Sandisk

Where we land

Sandisk makes flash memory, the chips that store data in phones, laptops and the servers that run artificial intelligence, and it is earning extraordinary sums from them. Demand from AI data centres has collided with tight supply, chip prices have soared, and the company now keeps 62% of its sales as operating profit. We award it gold.

The share price is more sceptical. At today's price, investors are assuming that earnings will shrink over the next five years: a bet that the price boom ends and takes the margins with it. The question is whether that bet is too gloomy, or simply realistic for a business whose profits have always risen and fallen with the price of what it sells.

What the business is

All of Sandisk's revenue comes from one line, data storage drives and peripherals: flash chips sold as memory cards, as drives for personal computers and as large drives for data centres. The chips are NAND flash, the kind of memory that keeps data when the power is off. One maker's chip does much the same job as another's, so the price is set less by the brand than by how much the industry has produced against how much buyers want.

That is why the margins move so much. Today 71% of every sale is left after the direct cost of making the product, a measure called gross margin, and 62% remains as operating profit after the cost of running the company. Those are remarkable figures for any hardware maker, but they are not typical for this one. Over five years the middle value of its gross margin has been 35%, varying by 14 points either way. This is not a stable-margin business, and today's figure sits at the very top of a wide range.

Why now

AI servers need vast amounts of fast storage, flash supply is tight, and contract prices, the prices large buyers agree with chipmakers for regular deliveries, have climbed steeply. Because higher prices fall almost straight to profit, the effect on earnings has been dramatic.

The shares, meanwhile, have been under pressure. Sandisk reported on 29 October: sales of $11.0 billion, nearly five times a year earlier, earnings per share above what analysts expected, and a forecast of a bigger quarter still. The shares fell 13% the next day all the same, leaving them at half their June peak. On our own price data they are up 5% over six months, a reminder of how far they had run before they fell.

On quality, Sandisk is among the strongest companies we rate, with the widest operating margin among storage companies, yet its shares trade at a modest multiple of earnings. Exceptional profits, a falling share price and an undemanding valuation: that combination is what makes Sandisk worth examining, and what makes it hard to judge.

The case for

Start with the scale of the turn. Sales rose 175% in a year. Operating profit went from $507m to $12.5bn. Twelve-month net profit, what is left after interest and tax, swung from a loss of $1.6bn to a profit of $11.4bn.

The analysts who follow the company do not think the move is over. On average they expect earnings per share to grow another 30% next year, and their forecasts have moved in one direction: in the last 45 days, estimates were raised 36 times and cut only 3 times.

The case for, then, is that the share price assumes an end to the boom that the people closest to the numbers do not yet see. The shares have lost ground over the past three months even as the measured quality of the business climbed sharply over six. The surge reflects a pricing cycle as much as a lasting shift, so the case rests on that cycle lasting longer, or ending more gently, than the market assumes.

Where the growth could come from

Almost everything about Sandisk's near-term growth comes down to the price of NAND chips. Because it sells only flash storage, nearly all of today's earnings growth comes from getting more for every chip rather than from selling vastly more of them, and each further price rise lifts earnings faster than sales. One headline puts the jump in NAND prices so far at an estimated 70%. The industry tracker TrendForce sees a further rise of 10% to 15% this quarter, a much smaller step. Micron, a rival memory maker, expects memory supply to be much tighter in 2028. If those forecasts hold, the growth analysts expect looks reachable; if the cycle turns, as memory cycles always have, it does not.

The second force decides how much of the boom reaches the bottom line. Sandisk has locked part of its supply into long-term contracts with customers, trading some price upside for visibility, and our figures do not show how much volume sits under them. One headline argues these contracts limit how much the company can gain from even tighter supply, and Bernstein, a research firm, has weighed which memory stock has more room to rise. If the contracts cap gains on the way up but cushion prices when the cycle turns, earnings become steadier and could earn a higher valuation. If they only cap the upside, rivals selling at market prices will grow faster.

Beneath the price cycle sits a possible structural line: very large, power-efficient flash drives for AI training and inference clusters, which in some uses are displacing hard disks. Our figures do not separate these data-centre drives from consumer and PC products, but the pattern fits a rising share of them, and they most likely carry richer margins than memory cards or PC drives. A bigger slice would make earnings less hostage to the chip price and could keep growth going after prices peak. Data-centre sales outgrowing the company total, and new approvals at large cloud or AI-server customers, would confirm it.

