AI × THE DOT-COM BUBBLE
Is the AI boom 2000 all over again? And if it pops, who is left standing?
The dot-com bust wiped out many companies, yet its technology and infrastructure were picked up by others, and the real winners came afterwards. This site lays the AI boom alongside that history on the same yardsticks, and every day checks the warning signs that preceded the 2000 crash.
Where are we?
46 months after ChatGPT's launch (30 Nov 2022), the Nasdaq Composite stands at 2.36 times its starting level. The same 46 months after Netscape's IPO (9 Aug 1995), in Jun 1999, the Nasdaq Composite was at 2.47 times, and it went on to 5.02 times at its peak in month 55.
Heat check: 2000's peak vs now
The same measures at the 2000 peak and today. Where today is higher, the bar turns hot.
Private investment in IT equipment and software ÷ nominal GDP. Q4 2000 → Q2 2026
2000 side: 1999–2000 high. Mar 2000 → Sep 2026
2000: Cisco at its peak; now NVIDIA. Mar 2000 → Sep 2026
On trailing earnings. 2000: Cisco at its peak; now NVIDIA. Mar 2000 → Sep 2026
Includes servers. 2000 side: 1997–2000 high. Q3 1997 → Q2 2026
Warning signs: 0 of 5 lit
Conditions that were in place when the 2000 bubble broke. A lit lamp is not a forecast of a crash; it means one of the 2000 conditions is present.
-
Rate hikesOff-0.70pt (3.63%)
Policy rate up 0.5pt+ on a year earlier
2000: the Fed hiked 1.75pt from June 1999; the Nasdaq peaked mid-cycle.
-
Credit tighteningOff+0.20pt (2.80%)
High-yield spread 1pt+ above its 6-month low
2000: the junk bond market shut and the debt-funded carriers ran out of money first.
-
Spending beyond earningsOff83%
Big-five capex above operating cash flow (last 4 quarters)
2000: the carriers spent far beyond their earnings and borrowed the rest.
-
Chip stocks breakingOff+24.2% vs 200-day
PHLX Semiconductor Index below its 200-day average
2000: the chip index broke its 200-day average six months after the peak (September 2000); later rebounds did not last.
-
The leader stallsOffNVIDIA -4.5%
Largest chip stock 25%+ below its 52-week high
2000: Cisco lost over 70% of its value within a year of the peak.
Three lessons from 2000
- The winners were not those who built the infrastructure but those who picked it up cheaply
The fibre builders went bust from overbuilding, but the fibre stayed, and Google and YouTube grew on top of it once it was cheap.
- The pick-and-shovel sellers did not escape
Cisco and Intel had real sales and profits, but when customers stopped spending their shares fell 80–90% and took more than 25 years to regain their 2000 highs.
- Survivors had light debt and cash coming in
The casualties were debt-funded builders (the carriers) and firms that burned their cash before revenue arrived (Webvan, Pets.com). Amazon's shares fell over 90%, but it reached profitability before the money ran out.
The full story: timeline, who went bust, who picked up the pieces →
If it pops: layer by layer
| Layer | Same layer in 2000 | Revenue growth | Operating margin | Capex / revenue | P/E | P/S | Warning flags |
|---|---|---|---|---|---|---|---|
| Physical infrastructure (data centres, power) | Fibre networks and carriers | +26% | 19% | 12% | 27x | 3.0x | 2 |
| Chips and networking gear | Telecom gear (Cisco, Lucent, Nortel) | +56% | 48% | 3% | 44x | 18.9x | 2 |
| Cloud and compute landlords | Hosting and data centres (Exodus etc.) | +20% | 32% | 35% | 21x | 6.4x | 6 |
| Services and applications | Portals and e-commerce (Yahoo, Amazon, Pets.com) | +22% | 36% | 1% | 28x | 8.5x | 2 |
Medians of the companies listed on the Survivors page (latest four quarters from SEC filings).
Updated 28 Sep 2026 10:50 JST · prices to 25 Sep 2026