Bubble Watch

1835 → today

200 years of bubble concentration

Six times in two centuries, one theme swallowed the stock market — railroads, the 1929 utilities, the Nifty Fifty, Japan, dot-com, and now AI. Every previous episode peaked between 36% and 63% of the market… and then broke. The AI Big 10 just reached 40%.

…and what happened next

Railroads1873 · 63%

America's first tech mania — railroads were the internet of the 1800s. Investors financed thousands of miles of track, much of it never profitable, until rail stocks were most of the entire market.

Northern Pacific collapse sparked the 1873 panicJay Cooke & Co. failedUnion Pacific bankrupt 1893Erie Railroad bankruptReading Railroad bankrupt

→ Panics of 1873 & 1893 — a third of US railroads went bankrupt.

Utilities / telco / industrials1929 · 36%

The 'new era' stocks of the Roaring Twenties — electric utilities, telephone and industrial holding trusts, pyramided with 10:1 margin loans into the 1929 boom.

RCA -98%General Electric -90%US Steel -92%Insull utility trusts wiped outGoldman Sachs Trading Corp -98%

→ 1929 crash — the market fell 86% and took 25 years to recover.

Nifty Fifty1972 · 40%

The ~50 'one-decision' blue chips of the early 1970s — Coca-Cola, IBM, Xerox, Polaroid, Disney. The story: quality so good you buy at any price and never sell. They traded at 50–90× earnings.

Polaroid -91%Avon -86%Disney -82%Xerox -71%Coca-Cola -69%IBM -59%

→ 1973–74 bear — the 'one-decision' stocks fell 60–90%.

Japan (% MSCI ACWI)1989 · 44%

The 1980s Japanese miracle. At the 1989 peak Japan was 44% of world market cap, the Imperial Palace grounds were 'worth more than California', and 8 of the world's 10 biggest companies were Japanese.

NTT world #1 → -74%Nomura -70%Industrial Bank of Japan -90%+Sony -60%Mitsubishi Estate -80%

→ Nikkei fell 82% — needed 34 years to reclaim its 1989 peak.

TMT (dot-com)1999 · 41%

Tech-Media-Telecom — the dot-com internet boom. Anything with '.com' soared; Cisco briefly became the world's most valuable company at ~200× earnings; most dot-coms had no profits at all.

Yahoo -97%Sun Micro -96%Amazon -94%Cisco -86%Intel -82%Microsoft -65%Pets.com bankruptWorldCom bankrupt

→ Nasdaq fell 78% — 15 years back to its 2000 peak.

AI Big 10 (Mag 7 + AVGO, AMD, MU)2026 · 40%

The generative-AI boom since ChatGPT — ten chip and platform giants (Mag 7 + Broadcom, AMD, Micron) funding a half-trillion-dollar-a-year datacenter build-out on very real profits.

NVDA ?MSFT ?AAPL ?GOOGL ?AMZN ?META ?TSLA ?AVGO ?AMD ?MU ?

→ ? — you are here.

The pattern

What every crisis has in common

Different decade, same recipe. Across two centuries, the same five ingredients show up at every top — and one of the most reliable tells is the mega-IPO: the era's biggest, hottest listing almost always arrives within months of the peak, because record demand for paper is exactly what a top is.

• 1A 'new era' story — railroads, electricity, the internet, AI. The technology is always real; the prices are not.
• 2Cheap money & leverage — 10:1 margin loans in 1929, the yen carry trade of the 1980s, 1% Fed rates 2003–07, zero rates 2020–21.
• 3Retail euphoria — shoeshine-boy stock tips in 1929, day-traders in 1999, meme stocks in 2021, AI-picked portfolios now.
• 4A capex arms race — railroad track, fiber-optic cable, datacenters. The boom's spending becomes the bust's overcapacity.
• 5A mega IPO near the top — see the table below.

