This week the AI rally finally took a real punch to the gut which it does from time to time. The priced-in lesson we have discussed on this site scaled all the way up from individual stocks to the whole market, and Cerebras kept drifting back toward where it started. Three items, all connected. Let’s get into it.
1. Broadcom and the AI bar that finally couldn’t be cleared
Two weeks ago I wrote about NVIDIA’s (ticker NVDA) earnings. Strong numbers, expectations already in the price, and the stock fell. Last week I wrote about Dell (ticker DELL). Strong numbers, expectations nowhere close to the actual result, stock exploded 32% in a day. The principle in both cases was the same one I keep coming back to: the market does not necessarily move on good earnings. The market moves on the difference between what was expected and what arrived.
This week was Broadcom’s (ticker AVGO) turn, and it gave us the sharpest version of the lesson yet.
Broadcom reported earnings Wednesday after the close. Revenue of $22.2 billion, only slightly below consensus estimates. EPS was $2.44, which did beat the estimates. AI semiconductor revenue (a component of the overall revenue) was $10.8 billion, up 143% year over year. Generally, 143% growth would be considered enormous. The company guided that Q3 revenue would be $32.4 billion which was also above expectations. By any normal measure these are unbelievable numbers from a company executing at the heart of the AI buildout.
The stock fell 14% the next day, and closed Friday at $385.73. From its Tuesday closing high of $481, that’s a one-week haircut of about 20%.
Why? The Broadcom CEO did not raise the company’s long-term AI revenue target. He reiterated the existing $100 billion 2027 figure. The market was hoping he would raise it. The disappointment was not that the company performed poorly. The disappointment was that the company performed exactly as well as expected, and the market had been pricing in something better.
Stop and read that again, because it is genuinely remarkable. A company that doubled its AI business in a year, beat earnings, raised near-term guidance, and reiterated a $100 billion long-term target was punished because it did not exceed the bar by enough. That is not a sign the AI rally is over. It is a sign the bar has gotten so high that anything less than “extraordinary plus a little extra” now reads as disappointing.
Three weeks, three lessons, same principle: NVIDIA showed us that priced-in good news can result in a falling stock price or, at least, not the positive reaction one new to all of this might expect. Dell showed the opposite: good news not priced in can result in a significant stock price increase. Broadcom showed us that even great news falls hard when the market was secretly expecting greater. The further you stretch expectations, the more news a company needs to deliver just to stay flat.
Personal note, in keeping with the honesty I try to bring here. I owned Broadcom before all of this. Earlier in the run I trimmed some of the position around $427, taking some profit off the table. The stock kept going up afterward — to a high of $481 — and I did not look back. One of the disciplines I try hardest to keep is to not worry if I left profit on the table after exiting a stock. The trade was sound when I made it. Now the stock is back around $385, well below where I sold it, and I will likely rebuild some of that position at the lower price. Same stock. Same long-term thesis. The market handed me both sides of the same trade. That is what discipline actually does even though it absolutely does not work every time. Build your own rules, stick to them. If ever you find yourself unable to do so, please seek an advisor who will certainly take your emotions out of the equation.
2. The worst Nasdaq day in over a year, and why it wasn’t noise
Friday was rough across the entire market. The Dow lost 695 points, down 1.35%. The S&P 500 lost 200 points, down 2.64%. The Nasdaq lost 1,121 points, down 4.18% — its worst single day since April 2025. The Russell 2000 lost 3.47%. Chip stocks led the rout. NVIDIA fell 6.2% to close at $205.10. Broadcom’s slide continued. Even names that had been quietly riding the rally for months got hit.
Now here is the part that matters most. There are two very different reasons a market can drop hard, and a regular investor needs to be able to tell them apart. I have written about this distinction before, but Friday gave us a textbook case.
Reason one: noise. The market reacts to news that does not actually change the underlying businesses. We saw that earlier in the week. Iran stopped negotiating with the United States and threatened to block the Strait of Hormuz. Oil jumped about 6% on the news. Equity futures swung from positive to negative inside an hour. None of that changed anything about whether Dell can sell servers or whether Micron’s (ticker MU) memory is in demand. The headlines were real. The market reaction was real. But the companies were exactly as good Monday morning as they were Friday at close. That is noise. Painful for prices in the short term, irrelevant to the long-term value of the businesses themselves.
Reason two: actual repricing of expectations. The market processes new information that truly changes what assets should be worth. That is what happened Friday.
