Jim Cramer used his June 8, 2026, Squawk on the Street appearance to deliver a direct warning to anyone riding the AI chip rally. His line, after a brutal Friday that saw the Philadelphia semiconductor index post its fourth-worst decline ever: “Unless you have accelerated earnings, your stock is pretty much done. Got to go down 50% and then it doesn’t come back. Sometimes it doesn’t come back for years and years and years. So I just don’t like parabolic moves. They are indeed as dangerous as you think they are.”
If Cramer is right and you own a parabolic AI name, you can lose half your capital and wait years to recover it. If he is wrong, his rule still costs you the next leg of one of the largest earnings cycles in tech history. The test is whether earnings are actually accelerating.
Cramer’s framework is sound, but the data disagrees
Cramer is correct that stocks doubling on no earnings change crater. The problem is that AI semiconductor names are posting some of the sharpest earnings acceleration ever recorded in large-cap tech.
Micron Technology (NASDAQ: MU | MU Price Prediction) exemplifies this. EPS went from $4.78 in fiscal Q1 2026 to $12.20 in Q2, with management guiding $19.15 for Q3.
Revenue grew 196% year over year, and operating income jumped 810%. The stock is up 203% year to date. Quadrupling earnings against a tripling stock is multiple compression.
SanDisk (NASDAQ: SNDK) shows the same pattern more sharply. EPS climbed from $1.22 to $6.20 to $23.41 across three quarters, with Q4 guidance of $30 to $33. The datacenter segment grew 645% year over year.
The stock is up 557% year to date, but EPS is tracking from $1.22 to a guided $30-plus in nine months. That is earnings acceleration outrunning price.
Broadcom (NASDAQ: AVGO) is the cleanest counterpoint. AI semi revenue went $8.4B to $10.8B with Q3 guidance of $16B, growth accelerating from 106% to 143% to a guided 200%-plus.
The stock is up only 12% year to date and just took a hit. Earnings accelerated, multiples compressed, and Cramer is calling it dangerous.
The metric that determines Cramer’s warning
The single number that matters is the ratio of price change to earnings change over the same period.
If a stock doubled and EPS tripled, the multiple shrank, and the rally is supported. If a stock tripled and EPS rose 20%, the multiple expanded, and Cramer’s 50% rule applies.
Run it on NVIDIA (NASDAQ: NVDA). Revenue growth accelerated from 56% to 62% to 73% to 85% year over year across four quarters. The stock is up 10% year to date, with a P/E at 31.
That looks like multiple compressions in real time. Run it on Alphabet (NASDAQ: GOOGL): Q1 EPS of $5.11 beat consensus by 94%, Cloud backlog nearly doubled to $460 billion, and shares are up 18% year to date. Again, earnings ahead of price.
Micron is where the comparison gets uncomfortable. Prediction markets currently show only a 39% probability that MU stays above $1,000 by the end of June, and the AI fair-value model implies 39% downside. That is where Cramer’s rule has teeth.
How to test this yourself
Pull up each AI-exposed name you own and write down two numbers: percentage price change over the last six months, and percentage EPS change (trailing plus next-quarter guidance) over the same window. If price is growing faster than earnings, your multiple expanded and Cramer’s parabolic rule is your risk. If earnings outpaced price, the multiple compressed, and the stock got cheaper while you held it.
David Faber’s point on the same segment is the second test: watch hyperscaler capex. Alphabet’s $175 billion to $185 billion 2026 capex guide is the leading indicator. If that number gets cut, the earnings acceleration thesis breaks, and Cramer’s warning becomes the base case.
Cramer’s rule is right. His chosen targets, for now, are the ones where the math disagrees with him.
Leave a comment