Contrarian Corpus
short seller research note initial thesis
2025-09-01 · 35 pages

Pure Storage, Inc. PSTG

Pure Storage's hyperscaler dream is hype: HDDs dominate at 5-6x TCO advantage, the Meta deal is replicable, and Pure deserves a peer multiple, implying ~55% downside to $35.

N 4 Narrative
V 3 Visual
C 3 Craft
Source URL unavailable

Thesis

Kerrisdale is short Pure Storage, a $27bn enterprise storage vendor whose valuation reflects unwarranted excitement around an alleged hyperscaler breakthrough. HDDs shipped 1,052 exabytes into data centers in 2024 versus Pure's 4EB, and Kerrisdale's own TCO model shows flash carries a 5-6x cost penalty versus 30TB HDDs, undermining Pure's claim that QLC flash will displace bulk storage. The recently announced Meta deal is a narrow software license to a single AI data center, with Meta buying flash directly from suppliers and using ODMs for hardware; engineers and competitors describe Pure's capabilities as replicable within 18-24 months. Meanwhile, ultra-high-performance players (DDN, WEKA, VAST Data, Hammerspace) are encroaching on Pure's core competencies. Valuing Pure at 3x revenue or NetApp's 10x EBITDA implies a $35 target, ~55% downside.

SCQA

Situation

Pure Storage is a $27bn enterprise all-flash array vendor that rode the AFA adoption wave and now trades at an 80%+ EV/revenue premium to history on hopes of a hyperscaler breakthrough.

Complication

Hyperscalers ship over 1,000EB of HDDs annually because flash carries a 5-6x TCO penalty for bulk storage; Pure's much-touted Meta deal is a narrow, replicable software license, not durable hyperscaler revenue.

Resolution

Short PSTG: re-rate Pure to its true peer NetApp on revenue and EBITDA multiples rather than the SaaS or AI-infrastructure comps bulls invoke.

Reward

A blended 3.0x CY2026 revenue and 10x CY2026 EBITDA valuation yields a $35 price target, representing roughly 55% downside from the recent $77.89 share price.

The three reasons

  1. 1

    HDDs retain a 5-6x TCO advantage over flash at hyperscale, Pure's TCO math is misleading

  2. 2

    Meta deal is narrow, replicable, and unlikely to expand to Google, AWS, or Microsoft

  3. 3

    Pure trades at 95% premium to NetApp despite weaker margins and core market stagnation

Primary demands

  • Investors should short PSTG shares
  • Reject the narrative that QLC flash will displace HDDs at hyperscalers
  • Discount the Meta software-licensing deal as narrow, replicable, and non-recurring
  • Re-rate Pure to peer storage multiples (NetApp), implying ~55% downside to ~$35/share

KPIs cited

TCO disadvantage of QLC flash vs HDDs at hyperscale
5-6x cost premium per Kerrisdale's adjusted Solidigm-style model (4.9x implied)
2024 data center exabyte shipments
1,357EB total: 1,052EB HDDs, 235EB enterprise SSDs, 70EB tape; Pure shipped just 4EB
Pure cumulative exabytes FY2016-FY2025
~15EB total — less than annual hyperscaler tape shipments
EV/NTM revenue multiple
5.0x average since 2024 vs. 2.8x average 2016-2023; recent spike to 6.3x
Premium to NetApp
95% on CY2026 revenue (6.3x vs 3.2x); 171% on CY2026 EBITDA (26.9x vs 9.9x)
Pure//E gross margin guidance
Low-to-mid 40s — sacrificing margin to penetrate hyperscaler market
Depreciation accounting change benefit
Useful life extended 4→5-7 years, boosting FY2025 net income by $20.1m / $0.06 EPS
Interest income contribution to GAAP net income
$76m gross / $55m tax-effected = ~52% of FY2025 GAAP net income
Total revenue growth FY2025
12% vs 26-29% historically; product revenue down to 5%
Meta exabyte deployment guidance
1-2EB testing in 2025; low-double-digit 10-12EB production in late 2026/early 2027

Pattern membership

Where this document fits across the library's 12 rhetorical / structural patterns.

Precedents cited

  • Western Digital's $16bn acquisition of SanDisk (vertical flash integration rationale)
  • EMC's $430m acquisition of XtremIO (prior AFA category history)
  • VAST Data's reported $30bn financing with NVIDIA/Google (competitive benchmark)

Notable slides (6)

Notes

Classic Kerrisdale long-form short report (~33 pages of text + disclaimer). Heavy reliance on expert calls (former Pure employees, hyperscaler architects, VAST/DDN/WEKA executives) for evidentiary anchoring. Strong rhetorical move on page 11: reproduces Pure's own marketing slide ('Even if HDDs Were FREE, Pure's TCO is BETTER') to set up the contradiction. CEO quote contradiction on page 9 ('no more hard disk drives will be sold after 2028') and page 23 (Giancarlo conceding Pure is not in Meta's newest cluster). Visual style is institutional/blue-Times-Roman text-dump with embedded data tables and line charts — competent but not editorially crafted. Date inferred from cover ('September 2025') and filename; specific day not stated, defaulted to 2025-09-01.