Wall Street has a habit of partying hard right before the crash. Recent weeks felt like a 1990s déjà vu with indexes at record highs, valuations stretched beyond reason, and a handful of mega caps holding everything together. Investors talked about a "new era," but history's warning lights are flashing red.

From Robert Shiller's century-old valuation metric to Warren Buffett's favorite market-to-GDP gauge, the verdict is the same. Grayhair's indicators are showing that this is one of the most expensive markets in history. AI hype brought a fresh punch bowl at the party, but the boom depends on an energy grid that's already buckling.

Third Priciest On Record

The Shiller cyclically-adjusted price-to-earnings ratio (CAPE), created by Nobel laureate Robert Shiller, takes the past decade of inflation-adjusted earnings and smooths out recessions and short-lived shocks. Historically, it averages about 17. Today? Nearly 39 — among the highest readings in 150 years of data.

As of Wednesday, the Nasdaq traded at about 105% of the U.S. economy's entire output — a first in history. Meanwhile, the Wilshire 5000, the broadest U.S. equity index, was valued at a record 212% of GDP. Buffett has long warned that anything above 140% is "playing with fire."

No wonder Berkshire Hathaway ( BRK/A ) is sitting on its largest-ever cash pile, with a rare purchase being UnitedHealth ( UNH ) , and only after the stock tanked over 60%.

History shows that, when valuations stretch this far, the following years are lean. After prior CAPE readings above 37, the S&P 500's next one-, two-, and three-year returns were all negative. The market doesn't forgive excess forever.

Crowded At The Top

Over the last five years, the Magnificent Seven stocks have racked up a 335% average gain, versus 92% for the S&P 500 overall. This exclusive group now makes up one third of the index, with Nvidia ( NVDA ) alone able to drag the market up or down – seemingly on a whim.

The internet hype in the late 1990s worked the same way — a few mega-cap names dragging the entire market higher while the rest of the index limped along. Market breadth, the share of stocks actually contributing to gains, is now historically thin, showing the increased market fragility.

Even Sam Altman, CEO of OpenAI, admitted the obvious in an interview for The Verge.

"Are we in a phase where investors as a whole are overexcited about AI? … My opinion is yes."

Seize The Means Of (Electricity) Production

The AI boom may be real, but its foundation rests on something boring and finite: electricity. Department of Energy found that data centers already consume 4.4% of U.S. power, and by 2028, that could rise to as much as 12%. That's up to 580 terawatt hours, the equivalent of more than five dozen nuclear plants' worth of juice.

Virginia is the epicenter — its data centers already swallow nearly a quarter of all state electricity, in high-growth scenarios, that could jump to 50%. Power grids are straining, utilities can't keep up, and 56% still comes from fossil fuels.

Altman himself notes that OpenAI will spend "trillions of dollars on data center construction in the not very distant future."

Thus, the dream growth that the market is optimistically pricing to perfection right now is not about chips and software, but rather about having the juice for it. And unless renewable and nuclear buildouts speed up, the biggest bottleneck to the AI dream won't be innovation — it'll be power.

However, before that elephant in the room becomes unignorable, the market might see the bubble pop.

"Someone is going to lose a phenomenal amount of money. We don't know who, and a lot of people are going to make a phenomenal amount of money," Altman said, reflecting on insane valuations of some startups that are a "three people and an idea."

Still, he believes that, on the whole, this would be a huge net win for the economy. After all, dot-com turned out just the same.

Read Next:

From Elevated To Worrisome — Gravity Always Wins, Billionaire Investor Says

Image: Shutterstock