The Big IPO tell — landmark listing vs market peak

Crisis	Landmark IPO	IPO date	IPO → peak	IPO fate
1929 Great Crash Sep 1929	Goldman Sachs Trading Corp Hottest investment trust of the boom	Dec 1928	9 mo before peak	-98%
	Insull Utility Investments Chicago utility empire holding trust	Jan 1929	8 mo before peak	wiped out
	Blue Ridge Corp Goldman's 2nd trust — leveraged on the 1st	Aug 1929	1 mo before peak	-99%
Nifty Fifty 1973–74 Jan 1973	Intel The hot tech listing of the era	Oct 1971	15 mo before peak	-80% by 1974
	Playboy Enterprises Icon-brand IPO of the go-go years	Nov 1971	14 mo before peak	-83%
	Levitz Furniture Glamour retailer at 100x earnings	Jul 1968	4.5 yrs before peak	-97%
Japan 1989 Dec 1989	NTT Biggest IPO in history at the time	Feb 1987	34 mo before peak	-74%
	Recruit Cosmos Shares-for-favors scandal that shook Japan	Oct 1986	38 mo before peak	scandal, collapsed
	Japan Airlines Full privatization at the boom's height	Nov 1987	25 mo before peak	~-70% by 1992
Dot-com 2000 Mar 2000	Palm Day one worth more than GM	Mar 2000	same month as peak	-95%
	Pets.com Sock-puppet mascot, no profits	Feb 2000	1 mo before peak	bankrupt in 9 mo
	theglobe.com +606% on day one — a record	Nov 1998	16 mo before peak	-99%, delisted
	AT&T Wireless Biggest US IPO ever at $10.6B	Apr 2000	1 mo AFTER peak	-70%+
GFC 2008 Oct 2007	Blackstone Biggest US IPO since 2002	Jun 2007	4 mo before peak	-80% by 2009
	Fortress Investment First US hedge-fund IPO	Feb 2007	8 mo before peak	-95%
	MF Global Broker spun out at the credit top	Jul 2007	3 mo before peak	bankrupt 2011
Rate-hike bear 2022 Jan 2022	Rivian $86B day one — worth more than Ford	Nov 2021	2 mo before peak	-93%
	Coinbase Crypto exchange at the crypto top	Apr 2021	9 mo before peak	-90%
	Robinhood The meme-stock broker itself	Jul 2021	6 mo before peak	-90%
AI 2026 ?	CoreWeave Biggest AI-infra IPO ever	Mar 2025	?	?
	Circle +168% day one — stablecoin mania	Jun 2025	?	?
	Figma +250% day one — hottest tech IPO since 2021	Jul 2025	?	?

Nineteen flagship listings across six historical episodes — almost every one hit the tape within months of the top (Palm listed the same month as the peak; Blue Ridge one month before Sep 1929). Record IPO waves don't cause the crash — they mark the moment demand peaks.

The fuel

USD liquidity — loose or tight right now?

Right now: LOOSE, and getting looser. Every bubble runs on fuel, and all four dollar dials point the same way — money supply at a record high, the Fed cutting rates, its balance sheet growing again, and the $2.5T overnight parking lot (reverse repo) fully drained back into markets.

Loose liquidity is why 'expensive' can keep getting more expensive. It is also the classic bubble-popper in reverse: 1929, Japan 1989, 2000 and 2022 all ended when the money got tight — so the number to watch is not the price, it is the fuel.

The dashboard

How stretched is this market?

Three of the four classic bubble gauges are at or near all-time records. Only one — the price investors pay per dollar of actual profits at the top — looks tamer than 2000.

Concentration

Never this few, never this big

The S&P 500 is 500 stocks, but 10 of them are 44% of the money. Concentration researchers at BofA and Bianco Research call it the most concentrated US market in a century.

Top-10 share of S&P 500 market cap

The dot-com peak — the previous concentration scare — topped out near 25%. Today is ~44%.

The mega-caps, by market cap ($T)

NVDA · Nvidia

$5.55T

P/E 29

AAPL · Apple

$4.83T

P/E 38

GOOGL · Alphabet

$4.15T

P/E 17

MSFT · Microsoft

$3.78T

P/E 28

AMZN · Amazon

$2.67T

P/E 20

META · Meta

$1.83T

P/E 27

AVGO · Broadcom

$1.70T

P/E 46

TSLA · Tesla

$1.39T

P/E 330

Nvidia alone ($5.55T) is worth ~9.3% of the entire S&P 500 — the largest single-stock weight ever recorded. The Magnificent 7 together are ~$24.2T, roughly the GDP of the US.

Valuation

Priced like 1929 and 1999

Valuation never times the top — but it has always set the size of the fall. Every era that paid these prices spent a decade or more earning them back.

The mania

The AI trade is the market

Since ChatGPT (Nov 2022) the S&P 500 has roughly doubled — and about 70% of that gain came from AI-linked stocks. The bull case and the bubble case are now the same trade.

History

What crashes actually look like

Seven times in a century the US market lost a quarter or more. The pattern repeats: the higher the starting valuation, the deeper the fall and the longer the climb back.

The playbook

Anatomy of a bubble

Economist Hyman Minsky and Jean-Paul Rodrigue mapped the phases every bubble repeats. Ask yourself honestly which label is closest to today.

So what

Then vs now — and what to do

2026 is not a copy of 2000: today's giants earn enormous real profits. But concentration and the price of the whole market are past every historical extreme.

• Concentration means index funds are no longer diversified — know what you actually own.
• High CAPE has never predicted the month of the top, but it reliably predicts low 10-year returns from here.
• Real profits at the top (unlike 2000) mean a crash needs an earnings disappointment, not just a mood change.
• Crashes are survivable when you hold cash, size positions honestly, and never use leverage into a euphoric top.

Educational research, not investment advice. Figures are approximate, hand-curated from public sources: Robert Shiller's data, currentmarketvaluation.com, Yahoo Finance, BofA / Bianco Research via Adam Tooze's Chartbook #396.