The Rhyme: AI Stocks Crack on Circular Deals & Its 2000 Echo

"History doesn't repeat… but it rhymes." — Mark Twain

Monday closed ugly. The Nasdaq Composite fell 5.8% on its heaviest volume since the August 2024 yen unwind, and the selling didn't come from the Fed or the Middle East or a tariff headline.

It came from a fifteen-page research note published before the open by a mid-sized sell-side shop, walking quietly through what most traders already suspected — that the biggest AI chip customers were also the chip makers' equity partners, and the money had been moving in a circle for two years. Nvidia led the drop. Microsoft, CoreWeave, Oracle, and AMD followed within the hour.

Nasdaq Composite − 5.8% | Nvidia − 11.3% | VIX 34.6

The last time a concentrated tech market cracked on circular-money concerns, the calendar was within six days of today's date. It was April 14, 2000.

The March consumer price report hit the tape at 8:30 in the morning on Friday, April 14, 2000. Headline inflation came in at 0.7% for the month against an expected 0.4%. Core was hot too. By the time the opening bell rang at the New York Stock Exchange, Nasdaq futures had already given back a hundred points, and the Street was bracing.

It was supposed to stabilize after the first hour. It didn't. By ten a.m. the Nasdaq had given back another two hundred, and the big names everybody owned were breaking in lockstep. Cisco, the most valuable company in the world the week before. Intel. Sun Microsystems. Oracle. JDS Uniphase. Every one of them had been a ten-bagger in eighteen months. Every one of them was suddenly down double digits before noon.

By the close, the Nasdaq Composite had fallen 355 points — 9.67% — the worst single-day point loss in its history at the time. Nearly a trillion dollars of market value was gone by the time traders left the floor for the weekend. The champagne was still in the ice buckets at the restaurants on West Broadway where tech bankers had been celebrating on Monday. By Friday night those same bankers were standing in the street on their phones.

What had cracked that week wasn't the economy. The economy was still fine. What cracked was the quiet math nobody had wanted to look at — that Cisco was lending money to telecoms so they could turn around and buy Cisco gear, that Lucent was taking equity stakes in startups that bought Lucent switches, that Nortel was selling to companies it had helped finance. The revenue was real on the books. The cash wasn't. Somebody on the Street finally wrote it down in a way the buy-side couldn't ignore, and the dot-com peak of March 10, 2000 — Nasdaq at 5,048 — turned overnight into a ceiling nobody was ever going back to.

The script on Wall Street today rhymes almost word for word. The vendors are Nvidia, AMD, and Broadcom instead of Cisco and Lucent. The customers are OpenAI, Anthropic, CoreWeave, and xAI instead of Global Crossing and WorldCom. The money runs in the same loop. The note that hit the tape this morning just did the math out loud.

Same mechanism, same concentration, same moment when somebody finally does the math out loud.

This is where the pattern breaks. Honest differences — not reasons to ignore the rhyme, but the angles that give a careful reader an actual edge.

Here is what happened after April 14, 2000. The Nasdaq rallied 19% off the lows over the next three weeks and every talking head on CNBC said the bottom was in. The buyers of that rally were mostly retail. By May 23 the index rolled over again. It ground lower for the next two and a half years, and by October 2002 it sat at 1,114 — down 78% from the March peak.

Cisco lost 89% of its value and took fourteen years to retrace half of it. Lucent fell from $84 to $0.55 before being sold to Alcatel for a discount. Nortel, once 35% of the Toronto index by weight, filed for bankruptcy. JDS Uniphase, Sun, WorldCom — gone, or reduced to name plates on somebody else's door.

But here is the part nobody tells at cocktail parties. The real companies survived. Microsoft was a great business in 2000 and is a great business in 2026. Oracle the same. Amazon fell from $113 to $5.51 and has returned something on the order of 45,000% from that bottom. The index killed its passengers. The operators who boarded at the back of the wreck made the money.

The smart money in 2000 did two things. First, they rotated into the sectors nobody had loved in a decade — energy, industrials, utilities, value. The S&P 500 Energy index returned 170% from March 2000 to October 2007 while the Nasdaq was still flat on the screen. Second, they waited. They didn't buy the first dip. They waited for the real bottom in October 2002 and the deeper one in March 2009, and when they did buy, they only bought the survivors.

When circular money unwinds, the index falls faster than the real businesses behind it. The lesson of 2000 is not that tech was a fraud. The lesson is that concentration breaks harder than breadth does — and the operators who buy the survivors after the break are the ones who get the compounding. Knowing which bucket a stock sits in matters more right now than knowing which direction the tape goes this week.

Watch the private credit BDCs. If spreads blow out this week while equities are still bleeding, it echoes August 2007, when the quant funds cracked first and told you the bigger unwind was six months away.