Friday morning, the May jobs report came out. The market was expecting roughly 80,000 new jobs. The actual number was 172,000 — more than double the consensus estimates. Prior months were also revised upward by another 93,000 jobs combined. By every measure, the labor market is stronger than people had been assuming.
Here is what made it visible in real time. I leave the house by 4 a.m. most days, and I have CNBC on by the time I am in the car. So I was watching pre-market trading Friday morning when the Bureau of Labor Statistics (BLS) — the government agency that releases the monthly jobs report — released the jobs number at 5:30 a.m. PST. Chip stocks had been drifting lower into the open already. The moment the jobs number hit, the entire market took a sharp leg down within minutes. You could watch the repricing happen — the bond market moved first, equities followed, and the whole tape turned within the same trading window. That is what a market processing genuinely new information looks like. Not noise. Not a panic. Just the system absorbing a data point that changed the math.
That should be good news, by the way. A strong labor market in real terms is what every American wants. But the equity market had been quietly pricing in the possibility of Federal Reserve rate cuts later this year. A strong jobs report kills that hope or, at a minimum, reduces the likelihood of it happening. If the economy is this strong and inflation is still elevated, the Fed has no reason to cut. Some traders are now pricing in the possibility of a rate hike later this year, not a cut.
Having interest rates higher for longer hit growth stocks the hardest, because so much of their value comes from earnings expected years in the future. When rates go up, those future earnings are worth less in today’s dollars. Chip stocks and AI plays are exactly the category that gets discounted hardest. That is why the Nasdaq took the worst of it on Friday, and why this drop is not the same animal as the Iran headlines from earlier in the week.
The Iran moves were noise. Friday’s drop was the market doing its job — repricing assets when the underlying assumptions changed. The bond market repriced Fed expectations the moment the jobs report hit, and the equity market followed. That is not a malfunction. That is the system working.
A regular investor who can tell these two situations apart — noise versus repricing — is genuinely operating at a higher level than most. When it is noise, you wait. When it is repricing, you reassess. Different responses, because they are different problems.
You can watch this exact dynamic play out in equities too. In my Weekly Notes from June 6, 2026, I describe the morning a strong jobs report rewrote Fed expectations within minutes and the entire stock market took a sharp leg down — the bond market moved first, equities followed, all in the same trading window. Same mechanic, different markets.
3. Cerebras keeps drifting back to where it started, and Anthropic gets in line
Three weeks ago I wrote about Cerebras coming public in the biggest IPO of the year. The post warned about the gap between the $185 insider IPO price and the $311 first-day close, and the much larger gap up to the $386 intraday high. The lesson was that the people who chased the hype near the open often got hurt.
Here is where Cerebras stands today. The stock closed Friday at $200.18. That is roughly 35% below its first-day close and almost 50% below its intraday high. The drift has been steady. Each week it gets a little closer to the $185 insiders paid. Every chase higher than that price is now underwater. The lesson did not require waiting long to see it confirmed.
And here is where the IPO discipline matters most going forward. Anthropic, the AI company behind the Claude assistant, confidentially filed paperwork with the SEC this week to begin the process of going public. It is not happening tomorrow. A confidential filing is the first formal step in a process that typically takes months. But it is happening. For me it is exciting. Out of college working for KPMG, a large accounting firm, I worked on multiple public offerings. It is very exciting and I am living vicariously through my younger self.
When Anthropic eventually lists, the same lessons that applied to Cerebras will apply to it. The headline price you read will not be the price you can buy at. The first-day pop, if there is one, is the most dangerous moment to chase. The hype crowd will not check the fundamentals. The disciplined crowd will. That said, nobody truly knows what will happen on day one of trading or what outside news from the world might disrupt things. The words here are just to be careful.
One last thought to close the week. Friday’s drop was real. The AI rally is not invincible. The bar is high, the Fed is not coming to the rescue, and even the best companies in the supply chain are showing they can fall when expectations get too far ahead of them. None of that is bad news for someone who built rules before the rally started. The rules that protect you during a rough day are the rules you set when nobody needed them. The moment you feel the temptation to abandon those rules — either chasing on the way up or panicking on the way down — is the moment to read them again.
The rally being interrupted is not the same as the rally being over. Sound businesses are still sound businesses. Some of them are now slightly on sale relative to last week. Whether to act on that, and how much, depends on the rules you already set. Not on what Friday’s headlines told you.
See you next Saturday.
Nothing here is personalized financial advice — just one person’s notes. The companies named are examples, not recommendations. Always do your own homework before you invest.
