Seagate Technology

FQ4 FY2026 turned the AI-nearline thesis from a demand story into a margin-and-cash story: US$3,629M of revenue (+48% YoY) above the high end of guidance, a record 52.7% non-GAAP gross margin in a 13th straight quarter of expansion, a 44.6% non-GAAP operating margin, US$1.1B of free cash flow in a single quarter, and a September-quarter guide of US$4.1B at a ~50% operating margin — underwritten by long-term agreements that management says already allocate the vast majority of nearline exabytes into calendar 2028.

• Growth accelerated rather than decelerated: FQ4 revenue of US$3,629M beat the top of the US$3.45B ±US$100M guide, and FQ1 FY2027 is guided to US$4.1B (~+56% YoY). A year ago the base case was a fade from 44% toward a ~20% floor; the company is running well above that
• Non-GAAP gross margin reached 52.7% (+1,480bps YoY), a 13th consecutive quarter of expansion, with non-GAAP operating margin at 44.6% and guided to ~50% next quarter on ~US$300M of opex — HAMR mix, pricing and operating leverage are compounding together, not trading off
• FY2026 free cash flow of US$3,105M (25.5% of revenue) with FQ4 alone at US$1,118M (30.8%), on capex of just 4.7% of revenue. This is the highest-quality part of the print: cash, not accrual margin
• The balance sheet was rebuilt in a year: net debt ~US$1.9B from ~US$4.2B, net leverage 0.4x from 1.8x, with US$1B of high-yield notes already retired in July and ~US$2.4B of gross debt targeted by the end of FQ1 FY2027
• Demand visibility is contractual rather than forecast — the vast majority of nearline exabytes are allocated into calendar 2028 under long-term agreements, with customers asking for 2029+ planning. That is a materially stronger position than a multi-quarter backlog
• The technology transition is on schedule and quantified: HAMR is at 40% of nearline exabytes with ~70% targeted by June 2027, Mozaic 4+ ships at up to 44TB and Mozaic 5+ targets 50+TB — the density lead that produces the cost-per-TB advantage is converting on a stated timetable
• Volume grew alongside price: 218EB shipped (+34% YoY) and 195EB nearline (+43%), so roughly two-thirds of the revenue growth is units, not pricing — the margin is not purely a pricing artifact

A ~US$194B market cap on a hardware manufacturer guiding a ~50%-of-revenue operating margin and taking ~20% more per exabyte year over year — extraordinary terms that, on this industry's own record, are the part of the cycle that mean-reverts. Concentration deepened rather than eased (~81% of revenue and ~90% of exabytes go to data center), capex stays deliberately capped at 4–6%, and the record cash is going to lenders rather than shareholders.

• Pricing is doing visible work: price per exabyte rose ~10% YoY in FQ4 and is implied ~20% YoY in the September quarter. HDD pricing power has historically been lent by scarcity, not owned — a single digestion quarter takes margin back faster than three years of mix built it
• A guided ~50%-of-revenue operating margin is not a defensible steady state for a component manufacturer. FY2026's non-GAAP operating margin was 36.5%, so the guided exit rate sits ~14pts above the year just completed — the gap is the cycle, and it closes in both directions
• Concentration got worse, not better: data-center revenue is ~US$2.9B, ~81% of total (+57% YoY), and ~90% of exabytes ship into data-center markets. A hyperscaler capex pause now hits substantially the whole model, with legacy edge/consumer far too small to cushion it
• Capex is deliberately held at 4–6% of revenue, so Seagate is monetising scarcity rather than building into it. That protects free cash flow and the duopoly's pricing, but it caps unit upside — and the same fixed-cost base is what makes the downswing violent if the 2028 allocations are renegotiated rather than honoured
• Valuation is a different objection than it was, not an absent one: at the Jul 31 close the shares trade at ~62x FY2026 GAAP EPS (US$13.90) and ~55x FY2026 non-GAAP EPS (US$15.58); the ~29x on the September guide annualised (US$7.30 × 4) exists only if that guided run-rate is durable, which is precisely the contested question
• The tape is violent — a US$144.75–US$1,145.00 52-week range, and the shares closed Jul 31 ~25% below the high — so positioning is crowded and sentiment carries no information about the business
• Record cash is not reaching shareholders: the dividend was held at US$0.74/share despite US$3.1B of FY2026 free cash flow, and only 0.5M shares were repurchased all year. Capital return is subordinated to debt paydown — sensible, but it means the shareholder case rests entirely on the multiple
• The two structural threats are untested by this quarter: QLC NAND $/TB continues to fall against the low end of nearline, and HAMR execution at 50TB+ nodes is still ahead — the 70%-by-June-2027 penetration figure is an execution commitment that can slip, and the margin story is directly levered to it