"History doesn't repeat… but it rhymes." — Mark Twain
◉ THE PRESENT
Broadcom doubled its AI revenue last quarter and the stock fell 14% anyway. That is the whole story of June 4th. The chip maker that became Wall Street's favorite way to own the AI buildout reported AI sales of $10.8 billion, up 143% from a year ago, then watched its shares get cut after CEO Hock Tan declined to raise the full-year target and said the company would sell "chips only" instead of full systems. The number was huge. It just wasn't bigger than huge, and that was enough.
AVGO −14% after hours | AI revenue $10.8B (+143% y/y) | CrowdStrike −11% | S&P 500 just snapped a 9-day win streak from a record above 7,600
When the best name in a trade puts up a doubling and still gets sold, the market is telling you something about the price, not the company. We have watched a networking-and-silicon king get crowned at the top of a technology boom exactly once before, and it happened on a spring afternoon in 2000.
◉ THE ECHO — MARCH 27, 2000
For one week, a company that made boxes was worth more than any company on earth.
On the afternoon of March 27, 2000, Cisco Systems traded at $80 a share. Two days earlier it had passed Microsoft to become the most valuable company in the world, worth $579 billion. Most Americans had never used a Cisco product and never would. The company sold routers and switches — the plain metal boxes that moved internet traffic from one place to another. But every dot-com, every phone company, every start-up burning venture money on a website needed those boxes, and Cisco sold the picks and shovels to all of them. John Chambers, the CEO, told anyone who would listen that the internet would change everything, and for a while the stock acted like he was right.
The growth was real. This is the part people forget. Cisco's revenue was climbing more than 50% a year. Orders were pouring in faster than the factories could fill them. Chambers ran a famous "virtual close" that let him see the company's numbers almost in real time, and the numbers were beautiful. Analysts competed to raise their targets. The stock had risen 236% in about fifteen months. At the peak it traded at 201 times earnings, which means a buyer was paying two centuries of profit for a single share and calling it cheap.
Nobody rang a bell. There was no bad quarter, no scandal, no warning. Cisco kept beating estimates straight through the top. What changed was subtler and harder to see. The customers buying all those boxes were dot-coms running on borrowed money and telecom upstarts laying fiber no one was using yet. Some of them were buying Cisco gear with financing Cisco itself had arranged. The demand looked like a wall. It was actually a balloon.
By the time anyone understood the difference, the most valuable company in the world had already seen its best day. March 27 was the high. It would not be seen again for a very long time.
◉ THE RHYME — WHAT'S IDENTICAL
The danger was never that the king's business would shrink. It was that the price had already booked a future that even a doubling couldn't beat.
◉ THE DIVERGENCE — WHAT'S DIFFERENT THIS TIME
The rhyme is close, but a few notes are off. These are where your edge lives.
The multiple is not 201. Cisco at its peak traded at two hundred times earnings on a hope. Broadcom makes real money — tens of billions in free cash flow — and trades at a fraction of that. The price is stretched, not deranged. A name can be expensive without being a 2000 Cisco, and that gap matters when you decide how far a crack can run.
The customers can pay. Cisco's demand came from companies burning venture capital and selling stock to fund fiber that sat dark for years. Broadcom's biggest customers are the most profitable companies on the planet, funding AI chips out of cash flow, not loans. The buildout could still slow. But it is not standing on the same borrowed sand.
Broadcom is not a one-trick box maker. Cisco was routers and switches and little else. Broadcom is half a software company now, with steady, sticky enterprise revenue from VMware sitting underneath the AI fireworks. When the cycle turns, there is a floor under it that Cisco never had.
The macro is heavier this time. In March 2000 the Fed was the only worry. Today the same week brought fresh US-Iran strikes, oil back near $96, and the 30-year Treasury yield pushing toward 5%. A momentum trade priced for perfection is being asked to hold up against a rising cost of money — a headwind Cisco's bull market never faced at the top.
◉ THE RECKONING — WHAT HAPPENS NEXT
Here is how it played out for Cisco, because the timeline is the lesson. The stock peaked on March 27, 2000, and the business kept looking fine for almost a full year. Revenue rose. Chambers stayed bullish. The believers held, and for months they looked smart, because the numbers gave them no reason to sell.
Then the wall turned out to be a balloon. In the spring of 2001, the orders that had outrun the factories suddenly stopped. On April 16, 2001, Cisco took a $2.25 billion charge to write off inventory it had built for demand that never came. Quarterly profit fell 77%. The same real-time numbers that had looked so beautiful now showed a company that had been building boxes for customers who were going bankrupt. The stock, already down hard, kept falling. From the $80 peak it lost about 80% of its value and wiped out more than $400 billion. It would take roughly two decades — until 2024 — before Cisco's share price saw March 2000 again.
The smart money didn't need to call the exact top. It noticed something simpler: when the best company in a boom puts up great numbers and the stock falls anyway, the marginal buyer is gone. Price stops rewarding growth. That is the signal, and it showed up at Cisco a full year before the inventory charge confirmed it. The ones who trimmed into strength, rotated toward boring cash flows, and stopped paying any price for the leader kept most of what the buildout had handed them.
Broadcom is not Cisco, and 2026 is not 2001. But last night the best name in the AI trade doubled its AI revenue and got sold. That is the same tell, arriving at the same place in the cycle.
When the leader's blowout quarter is met with a 14% drop, the news to watch isn't the earnings — it's the reaction. A market that punishes a doubling is a market that has stopped paying up for growth. In 2000, that shift showed up roughly a year before the fundamentals broke. The pattern doesn't tell you to sell the AI buildout. It tells you to stop paying any price to own the king of it.
◉ TOMORROW’S WATCH
Watch whether the selling stays inside the chip complex or leaks into the hyperscalers' capex guidance — the dollars that actually fund Broadcom's order book. If a cloud giant so much as hints at trimming AI spending, that is the 2001 inventory moment arriving early, when the wall of demand quietly turns into a balloon.
