Decompose Internet bubble

In the 2000–2002 dot-com bust, the Nasdaq Composite fell 78% (5,048 to 1,114) and the Nasdaq-100 fell ~82%. Roughly 75% to 85% of that drawdown was pure revenue-multiple (P/S) compression.

The S&P 500 Information Technology sector P/S ratio fell from ~7.8x at the March 2000 peak to roughly 1.5x–2.0x at the 2002 trough—a ~75%–80% multiple contraction. At the peak, roughly 11% of the S&P 500 by weight traded above 20x sales; by late 2002, those multiples had been wiped out by 80%–95%.

Cisco: 38x sales to ~3x sales

Intel: ~15x+ sales to 3.8x sales

Amazon: ~23x–30x sales to ~0.65x–0.70x sales


There are two side of the bubble:

1) For Telecom Equipment Makers (Double Whammy: Multiple + Earnings): Lucent, Nortel, and JDS Uniphase got hit by both. Carriers borrowed hundreds of billions to lay fiber (85%+ of which sat unused/”dark” in 2001–2002), and vendors used vendor financing to keep orders flowing. When credit dried up, both the 100x multiple and the revenue disappeared simultaneously—leading to 90%–99% wipeouts and bankruptcies.

2) For Software & Internet Survivors (Pure Multiple Bubble): Microsoft’s earnings nearly tripled over the next decade and Amazon’s revenue grew 15x, yet both stocks crashed 65%–95% and took 10–16 years to break even. The businesses weren’t broken—starting at 25x–30x sales or 70x earnings simply left zero mathematical path to positive returns.


From this perspective, internet bubble is more of a valuation bubble especially for the internet & software side, while the telecom equipment maker did have a small earnings bubble.

While currently, AI bubble is more of an earnings bubble on the hardware side, while the AI labs/software side could be a valuation bubble (no earnings!).