Western Digital (WDC) Q4: revenue +44% and margins soar, HAMR gap is the story

$WDC printed $3.75B revenue (+44% YoY) and $3.56 non-GAAP EPS on the fiscal fourth quarter, beating consensus on both lines, with cloud revenue at $3.3B and 89% of the mix. The DRAM / HBM Memory bubble's HDD leg carried through the print intact on demand and margin; the tape still marked the stock down roughly 11% after hours on the HAMR execution gap the QA thesis had flagged as a known bear leg. This recap walks the segment numbers, the read against $STX, and the Q1 FY27 guide.

The print

$WDC reported its fiscal fourth quarter (ended July 3, 2026) after the close on August 5. Every line beat consensus.

| Metric | Q4 FY26 actual | Consensus | YoY | |---|---|---|---| | Revenue | $3.75B | $3.71B | +44% | | Non-GAAP EPS | $3.56 | $3.27 | +122% (more than doubled) | | GAAP EPS | $8.21 | n/a | n/a | | Non-GAAP gross margin | 54.4% | ~53% | +970 bps FY | | GAAP gross margin | 54.1% | n/a | n/a |

The GAAP-to-non-GAAP EPS gap is unusually wide this quarter, tied to the SanDisk spinoff completion and related one-time items; the $3.56 non-GAAP number is the consensus comparison surface.

End-market mix is where the AI-storage thesis actually lives:

• Cloud: $3.3B, +43% YoY, 89% of total revenue. This is the hyperscaler + enterprise-storage book pulling every ounce of the print.
• Client: $225M, ~6% of the mix.
• Consumer: $187M, ~5% of the mix.

Full fiscal year 2026:

• Revenue: $12.9B, +36% YoY.
• Non-GAAP diluted EPS: $10.22 (more than doubled).
• Gross margin: expanded 970 basis points FY to 49.1%.
• Free cash flow: $3.5B.
• Capital returned: $3.1B to shareholders, including $1B of share repurchases in Q4 alone.

Guidance for Q1 FY27, verbatim from the release:

| Target | Q1 FY27 guide | |---|---| | Revenue | $4.1B ± $100M | | Non-GAAP gross margin | 55% to 56% | | Non-GAAP diluted EPS | $4.00 ± $0.15 |

At the midpoint, that guide is another sequential step up on both revenue and margin. CEO Irving Tan framed the setup on the call: "The inflection we have discussed in the past, from AI training to inference to agentic AI, has only become more pronounced," and separately "while compute cycles can be reused, data compounds." That is the memory-bubble thesis in one line.

DRAM / HBM Memory read-through

The QA thesis on /stocks/wdc frames Western Digital as leg two of the HDD duopoly with $STX, both riding the AI cold-storage layer: exabytes of warm + cold storage backing every hyperscaler training program, with NVMe handling the hot tier and HDD handling everything else by economics. The demand read from this print sits inside that thesis cleanly. Cloud revenue is 89% of the business and grew 43%. Full-year gross margin expanded almost 1,000 basis points. The duopoly is pricing rationally. A cyclical HDD business that used to trade at low-single-digit gross margin in downturns is now printing 54% at the top of an AI-driven cycle. The DRAM / HBM Memory bubble read-through is that the storage layer is participating in the same allocation-tightness setup the DRAM oligopoly is on, one tier down the stack.

Where the print does update the thesis is on the execution leg. The QA writeup explicitly notes "WDC's HAMR roadmap is 1-2 quarters behind STX's Mozaic 3+ but ramping." The tape's -11% reaction after this print says that leg went from a footnoted timing risk to a hard competitive gap the market is now marking. Seagate is shipping 44TB HAMR drives at hyperscale today. Western Digital's HAMR is still in the qualification phase. Management pushed back with roadmap language on next-generation ePMR and HAMR products, but the print did not include a shipping HAMR SKU, and after a stock that ran roughly six-fold over the trailing twelve months, "still qualifying" is not what the tape needed to hear.

Net read: the memory thesis on $WDC is intact on demand and unit economics, and it just took a real hit on relative-execution timing. Neither of those is fatal; both are observable.

Peer reaction

$WDC's closest correlated names in the QA universe are $STX (0.88 correlation, the direct HDD duopoly comp), $SNDK (0.72, the recently-spun-off NAND business), $MU (0.71, the DRAM/HBM anchor of the same bubble), and $LRCX (0.67, the semi-cap play on the same capex cycle).

