Second order memory play likely to double in a year

Everpure (NYSE: P) The AI Infrastructure Re-Rating

Current price: ~$78 | My fair value estimate: ~$$150 | Timeframe: 12+ months

The market is pricing Everpure as an enterprise storage company. That is the wrong category. Once two or three analysts figure out what is actually happening inside this business, the multiple closes toward AI infrastructure peers, and the stock moves $70–$80 from here without a single change in fundamentals. Let me walk through why.

What Everpure Actually Does (And Why It Matters Now)

Everpure, formerly Pure Storage, rebranded and relisted as NYSE: P in April 2026 makes all-flash storage hardware and the software stack that runs on it. That description sounds boring. The engineering underneath is not.

Every competing storage system (NetApp, Dell EMC, HPE) uses commodity SSDs. Each SSD contains its own Flash Translation Layer, a firmware layer that makes NAND flash behave like a spinning hard disk. This compatibility tax is expensive: each drive embeds its own DRAM controller, loses roughly 20% of raw capacity to overprovisioning, and remains a black box to the array operating system above it.

Everpure's DirectFlash Modules eliminate the per-drive FTL entirely. Purity OS, their proprietary operating system, manages NAND directly at array scale, balancing writes, handling garbage collection, and optimizing data placement across every die simultaneously. The measurable outcome: DirectFlash drives fail at 0.15% annually versus 0.90% for commodity SSDs, six times better reliability, measured across their production fleet since 2017.

Competitors cannot replicate this by rewriting their software. They would need to redesign the flash module, qualify it with NAND foundries, and rebuild their OS from scratch. That is a 3–5 year program with no guaranteed outcome. Everpure has been compounding this advantage for over a decade.

The Financial Picture

FY2026 closed strong: $3.663 billion in revenue, up 16% year-over-year, with Q4 crossing $1 billion for the first time in company history. Non-GAAP gross margin held at 72.1%. Non-GAAP operating income came in at $635 million (17.3% margin) — a number that grew 47% in Q4 alone. Free cash flow for the year was $616 million.

The forward-looking numbers are more interesting than the trailing ones. Remaining Performance Obligations, contracted future revenue already on the books, grew 40% year-over-year to $3.67 billion. That is not a soft leading indicator; it is signed contracts. FY2027 guidance calls for $4.3–4.4 billion in revenue (+17–20%) with non-GAAP operating income growing 26%. Management has beaten guidance consistently.

Subscription ARR hit $1.92 billion, growing 16%. The company returned $343 million to shareholders via buybacks in FY2026, reducing the share count while holding $1.5 billion in cash. With ~330 million basic shares outstanding (~345 million diluted), the stock trades at roughly 6.7x EV/FY2026 revenue. That is cheap for a business growing at 16–20% with 72% gross margins and expanding operating leverage.

Why the Base Case Is Already Priced and Where the Real Bet Is

The consensus analyst target sits at $93–$95. At current prices around $78, that's 20% upside and represents a reasonable base case: Everpure executes FY2027 guidance, revenue lands at ~$4.35 billion, the EV/Revenue multiple compresses slightly to 6.0x, and the stock drifts toward $82–85. Fine, but not why you build a position.

The real bet is a multiple re-rating event, driven by three specific catalysts.

The hyperscale design win. Everpure secured its first EXA-scale hyperscale customer in FY2026. The company expects hyperscale revenue to ramp materially in FY2027. One additional hyperscale win at this scale represents $200–300 million in incremental product revenue in year one — more than the company's entire quarterly subscription revenue from two years ago. At that point, Everpure is no longer a story about enterprise storage share gains; it is a story about being a critical supplier to the largest infrastructure buildout in history. The multiple changes.

The market the analysts are using is too small. The most aggressive credible third-party forecast for the all-flash array market, from Data Insights Market Research, puts total AFA market size at $71.6 billion by 2027, growing at a 22.8% CAGR. Technavio projects $27.7 billion in incremental flash storage growth by 2029 alone at a 27.5% CAGR. Everpure currently holds roughly 10–12% of the AFA segment. Hold that share against a market growing at 22–27% and revenue lands at $4.8–5.2 billion for FY2027, above guidance, with operating leverage flowing through. Most models use IDC's more conservative 17% figure.

Pure KVA is a GPU cost story, not a storage story. This is the one the market has not priced at all, and I'll come back to it below.

Three Share Price Scenarios

Base case (~$82–85): Guidance execution, no hyperscale acceleration, multiple compresses modestly to 6.0x EV/Revenue on $4.35 billion in FY2027 revenue. Market cap ~$27.6 billion, ~$79–85 per share at ~347 million diluted shares. Essentially current fair value.

Bull case (~$107–112): Revenue lands at $5.0 billion (market growth at high end of analyst range plus a second hyperscale win). Non-GAAP operating margin expands to 22% as hyperscale revenue has near-zero incremental sales cost. Multiple holds at 7.0x EV/Revenue as the hyperscale narrative drives sentiment. Market cap ~$36.6 billion at ~340 million diluted shares. Non-GAAP EPS around $2.85–3.10, implying 35–39x P/E — consistent with high-growth infrastructure comps. This is ~40% upside from here.

Maximum bull (~$150–160): The market begins pricing Everpure as AI infrastructure rather than enterprise storage. A re-rating to 9.5x EV/Revenue requires only one or two additional hyperscale customers plus evidence that Pure KVA is generating platform-level revenue. Revenue at $5.2 billion, ~335 million diluted shares, enterprise value ~$49 billion. NVIDIA trades at 20x revenue. Arista trades at 12x. A storage company that demonstrably reduces GPU capex requirements gets re-rated somewhere in between. That re-rating is worth ~$75 per share from current levels.

