Western Digital Beat Its Own Guidance and the Shares Fell Anyway: AI Storage Has Switched From Demand to Price

Western Digital cleared its own revenue guidance and the consensus on every headline line. Revenue of $4,381M was above the company's $4.0B–$4.2B guided range and the roughly $4.14B consensus; non-GAAP gross margin reached 58.1%; non-GAAP diluted EPS of $4.64 beat the $4.08 consensus by 14%. The board raised the quarterly dividend 17% to $0.175 a share, and the company repurchased $1,001M of stock during the quarter.

Then the tape rejected it. The shares had already fallen about 7% the day before the print and closed down 6.43% at $354.19 on Oct 22, caught in a broad repricing of AI-linked stocks; after the release they jumped roughly 4% and then reversed, and were down a further 5.4% at $335.01 in late after-hours trading. The market switched the subject from demand to price, and the earnings call is where the mood changed — from the revenue beat to the cost curve and the capital spending plan.

The beat is real; its composition is what the market argued with. Applying the company's own two disclosures — 22% more exabytes and 28% more price per terabyte — to the prior-year base of $2,818M, about $620M of the $1,563M revenue increase is volume and about $942M is price and mix. The market has to underwrite three-fifths of the growth as price, and price is the line that is hard to extrapolate.

Guidance implies another leg up. The $5.1B December-quarter guide is 69% above the year-ago quarter, and management told investors the September quarter was the floor for exabyte growth, accelerating through fiscal 2027 and 2028 on long-term nearline exabyte growth above 25%. If exabytes compound in the mid-20s and the price line normalises, the revenue line converges quickly — which is precisely the arithmetic the market is refusing to underwrite.

Two forces pushed price up faster than cost came down. Supply is allocated years in advance — Amazon, Microsoft, Alphabet and Meta are among the buyers — and some of that capacity now sells under multi-year agreements carrying scheduled price increases at contract anniversaries. The remainder sells at spot prices that management attributed directly to tight supply, which is the least durable part of the book.

Mix did the rest. Shipping a 40TB ePMR drive instead of a 30TB drive raises revenue per terabyte without adding a single drive unit, and areal density spreads fixed factory cost across more terabytes. That is why cost per terabyte still fell 5% even while the DRAM and NAND inside every drive got more expensive. Memory prices have turned into a tax on the disk business rather than a tailwind, and management capped near-term cost-per-terabyte improvement below its long-run 10% target for exactly that reason.

Western Digital fell before it reported, and it fell with everything. On Oct 22, Meta beat on revenue and then raised the low end of its 2026 capital spending plan by $5B to $135B–$145B while disclosing $825B of future spending commitments, pushing free cash flow negative and its five-year credit default swap to a record high. Amazon fell 4.6%, Alphabet 3.2%, Microsoft 2.2%, the Philadelphia semiconductor index 6.7%, Sandisk and SK hynix more than 8%, Seagate more than 7%, and Micron and Western Digital more than 6%.

Once the numbers landed, the objection narrowed to storage itself. Goldman Sachs kept a Neutral rating and cut its 12-month target to $540 from $615, citing compressed multiples across hard-disk peers and the risk that a competitor expands capacity, while preferring Seagate's smoother transition to heat-assisted recording. Nobody was forecasting fewer drives being bought. The question was how much of a 28% price per terabyte survives the next contract cycle, and that is a harder thing to underwrite than a demand forecast.

The hard-disk business is a components business with an assembly step at the end, and the contracted economics sit in the components. Western Digital's own fiscal 2026 Form 10-K notes that some key components come from single-source vendors with no current alternative, which makes the Japanese supplier layer the earliest observable signal of how long pricing can hold. Component suppliers have already reset price, while the drive maker is still selling at yesterday's price.

Eighty-eight percent of revenue — $3.87B of $4.38B — came from cloud, up 54% year over year, and the buyer list is short. Three customers accounted for 16%, 15% and 13% of fiscal 2026 net revenue, and the top ten rose to 73% from 68% in fiscal 2025 and 55% in fiscal 2024. A storage forecast has become a financing forecast, and that is precisely what the market repriced on Oct 22.

The four largest buyers are spending more, not less: Amazon raised 2026 cash capital spending to $220B from $200B, Alphabet plans $175B–$185B against $91.45B in 2025, Microsoft spent about $41B in its June quarter and guided to more than $50B for the September quarter, and Meta lifted the low end of 2026 spending to $135B–$145B. On those plans the four together are committing roughly $700B or more to capital expenditure in 2026. Demand for nearline capacity is not the problem.

Western Digital ended the quarter with $1,314M of cash against $789M of debt, a net cash position of $525M, and $4.0B of free cash flow over the trailing twelve months. It repurchased $1.0B of stock in the quarter and $3.0B over the trailing year, and raised the dividend 17% to $0.175 a share. The raise is a signal rather than an income story — $0.70 annualised is a 0.2% yield at the Oct 22 close — but the cash generation is real, and the balance sheet can fund the technology transition without the market's help.

One number in the reported results needs stripping out. Mark-to-market gains on the retained interest in Sandisk added $611M of non-cash income in the September 2025 quarter, $2.1B in the March 2026 quarter and $1.7B in the June 2026 quarter — roughly $4.4B across fiscal 2026, which is why fiscal 2026 GAAP diluted EPS of $24.28 sits far above operating earnings. Western Digital is partly long flash equity, and on an operating basis fiscal 2026 revenue of $12,919M carried a 48.9% GAAP gross margin.

The next four weeks answer the near-term question: is this pricing environment industry-wide or a Western Digital artefact? Seagate reports fiscal Q1 2027 on Oct 27, guided to $4.1B of revenue and $7.30 of non-GAAP EPS. A second beat with margins intact would confirm that the $5.1B guide and the 60%–61% gross margin range for the December quarter describe the industry rather than one company.

Western Digital's September quarter is the best available evidence that the AI build-out has moved from demand to physical constraint to margin: 248 exabytes, a 58.1% non-GAAP gross margin, $1.38B of free cash flow, and a December guide of $5.1B at a 60.5% gross margin midpoint. It is equally the clearest evidence that the market has changed what it pays for. A demand story gets a multiple; a price story earns scrutiny instead of a re-rating, and a $0.175 dividend on a $354.19 share price was never going to settle the argument.