened to other people, in another era.
Except now, serious institutions are dusting off that history lesson and pointing it directly at the AI boom. And the numbers backing up their argument are uncomfortably specific.
This isn't a prediction that AI stocks will crash tomorrow. It's a map of what happened last time — and where the dominoes could fall if history rhymes.
The Bank of England warned that AI stock valuations are 'comparable to the peak of the dotcom bubble.'
What just happened
The dot-com bubble peaked on March 10, 2000. Between 1995 and that peak, the Nasdaq Composite rose 600% — then fell 78% by October 2002. That boom was fueled by the same basic dynamic we see today: a genuinely transformative technology attracted a flood of venture capital and sky-high valuations for startups that had no real business model.
Fast forward to now. The Bank of England has warned that AI stock valuations are 'comparable to the peak of the dotcom bubble'. The concentration of AI stocks inside the S&P 500 has hit dot-com bubble peak levels. And the overall stock market is near its peak dot-com era valuation.
None of this means a crash is imminent. The dot-com bubble itself ran for years before popping. But the structural similarities are now too specific for serious investors to wave away.
First domino: AI's circular money machine
The worry isn't that AI is fake. It's that leading AI firms may be involved in a circular flow of investments that artificially inflates their stock prices. Company A buys cloud credits from Company B, which buys AI chips from Company C, which invests in Company A's platform. Each transaction shows up as revenue on someone's income statement.
This is the same dynamic that inflated dot-com valuations in the late 1990s: startups buying ads from each other, hosting services from each other, and counting it all as growth. The technology was real. The revenue was... Creative accounting.
Some analysts push back hard on this comparison. They argue that today's AI leaders have real earnings, unlike most dot-com companies. One analysis argues the AI boom isn't near a peak and is only at 'base camp'. Both sides have evidence. But the circular-investment concern is the specific mechanism that could turn a correction into a rout.
Second domino: The index concentration trap
The concentration of AI stocks in the S&P 500 has hit dot-com bubble peak levels. That's not a metaphor. It's a measurement. A small group of companies now represents an outsized share of the index's total value.
When a narrow group dominates an index, a downturn in that group drags the entire market down disproportionately. Millions of Americans own the S&P 500 through their 401(k)s and index funds. They think they're diversified. They're not — they're making a concentrated bet on AI whether they realize it or not.
This is exactly how the dot-com bust spread beyond tech. In 2000, it wasn't just Pets.com investors who got hurt. It was everyone holding the Nasdaq. The same structural risk exists today, just with different company names on the marquee.
Third domino: Semiconductors are the new Cisco
Semiconductor stocks haven't been this hot since the dot-com bubble. They're the picks-and-shovels play of the AI gold rush — the companies selling the hardware that makes AI possible. The business case is real. The question is whether the stock prices already reflect a decade of future growth.
Cisco's story is the cautionary tale. It survived the dot-com bust. It kept growing revenue. It remained a dominant company. And yet it took a quarter century — until late 2025 — for Cisco's stock to surpass its dot-com-era closing high.
Twenty-five years. An investor who bought Cisco at the peak in 2000 waited until 2025 to break even. The company was fine. The stock price was not. Even fundamentally sound companies can take decades to recover bubble-era valuations if investors overpay at the peak.
Cisco survived the bust, kept growing revenue, and remained dominant. It still took 25 years for the stock to recover.
Fourth domino: The venture capital freeze
The dot-com era produced a flood of venture capital and rapid growth in startup valuations. When public markets crashed, that funding evaporated. Thousands of startups that were months away from profitability ran out of cash — not because their technology was bad, but because the funding window slammed shut.
The same dynamic applies today. AI financing has drawn direct comparisons to the dot-com era. When public market sentiment turns negative on a sector, private market funding for that sector typically contracts with a lag of several months.
This is the domino most people miss. A correction in public AI stocks wouldn't just hurt shareholders. It would starve the next generation of AI startups of capital — potentially slowing the very innovation that justified the boom in the first place.
Fifth domino: Boring stocks become the winners
After major growth-stock bubbles burst, capital tends to rotate into value-oriented sectors with stable cash flows. Money doesn't vanish — it moves. And it tends to move toward companies with real earnings, reasonable valuations, and boring but predictable businesses.
That's exactly what happened after 2000. Value stocks massively outperformed growth stocks for nearly a decade. Utilities, consumer staples, healthcare — the sectors that nobody talked about at cocktail parties — delivered the best returns.
A partial deflation of AI valuations could trigger a similar multi-year rotation. You don't need a full-blown crash for this to play out. Even a modest repricing of AI stocks could redirect billions toward the parts of the market that have been ignored during the mania. The Pets.com lesson isn't just 'avoid bubbles.' It's 'pay attention to what everyone else is ignoring.'
The last time this happened
The Nasdaq rose 600% between 1995 and March 2000. The bubble was powered by a technology that genuinely changed the world — the internet. The problem wasn't the technology. It was the valuations.
At the peak, companies with no revenue were worth billions. Companies with real revenue were worth absurd multiples of it. And the index was so concentrated in tech that when the music stopped, the entire market fell.
The Nasdaq dropped 78% from its peak by October 2002. Trillions of dollars in wealth evaporated. But Amazon survived. Google launched. The internet kept growing. The technology was transformative even though most companies built on it failed.
That's the uncomfortable truth about bubbles. The believers are usually right about the technology and wrong about the price. The skeptics are usually right about the price and wrong about the technology. Both sides lose money.
What could go wrong
This time really is different. One argument against the AI bubble thesis is that, unlike many dot-com companies, leading AI firms have real revenue and earnings. If AI companies keep growing into their valuations — if revenue catches up to stock prices — then the bubble comparison falls apart. The dot-com analogy requires that valuations stay disconnected from fundamentals. If they reconnect, the bears are wrong.
The bubble runs longer than anyone expects. One analysis argues the AI boom isn't near a peak and is only at 'base camp'. The dot-com bubble itself lasted roughly five years from early mania to peak. If we're early in the cycle, selling now means missing potentially enormous gains. Timing a bubble is nearly impossible — ask anyone who shorted Nasdaq in 1998.
Value stocks stay cheap forever. The rotation into boring sectors requires capital to actually move. If AI keeps delivering, or if a new narrative captures investor attention, value stocks could stay unloved for years. Being early and being wrong feel identical.
Concentration doesn't cause a crash. High index concentration is a risk factor, not a trigger. The S&P 500 could remain top-heavy for years without a correction. Concentration makes the market fragile, but fragility isn't the same as failure.
Watchlist
| Ticker | Level | Status | Why |
|---|---|---|---|
| QQQ | Watch for sustained weakness | monitoring | The Nasdaq 100 ETF is the direct proxy for AI concentration risk. If the biggest AI names crack, this is where it shows up first. |
| SMH | Watch for divergence from QQQ | monitoring | The semiconductor ETF. Chip stocks are the Cisco of this cycle — if they start underperforming the broader tech index, it's an early warning signal. |
| VTV | Watch for relative outperformance | monitoring | Vanguard's value ETF. If capital starts rotating out of growth and into value, this is the clearest signal. Outperformance here means the fifth domino is falling. |
| XLU | Watch for inflows | monitoring | Utilities were a top-performing sector after the dot-com bust. Boring, stable cash flows become attractive when growth stories unravel. |
| XLP | Watch for relative strength | monitoring | Consumer staples — toothpaste, cereal, soap. Nobody gets excited about them during a boom. Everybody wants them during a bust. |
Get the weekly digest
One email every Saturday. New stories, new research, no upsell. Unsubscribe with one click.
