ARM vs SanDisk: The Forward Numbers Flip the Cheap-and-Expensive Read

Two overbought semiconductor names, one underlying question: how long does the earnings run last? Arm Holdings (about $265) and SanDisk (about $1,636) have both ridden the AI trade, and the same-day market data from Sept. 11, 2026 holds a clean valuation contrast at near-identical market caps. It also holds a trap for anyone who stops at the obvious screen: on trailing earnings, SanDisk looks like the bargain at a fraction of Arm's multiples while Arm looks like the expensive one. The forward numbers run the other way, and the divergence is where the decision actually sits.

The trailing numbers write an obvious headline

Ainvest data on the same basis, the past twelve months, shows two companies that look like different asset classes. Arm is the extreme compounder: 271.8x trailing earnings, 262.5x EV/EBITDA, 55x sales. SanDisk is the value table in the same row: 20.9x trailing earnings, 18.7x EV/EBITDA, 11.8x sales, for only slightly less total market value.

TTM valuation multiples — ARM vs SNDKTrailing basis, Sep 11, 2026

On trailing earnings, ARM carries an extreme P/E near 272x versus SanDisk at about 21x, with the same extreme-to-cheap split across EV/EBITDA and P/S.

Ticker	P/E (TTM) (x)	EV/EBITDA (TTM) (x)	P/S (TTM) (x)
ARM	271.8	262.5	55
SNDK	20.9	18.7	11.8

A reader handed just that table would tell you SanDisk is the buy — a memory maker that in a year went from a Western Digital spin-off most people ignored to a stock up nearly 590% on the year, priced at a fraction of Arm's trailing multiples with next to the same market cap. That is the surface read, and it is the wrong one.

The forward picture reverses the comparison

Trailing and forward multiples sit on different earnings denominators, and it is the forward figure — the profit the street projects for the periods ahead — that says what the price is paying for. Push both stocks to forward earnings and the two detach in the opposite direction. Arm's forward P/E sits even higher than its trailing multiple — about 545x against 271.8x, both Ainvest data. A forward multiple above the trailing one means the price is not assuming near-term GAAP earnings growth that would bring the ratio down; it means no quick earnings relief is priced in at all. The premium is being carried by a story, not by a forecast of immediate profit. SanDisk's forward P/E is not high — it is negative, because the provider consensus projects a near-term loss as the NAND cycle turns. A negative denominator has no meaning as a multiple, so the honest statement is directional: Arm's forward premium stays above its already-extraordinary trailing number, while SanDisk's forward figure turns negative. Both stocks, in opposite ways, are therefore betting on a change in earnings rather than on the number they are earning today. The whole cheap-versus-expensive framing collapses into one durability question.

The durability question runs through a pricing cycle

SanDisk's current profit is a shortage artifact, and the entire contest turns on how much of it survives. NAND flash stayed undersupplied through 2026 as hyperscaler AI spending pushed server-led enterprise solid-state drives toward more than half of all NAND shipments, lifting prices and SanDisk's margins to a peak.

That scarcity built SanDisk a $42 billion contract backlog — five multi-year agreements with financial guarantees. But the backlog covers only about one-third of expected bit production in fiscal 2027; the other two-thirds rides open-market pricing that the cycle will eventually crush as new capacity arrives. It is the cleanest possible map of the bull-bear gap. The bull, looking at the backlog and the multi-year contracts with price floors, sees earnings becoming less cyclical. The bear — here personified by Morningstar, which rates SanDisk as having no economic moat — sees a commodity at peak margins headed for a 2029 downturn as capacity comes online. That same stock has a Street consensus across 23 analysts of $2,188 against Morningstar's $1,000 fair value estimate — a more-than-twofold disagreement over how durable the peak is, with the current price sitting between the two. When the two camps studying the same company cannot agree whether it is worth $1,000 or $2,188, the difference is not a valuation quirk. It is the whole thesis: how much of peak-cycle earnings survives reversion. Now put the same durability lens on Arm. Its premium rests on a compounder story with real cash under it — data-center royalty revenue more than doubled as hyperscalers adopted its Neoverse chips, and management holds roughly $2 billion in committed orders it can fill only about half because foundry capacity is diverted to AI accelerators. That is growth that is real but supply-gated, and it is being paid for in stock rather than cash: Arm's GAAP operating margin runs far below its non-GAAP number on heavy stock-based compensation. The bull can credibly call Arm a durable compounder; what the bull cannot do is point to a forward P/E that promises the premium will be earned anytime soon.

The ruling

Stop at trailing multiples and SanDisk is the obvious value and Arm the obvious excess. The forward data says neither reading is what the market is paying for. Arm's price demands continued compounding with no near-term GAAP earnings relief; SanDisk's price already writes off a near-term loss. So the stocks are not cheap-versus-expensive — they are two bets on durability, and only one has its central risk sitting inside a pricing cycle the providers already model as reverting. The sharper edge belongs to Arm on the quality of its earnings base: a royalty-and-licensing model that compounds regardless of memory pricing, versus a margin that exists only while NAND stays scarce. But "better business" is not "better stock call" at Arm's 55x sales. Arm's tripwire is forward GAAP EPS — if consensus revisions push its already-lofty forward P/E still higher, the compounder premium needs a lower price. SanDisk's tripwire is the forward NAND price guidance in each quarterly report: if pricing holds, the cheap trailing multiple is a beaten-down trough on the cusp of the next cycle; if guidance softens into the second half, the "cheap" stock is confirmed as a cyclical-peak trap. The cheapest stock in the table is the one most exposed to a number the market has already decided is going to roll over — and that is the entire point of checking forward earnings before trusting the trailing story.