Because the -11% mark on WDC is specifically a HAMR-gap trade, the peer read splits by axis:

• STX is the direct beneficiary of the same trade. If the market rewards the HAMR incumbent versus the qualifier, Seagate's relative bid should compound through the next earnings cycle. The bear version is that WDC's guide-down on relative competitive positioning eventually resets HDD-duopoly pricing lower for both.
• SNDK is now a pure-play NAND read, uncorrelated to the HAMR question. Its Q4 FY26 print, along with the DRAM / HBM Memory supercycle framework, is the surface to watch there.
• MU and the DRAM leg of the bubble should be unaffected by the HAMR-specific reaction, but any read that "the tape is now punishing memory execution stumbles at cycle highs" is a signal for the DRAM names too, given eight of ten late-cycle memory signals were already flashing coming into August.

What worked

• Cloud revenue at 89% of the mix and +43% YoY is the number the AI-storage thesis is built on. It grew this quarter.
• Full-year gross margin expanded 970 basis points to 49.1%. Duopoly pricing plus mix-shift to high-capacity drives is showing up in the P&L exactly as the thesis argued.
• $3.5B in free cash flow on $12.9B revenue is a 27% FCF margin on a business the buy side used to model as low-single-digit through the cycle. That reset is real.
• $3.1B capital return in FY26, with $1B of buybacks in Q4 alone, at a share price roughly 6x higher than a year ago. Management is choosing to buy back at the top of the cycle, which is either conviction or a mistake; either way it is a signal.
• Guidance stepped up sequentially on revenue and margin. $4.1B midpoint versus $3.75B is another 9% sequential quarter, and 55-56% gross margin is another leg up.

What broke

• HAMR shipping gap. Seagate is shipping 44TB HAMR drives at hyperscale today; WDC's HAMR is still in qualification. This is the QA thesis's known bear leg, now activated. Until WDC has a shipping HAMR SKU in a hyperscaler qualification pass, every quarter is a chance for STX to widen the density lead.
• Guide barely cleared consensus after a 6x run. A revenue guide range that clears consensus by $60M-$160M is a good guide in most tapes. Not this one. After a ~200% YTD run, the tape needed a blowout, and $4.1B ±$100M did not deliver a blowout.
• Cyclicality is unresolved, not repealed. Storage is still storage. 27% FCF margins at the top of an AI-capex cycle are cycle-peak margins by definition; the question is how far back they compress in the next hyperscaler capex pause. This quarter's print does not settle that.
• 89% cloud concentration is a feature and a risk. Hyperscaler concentration means one large customer program pull-in or push-out swings a quarter's revenue by more than the guide range. That was tolerable when the multiple was 4x; at the current multiple, it isn't.

What to watch

• Next earnings. Q1 FY27 print, likely late October or early November 2026 based on WDC's cadence. The setup: $4.1B ±$100M revenue at 55-56% gross margin and $4.00 ±$0.15 EPS.
• HAMR qualification milestones. Any customer-named HAMR qualification pass or first-shipment announcement between now and Q1. This is the specific data the tape is asking for and the direct fix for the HAMR-gap trade.
• STX vs WDC relative performance. If $STX continues to shed less than WDC on any incremental HDD-negative headline, that is the tape confirming the competitive-positioning re-rate. If they trade one-to-one, the market is treating the HAMR gap as noise.
• Hyperscaler capex commentary. The next four to six weeks of hyperscaler earnings will re-anchor the cold-storage demand curve WDC is levered to. A single hyperscaler pulling back is a real read on the FY27 guide.
• Cross-check versus the memory-bubble late-cycle framework. The DRAM / HBM Memory writeup flags eight of ten late-cycle signals already flashing (weekly RSI 86, +1,873% off the 2023 low, forward PE 8.76x). WDC's post-print action is a live test of whether the tape starts pricing in cycle exit for the whole bubble or just the execution laggard.

Full framework on where the memory cycle sits today: Memory cyclicality: the supercycle that still has a cycle. The QA thesis for $WDC and the STX comp live on /stocks/wdc.

Educational research on structural market dynamics. Not investment advice. QuantAbundancia positions may change without notice.