SanDisk's $3,000 target is a memory-cycle warning, not a roadmap

When a stock trades near $1,400 and a respected broker says $3,000, do you read a roadmap or a warning label? We lean toward the second. Not because Bernstein is necessarily wrong, but because a target that more than doubles the share price tells you mostly how long the good news is being assumed to last.

What the target actually says

According to Yahoo Finance's coverage, Bernstein raised its SanDisk target to $3,000 from $1,700 on June 30, based on 11 times its fiscal 2028 EPS estimate. Read that sentence again: the number rests on an earnings estimate for fiscal 2028, in an industry whose defining feature is that prices do not stay where they are.

Meanwhile the stock has done its own thing. SanDisk closed at $1,371.12 on October 19, and 24/7 Wall St. notes it is up about 478% year to date but down 23.48% over the past month. It was trading above $2,000 as recently as late June. So the target is not only far above the price; the stock has already shown it can lose a lot of ground quickly. SanDisk reports on October 29.

The business is real; the extrapolation is the problem

None of this means SanDisk is a mirage. Fiscal fourth-quarter revenue rose 371.6% year over year, with an 84.6% GAAP gross margin, and demand from AI data centers is outrunning NAND supply. Management has also tried to dampen the cycle: its new business model agreements cover eight customers with at least $93.9 billion in revenue at floor pricing.

But look at what drove the numbers. About two-thirds of last quarter's sequential revenue growth came from higher prices, not more units. Price-led growth is the most perishable kind. And the contracts cut both ways: BigGo reports that contract volume is expected to earn around 80%, which in our reading limits the downside but also caps the upside that a $3,000 target seems to need. Floor pricing, as 24/7 Wall St. puts it, only limits how far revenue can fall, and the rest of SanDisk's sales move with market prices.

Then there is the P/E. A trailing multiple in the high teens looks cheap. We think that is the classic memory trap: multiples look lowest when earnings are at their peak. The low multiple suggests investors doubt the profits will last, whatever story the bulls tell about more predictable revenue. An 11x multiple in Bernstein's own method looks to us like a quiet admission of the same thing.

Correlated risk, not diversification

The other thing a big target hides is what the stock actually depends on. Jim Cramer's argument, as quoted by 24/7 Wall St., is that SanDisk and Western Digital are "the worst stocks to own" if fewer data centers get built. You can disagree with his tone and still accept the logic: both trade as one AI-spending bet. Microsoft's roughly $175 billion fiscal 2027 capex expectation is the tailwind, but the same source notes that about two-thirds of its quarterly capex goes to short-lived chips it can stop buying. The buyer keeps the flexibility; the supplier carries the cycle.

Compare Micron, which BigGo describes as selling HBM, DRAM and NAND, so a slump in one line is partly offset. SanDisk is NAND only. That is a strength in a shortage and, in our view, the whole risk in a glut.

Even the bullish voices hedge. Morningstar analyst William Kerwin calls SanDisk's AI-fueled growth "tremendous, but finite," per the Stocktwits headline. And the TipRanks-ranked investor Julia Ostian, who rates the stock a buy, said she would "rather split the position around October 29 than go all in before the guidance."

What to do with this

This is not personal advice, and we are not predicting where SanDisk goes. But the framework seems clear. Treat $3,000 as a scenario in which pricing stays extreme for years, not as a forecast. Decide in advance how much of a portfolio can sit in one AI-capex bet, remembering that owning several storage names is barely diversification. Size positions so a price reversal does not force a decision. And on October 29, watch the guidance and the contract pipeline more than the headline quarter.

In cyclical hardware, targets are cheap and discipline is not. The investors who come out of memory cycles well are rarely the ones who believed the biggest number; they are the ones who knew what they would do when the price fell.