Sandisk Stock Plunged 14% in a Day. Is the AI Memory Boom Cracking?

On July 2, Sandisk (SNDK +0.10%) stock fell about 14% in a single session, closing at $1,745. It wasn’t alone. The sell-off tore through the entire memory complex — Micron dropped, and shares of Samsung and SK Hynix fell sharply in Seoul.

For a stock that had risen more than 700% this year, a big one-day drop like that naturally raises the question: Is the artificial intelligence (AI) memory boom finally cracking?

A chart showing a stock price declining.

Image source: Getty Images.

What actually happened

Interestingly, the trigger for the stock’s sell-off came from outside the memory industry. A day earlier, reports that Meta Platforms (META +3.12%) plans to sell its spare AI computing capacity to outside customers stoked fears that the AI compute shortage is easing — and that was enough to rattle anyone betting on AI hardware. Memory, the hottest corner of that trade after a blistering rally, took the brunt.

It helps to remember how far these stocks had run. Sandisk had been one of 2026’s biggest winners before the drop, and its rivals had also soared. After a move like that, it doesn’t take much bad news to spark a violent pullback — profit-taking feeds on itself. What it didn’t take was any change at Sandisk. Nothing about the company’s own business shifted on July 2. This was a sentiment reset after a red-hot run, not a sign that demand for its chips had softened.

Sandisk’s business tells a different story

If the boom were truly cracking, you would expect it to show up in Sandisk’s numbers. It doesn’t.

The company — spun out of Western Digital early last year — makes NAND flash, the storage memory that increasingly feeds AI data centers. That is a different product from the high-bandwidth memory grabbing headlines, but the same AI build-out is driving it: models and their outputs have to be stored somewhere, and enterprise storage drives have become a booming business. In its fiscal third quarter (the period ended in early April 2026), revenue nearly doubled from the prior quarter to $5.95 billion, and non-GAAP (adjusted) gross margin jumped to 78.4% from 51.1% three months earlier. Adjusted earnings per share reached $23.41. The data center storage line, specifically, tripled — with revenue up 233% from the prior quarter to about $1.5 billion.

More telling still is what’s locked in. Sandisk said it has signed multiyear supply agreements worth about $42 billion in minimum contracted revenue — customers committing years in advance to secure supply. Management then guided for fiscal fourth-quarter revenue of $7.75 billion to $8.25 billion, well above the quarter it just reported. That is not the profile of a market about to roll over.

Sandisk Stock Quote

Today’s Change

(0.10%) $1.65

Current Price

$1746.65

Reset or top?

But maybe the question is less about whether the AI boom is cracking but more about whether or not some stocks benefiting from the AI boom have become overvalued.

After the plunge, Sandisk trades at about 59 times its earnings — but that multiple is set to shrink quickly, because profits are still climbing sharply. Management’s guidance for the just-ended fiscal fourth quarter alone calls for as much as $33 in adjusted earnings per share. Annualize a pace like that, and the stock changes hands for little more than a dozen times forward earnings — the market pricing in a downturn that hasn’t arrived.

And that is the central question with any memory stock: not whether demand is strong now (it clearly is), but when the cycle turns. Memory has always run in booms and busts, and NAND flash has historically been one of the most commoditized, price-sensitive corners of the chip world. Prices spike when supply is short, then crater when the industry overbuilds. Sure, that $42 billion in contracted revenue helps, but it doesn’t eliminate the risk of cyclicality in the overall sector. The bet buried inside Sandisk’s forward multiple is that this AI-driven upturn runs longer and steadier than the ones before it.

So is the boom cracking?

I don’t think July 2 was the crack. It looks far more like a valuation reset amid a shortage that is still very much intact — the demand signals, the surging margins, and that $42 billion backlog all point in the same direction. But I would stop short of calling the dip a clear-cut bargain. Memory stocks have a long history of punishing investors who mistake the top of a cycle for a new normal. Overall, I’d like to see an even cheaper valuation before I buy — one that gives me an even bigger margin of safety to handle a potential slowdown in the AI boom.

Similar Posts

Leave a Reply

Your email address will not be published. Required fields are marked *