Phase Space AI

Valuation

SentinelOne [S]

SentinelOne [S] — Valuation

As of 2026-07-29. Spot $18.53. Framework v1.5.1.

---

1. COMPANY STATE — declared first

STATE C — scaling but economically observable.

Not State A — deeply unprofitable on GAAP. Not State B — subscription revenue, no exogenous cycle. Not State D — and this is worth stating explicitly against the SHAZ precedent. S has $1,049.0m of revenue at a +74.0% gross margin, growing 20.8%, with zero debt, $812.5m of net cash, positive free cash flow, and a reverse DCF that solves cleanly across the entire parameter range. Nothing about S is economically unidentified.

Consequence: terminal margin must be BUILT; instrument is the two-dimensional surface; all three surface outputs required.

evidence_grade: C+. Lower than ZS's B, for four specific reasons: 1. NRR is retired and not reconstructible — the single most important operating metric for a land-and-expand business has no public number since 129% in FQ1 FY23. 2. RPO is disclosed rounded to $100m, making every RPO growth rate a ±4.3pp band. 3. Share count is unresolved — S is dual-class and all three EDGAR share-count concepts 404. Per-share figures below are on a basic 337.0m basis and will overstate value on a diluted one, by an amount not determinable from available data. 4. AV transcript coverage is 4 of 18 quarters, and the most recent available call is FQ2 FY26 — a year old. The FQ1 FY27 call (quarter ended 2026-04-30) is not in AV.

---

2. Terminal margin — BUILT, with the bridge

Base case: m_EBIT,T = 5.5% GAAP, at FY2032 (fiscal year ending January).

m_gross,T           76.0%      (GAAP TTM 74.0%, up from 68.1% three years ago; non-GAAP 79%
                                guided 78.5-79.0%. +2.0pp on continued scale in hosting and
                                support, capped below the non-GAAP figure because the ~4.5pp
                                GAAP/non-GAAP gap is SBC-in-COGS and intangible amortisation,
                                both of which are real)
  − R&D             26.0%      (TTM 33.2%; the clearest leverage line — S's absolute R&D is
                                sub-scale against CRWD's, so this cannot fall as far as a
                                mature comparator's)
  − S&M             34.0%      (TTM ~44%. The binding constraint: S competes for the same
                                deals as CRWD and MSFT and cannot cut go-to-market spend
                                without ceding the growth the case depends on)
  − G&A             10.5%      (from ~13%)
  − other            0.0%
  ─────────────────────────
  = m_EBIT,T         5.5%

Hard constraint: 5.5% ≤ 76.0% ✓. The bridge reconciles by construction.

The SBC reconciliation — and why the number is 5.5% and not 15%

S's SBC is 28.96% of revenue. Getting to a positive GAAP margin requires it roughly to halve. The base case has SBC dollars growing ~40% while revenue triples — arithmetically achievable, and S's record supports the direction: SBC has fallen 41.6% → 28.96% of revenue over three years, a 12.6pp decline. S has actually demonstrated SBC leverage; ZS has not (−0.20pp over the same recent window). That difference is why S's terminal margin is built with more confidence than its absolute level suggests.

Is 5.5% above S's trailing actual? Yes — by 35.2pp, from −29.70%. Justified by: the ex-SBC operating margin has already reached −0.74% from −8.12%; the company guides its first full year of non-GAAP operating profit in FY26 and has delivered five consecutive quarters of positive non-GAAP net income margin; and the ex-SBC series is improving at +7.38pp/yr, so mid-single-digit GAAP is roughly five years of continuation, not a break.

terminal_margin_source: own opex bridge, reconciled to the company's guided non-GAAP profitability through an explicit SBC assumption. Not a peer median. Not clamped.

What the screen used: 10.7%, "industry median of mature profitable peers (pre-profit subject)" — the universe median produced by the sic2 = None defect. Error +5.2pp, i.e. nearly double the built figure, in the direction that flatters the name. On S this matters far more than on ZS in relative terms: 10.7% vs 5.5% is a 1.95x overstatement of terminal earnings power.

---

3. The implied-path test — the Valuation Criteria

spot            $18.53      (Alpaca latest trade, 2026-07-29)
shares          337.0m      (BASIC. Dual-class; diluted not determinable — see §1)
net cash      +$812.5m      (VERIFIED — the screen had $656.8m; see S_Research.md §4)
EV            $5,432m       EV/Sales 5.18x
revenue       $1,049.0m     TTM to 2026-04-30
years                5      WACC 10%     terminal margin 5.5%

3a. The EBIT-basis exit multiple is declared UNIDENTIFIED. The primary test runs on a sales basis.