What the Market Is Not Pricing

Here is where it gets interesting. The conventional bull case, hyperscale wins, AFA share gains, ARR expansion, is legitimate but well-understood. The following arguments are almost entirely absent from sell-side research.

Pure KVA turns storage budget into GPU budget. When an enterprise deploys Everpure's Key-Value Accelerator, FlashBlade persists KV cache states across LLM inference sessions. GPUs stop recomputing prefill for repeated prompts. Benchmarks show up to 20x faster inference with NFS, 6x with S3. Here is the business implication: a $3 million FlashBlade system that reduces GPU requirements by 15% at an enterprise running a $40 million GPU fleet pays for itself in months. That converts FlashBlade from an IT infrastructure purchase into an AI compute cost-reduction strategy. The buyer is no longer the storage team with a $15 million annual budget. The buyer is the AI infrastructure team with a $150 million budget. No model has repriced Everpure on this basis. When it does, the TAM recalculation is enormous.

DirectFlash is an unquantified energy asset. Data center power capacity is now a genuine binding constraint on AI infrastructure growth. The narrative focuses entirely on GPU power draw. Storage is invisible in this conversation, but it should not be. At petabyte scale, commodity SSD arrays consume substantially more power per usable terabyte than DirectFlash, because each COTS drive carries independent controller overhead that scales linearly with drive count. DirectFlash concentrates this overhead at the controller level, where it scales logarithmically. A hyperscaler constrained to 500MW can store more data per megawatt with DirectFlash than with any commodity alternative. Everpure has never quantified this publicly in watts-per-terabyte terms. The first analyst who does will find a competitive moat layer that does not appear in any current valuation.

The true Evergreen churn rate makes every DCF wrong. Evergreen customers cannot churn at normal rates. Switching requires a forklift migration, data re-ingestion, retraining of administrators, and renegotiation of every backup and replication policy — in short, all the disruption that Evergreen was specifically designed to eliminate. The real churn rate is almost certainly below 3%, possibly below 2%. No one has confirmed this publicly. But RPO growing 40% year-over-year while the customer base grows at a much slower rate implies extensions and expansions, not churn. If the market's DCF models assume 7–8% implied churn and the true number is 2%, the LTV per customer is 3–4x higher than consensus assumes. That reprices the stock without touching any growth rate assumption.

1touch.io is a Cloudflare moment, not a feature. The acquisition is universally analyzed as a data discovery add-on. The actual opportunity is larger: Everpure will soon know more about the shape, sensitivity, location, and regulatory exposure of enterprise data than the enterprises themselves, because it sits on the storage layer where all of it lives. Cloudflare made a similar realization when it understood it was sitting on the internet's routing fabric, it built a security intelligence business on that substrate. Everpure, with 1touch.io integrated, can sell data governance, compliance automation, and AI-readiness scoring as a SaaS layer on top of storage it already manages. The customers are locked in. The data flows through Everpure's systems already. The marginal cost is minimal. Analysts are pricing this as a modest product enhancement. It is a potential $1 billion ARR business hiding in plain sight.

The backup market is about to consolidate into primary storage. Enterprises today run two storage stacks: primary (fast, live data) and secondary (slower, backup and DR). Rubrik, Commvault, Cohesity, and Veeam exist because those stacks are historically separate. Everpure's SafeMode snapshots, synchronous ActiveCluster replication, and Evergreen//One's zero-RPO capabilities already compete with purpose-built backup appliances for many workloads. As QLC NAND continues falling toward HDD cost parity per terabyte, the performance and economic gap between the two tiers closes. When enterprises realize a single Everpure platform can handle both functions, the secondary storage budget migrates to primary. Rubrik's ~$5 billion market cap and Cohesity's ~$3 billion valuation represent storage spend that eventually gravitates here. This consolidation does not appear in any consensus model.

The Honest Risks

The hyperscale bet is the highest-leverage variable and the least certain. If the first EXA customer does not become a reference account that closes a second win, the bull case compresses to the guidance range.

VAST Data, Weka, and DDN compete directly on FlashBlade's turf for AI storage workloads. Their architectures are newer and their sales motion is more focused. They are small today but growing fast.

NAND commoditization is a long-run headwind. As QLC costs fall, the raw cost advantage of DirectFlash over commodity SSDs compresses. The software moat has to carry more weight over time.

Insider selling has been notable: co-founder John Colgrove has sold shares in Q1 and Q2 2026. The amounts are small relative to his holdings, but worth watching.

My Assumptions

I believe the all-flash array market grows at 20–23% annually through FY2027, consistent with the Data Insights and Technavio high-end scenarios. I believe Everpure holds or slightly grows its market share. I believe the hyperscale customer adds at least one more significant logo in FY2027. At those assumptions, FY2027 revenue lands around $4.8–5.0 billion. I apply a 7.0–8.0x EV/Revenue multiple reflecting the transition narrative, not yet AI infrastructure multiples, but a step toward them. That produces a fair value estimate of $112–$130.

If the re-rating happens, if KVA attach rates start appearing as a separate disclosure line, if a second hyperscale logo is announced, if 1touch.io starts contributing SaaS revenue, the 9.5x scenario and a $150 fair value becomes a reasonable 24-month target rather than a tail outcome.

At $78, the market is betting this stays a storage company. I am betting it becomes something else.

Positions disclosed: long NYSE:P. This is not financial advice, it is my personal investment thesis. All numbers sourced from Everpure FY2026 8-K, DEF 14A filings, IDC, Technavio, Data Insights Market Research, Emergen Research, and public analyst consensus data as of May 2026.