SanDisk (SNDK) $4,000 Price Target: The Case Rests on the Multiple, Not the Earnings
The arithmetic behind a $4,000 SanDisk share price is unremarkable. Take consensus fiscal 2027 earnings of roughly $168 per share, apply a price-to-earnings multiple of 22, which is about where the S&P 500 sits, and the answer falls out. Evercore’s bull case gets there a different way, assuming better than $300 in earnings power against a headline target of $3,100. Neither calculation requires SanDisk to beat anything. Both require only that the market agree to value NAND flash earnings the way it values the average large-cap American company.
It refuses. At roughly $1,350 and a market capitalisation near $211 billion, SanDisk trades at a high single-digit multiple of what the sell side believes it will earn in the fiscal year that just started. That gap is the entire investment debate. Demand is not in dispute. Pricing is not in dispute. What is in dispute is duration, and no quarterly result can settle it.
Fiscal fourth quarter revenue came in at $8.97 billion, up 371.6 percent year over year and 51 percent sequentially, against management guidance of $7.75 billion to $8.25 billion. Adjusted earnings were $39.25 per share against a $30 to $33 guide and a consensus band in the low thirties. Net income was roughly $6.9 billion. Free cash flow was $7.08 billion in the quarter, $5.04 billion on an adjusted basis after payments and deposits tied to the new long-term supply agreements, and $8.74 billion for the full year against $238 million a year earlier. The September quarter is guided to $10.3 billion to $10.8 billion.
The shares fell. That reaction is the most informative data point of the week. A company that grew revenue by a factor of nearly five, beat its own guidance by six dollars a share, and guided the next quarter up another fifteen to twenty percent sequentially was marked down, because none of those numbers speak to fiscal 2029. The market is not pricing SanDisk as a company with $168 of earnings power. It is pricing a company that will earn something like that for a while and then revert, and it is discounting the interval accordingly.
Which is why Elon Musk moved memory stocks more than SanDisk’s own print did. On the SpaceX post-earnings call this week, Musk identified memory as the binding constraint on AI infrastructure, put supply growth at roughly 20 percent a year against demand growth of 200 percent or higher, and drew the obvious conclusion about price direction. He also thanked Samsung, Micron and TSMC by name, which listed-company chief executives do not ordinarily do for suppliers.
The significance is not the figures, which are estimates offered conversationally. It is the direction the testimony runs. Every memory supplier has an interest in describing its market as structurally tight, and the market discounts supplier guidance accordingly. A buyer describing the same tightness, on his own earnings call, while explaining why his compute build is constrained, is evidence of a different quality. The bullish memory case has spent two years resting on the word of people who sell memory. It now rests partly on the complaints of the people who buy it, and that is a materially harder thing to dismiss as talking the book.
There is a transmission problem, and anyone underwriting $4,000 on the strength of that call should state it. Musk was describing the constraint on AI compute, which is principally DRAM and high-bandwidth memory. SanDisk makes neither. Its participation in the shortage is derived rather than direct: wafer capacity diverted toward HBM tightens everything else, inference workloads push key-value cache and model weights into storage tiers, and QLC enterprise SSDs absorb capacity that DRAM economics no longer justify. Those channels are real and they are visible in the datacenter revenue line. They are also one step removed, and a step removed is where cyclical arguments usually break.
The competitive question is where this gets interesting, because SanDisk’s advantage is no longer primarily technical. BiCS8 is competitive, not singular. Samsung, SK Hynix through Solidigm, Kioxia as joint venture partner, and YMTC all produce NAND, and the industry’s history is one of discipline collapsing the moment anyone believes a shortage is permanent. What SanDisk has built instead is contractual. The new business model agreements reportedly represent about $62 billion in minimum committed revenue backed by better than $11 billion in guarantees and prepayments, with floor pricing in the region of $0.29 per gigabyte. That structure converts a commodity producer into something closer to a contracted utility, and a contracted utility is exactly the kind of business that earns a market multiple.
The same structure caps the upside. Floors set near current average selling prices mean the contracted book does not participate fully in a further spike, and the more bits move under contract, the more fiscal 2027 upside is traded away for fiscal 2029 visibility. Capital intensity sits largely inside the Kioxia joint venture rather than on SanDisk’s own balance sheet, which is why cash conversion looks the way it does, and which also means capacity decisions are shared rather than unilateral.
Management has taken a side. The board authorised an additional $14 billion of repurchase capacity, bringing the total remaining authorisation to $15.5 billion, roughly 6.6 percent of shares outstanding. At a high single-digit multiple of forward earnings, retiring stock is arithmetically the highest-return use of capital available to the company, and the size of the authorisation is a statement that management believes the earnings are more durable than the tape does. It is the same bet Evercore is making, expressed in cash rather than in a target price.
Base case tracks the $2,200 consensus, with fiscal 2027 earnings landing somewhere between $170 and $226 and the multiple staying compressed in the ten to thirteen range. Bull case is the $3,000 to $4,000 band, and it requires no operational surprise at all, only multiple normalisation toward the high teens, which in turn requires the market to accept that post-2028 floors are real. Bear case is a cohort derating rather than a company-specific failure: AI capital expenditure growth decelerates, every memory name compresses together, contract floors become ceilings on the way down, and fiscal 2028 estimates fall while the multiple stays where it is. The low end of the published target range is already $1,000.
The Investor Day on August 13 has one number that matters more than the rest of the presentation. What share of fiscal 2027 bits is committed under long-term agreement? Above half, and the durability argument acquires a figure the market can model against. Below it, $4,000 remains arithmetic.