The force that could end the boom comes from outside. New supply is what usually ends a memory price boom, and Chinese chipmakers are advancing their technology and adding capacity. Sandisk's shares fell 6% on news of China's progress in DRAM, a different type of memory chip. A Chinese flash maker adding meaningful volume would loosen the tight supply behind today's prices and pull earnings growth below expectations.

Finally, there is an option rather than an engine. Sandisk sits largely outside the boom in high-bandwidth memory, the stacked DRAM that feeds AI processors and has enriched other chipmakers. One plausible route in would be stacking flash chips to do a similar job, more cheaply and with more capacity, which would open a new and less price-driven market. It remains speculative until a named AI chip or system maker announces a design that uses it.

The case against

The strongest argument against is that today's margin is an outlier. As recently as June 2025, Sandisk was losing money on a twelve-month basis. Its run of profits is short, so these margins have never been tested through a downturn, and a business that has swung so widely between thin and fat margins can swing back.

Comparison with other storage companies makes the same point from another angle. Western Digital and Seagate run operating margins of around 40%; NetApp, which sells storage systems, manages 26%. Sandisk beats them all, yet its shares trade at only 16 times earnings, about half Seagate's 34 times and close to Western Digital's 14. Investors are paying less for each dollar of Sandisk's profit because they doubt it will last. NetApp's long, steady record of profits shows what they reward instead, and Sandisk does not yet have such a record.

The price tailwind may also be easing. The rise forecast for this quarter is a fraction of the jump already banked, the contracts that steady earnings limit how much of any further tightness the company can capture, and the shares are acutely sensitive to any hint of new supply.

Some large investors have acted on those doubts. David Tepper sold his entire stake in the quarter the stock peaked, a fund using borrowed money to bet on AI was forced to sell, and one report flagged professional money leaving memory stocks. Others see the opposite: JPMorgan set a large price target, and Rosenblatt began coverage with a buy rating and a target of $2,400. Opinion is sharply divided, as one would expect at a point in a cycle where nobody can see the top.

What the price assumes

The share price can be translated into a growth rate. Suppose it stays flat for five years and, at the end, the shares trade at 18.5 times earnings, the middle value for every profitable listed company we cover. For that to work, earnings would need to shrink by 3% a year. In plain terms, today's price assumes Sandisk will be earning somewhat less in five years than it does now.

The company's own past points violently the other way. Over three years, operating profit has grown at an annual rate of 432%. Carried forward, that trend lands so far above what the price requires that the comparison stops meaning anything; it reflects how low the starting point was, not a plausible future.

The more useful counterweight is the analysts' forecast of 30% growth next year. Between that forecast and the price's assumption of gentle decline lies the entire debate. The market is not saying Sandisk is a bad business; it is saying this year's earnings are a peak, and the flood of profit has not shifted that view.

The other side of the medal

The risk we would keep in front of everything else is that today's margin rests almost entirely on rising chip prices. Sandisk has not become a different kind of company; it is the same flash maker selling the same kind of product for far more. A gross margin that has swung so widely over five years can fall as fast as it rose, and because prices feed almost straight into profit, a fall in chip prices would hit earnings harder than sales.

That matters more because, measured against its assets, sales and earnings, the shares are dearer than most of the companies we rate. The valuation looks modest next to its storage peers only while today's earnings hold; on a more ordinary margin it would not look modest at all. The first sign of the risk arriving would not be a loss but a pause: contract prices that stop rising and a gross margin that stops widening.

What to watch

The clearest signal is the price of the chips themselves. Industry trackers report NAND contract prices regularly, and those reports move before Sandisk's results do. As long as they keep showing increases from one quarter to the next, the 71% gross margin has room to widen; a stall would be the first hint that the boom is peaking. Watch too for Chinese flash makers winning business with big phone, PC or server buyers: a price stall alongside such wins would be a strong warning.

Two slower signs would argue for something more durable: management disclosing how much supply sits under multi-year contracts, and data-centre sales growing faster than the company as a whole. If price rises instead slow sharply or reverse and the gross margin slides back toward its middle value, the 30% earnings growth analysts expect would give way, and today's low valuation would turn out to be the price of a peak rather than value.

The numbers

Qualifications that apply to every figure above: gross margin has varied by 14% around a 35% median: not a stable-margin business.

The same industry, for comparison

Research for information, not personal investment advice or a recommendation to buy, sell or hold any security: see Important information. Data as at 31 October 2026.