Two independent reasons, both required by the framework:

(i) Instrument fragility at a 5.5% margin. The brief states: "if the name runs 1–3% operating margins (distribution, contract manufacturing), an EV/EBIT exit multiple may be the wrong instrument entirely." At 5.5% S is above that band but the sensitivity is severe — required CAGR moves 8pp for every 1pp of terminal margin near the base case, so the answer is determined almost entirely by the least observable parameter. Demonstration:

The EBIT basis returns FAIL at every point: required 53.1% CAGR against 20.8% demonstrated latest-quarter growth is −32.3pp. But the same table shows the verdict is entirely a function of a 5.5%-versus-18.7% terminal-margin judgement on a business currently at −29.7%. That is not a test with power; it is a restatement of the terminal margin.

(ii) The comparator set has no valid EBIT-basis member at S's exit growth. Required: n ≥ 5 mature profitable firms, matching operating model, and dispersion in the matched dimension. The security-software cohort's traded EV/EBIT multiples are FTNT 49.6x, PANW 180.2x, OKTA 141.3x, and CRWD, NET, TENB, ZS have no positive GAAP EBIT at all. So the set has three usable members spanning 49.6x to 180.2x — below the n ≥ 5 threshold, with dispersion so wide it carries no information. A class can be valid and carry zero information; that is a distinct failure from a wrong class, and both return UNIDENTIFIED. Declared.

3b. The sales-basis exit multiple — identified, from S's own history

The identity EV_T/EBIT_T = (1−t)(1−g/ROIC)/(WACC−g) at t = 21%, WACC = 10%, g = 4%, ROIC = 25% gives 11.2x EV/EBIT, which at a 5.5% terminal margin is 0.62x EV/Sales — implausibly low, and the implausibility is itself informative: it says a 5.5%-margin business does not warrant a sales multiple at all on a steady-state basis. So the identity cannot set the sales anchor here, and I use S's own trading history, per valuation.md.

Base exit multiple: 4.0x EV/Sales at FY2032.

4.00x is below today's multiple and just below the 2026 median, and that is deliberate and correct. Per valuation.md rule 3, a bull exit multiple below today's multiple is often right: a 20% grower becomes a low-teens grower by the exit year. It is NOT below every stated anchor — the NTRA double-haircut error is not repeated: 4.00x sits above the 2026 p25 of 3.90x and above the 2026 minimum of 3.21x.

Growth-matched validity of the sales anchor. Comparators whose growth brackets S's ~13% growth at exit: CRWD (25.6% latest, 8.33x), NET (33.5%, 38.3x), ZS (25.4%, 7.58x), FTNT (20.1%, 15.2x), OKTA (11.2%, 7.21x), TENB (9.6%, 3.03x). Growth dispersion 9.6%–33.5% brackets 13% ✓; n = 6 ✓; matching operating model ✓. The set spans 3.03x to 38.3x — again too dispersed to set a point estimate, which is precisely why the anchor is S's own history and the peer set is used only as a validity check.

3c. Solve 1 — required 5-year revenue CAGR, sales basis

Held fixed: WACC 10%, 5 years, revenue base $1,049.0m, EV $5,432m, shares 337.0m, net cash $812.5m. Solved for: revenue CAGR. Terminal margin is not a parameter on the sales basis — which is the point of using it.

3d. The margin — demonstrated − required

VERDICT: PASS. Margin +4.1pp against FY27 street consensus; +5.0pp against the latest actual quarter.

This is a genuine PASS, not a PASS WITH ARGUMENT. The implied path (15.8% CAGR) sits at or below what the business has already demonstrated on every comparator I can construct — trailing 3-year, TTM, latest quarter, last fiscal year, and forward consensus. No narrative is required. That is exactly the criteria.md PASS condition.

It is also a narrow PASS and must be read as one. +4.1pp against consensus is inside the noise band that +17.6pp against the trailing CAGR is not. The honest statement is: the price requires 15.8% from a business running at 20.8% and decelerating monotonically from 45.8% over eight quarters. If the deceleration continues at its recent pace, S reaches 15.8% within about two years, and the margin closes to zero. The PASS is real and time-limited, and that dependency is what the invalidation gap in S_Trade_Construction.md monitors.

3e. Sensitivity — over the exit multiple, never over probabilities

Given in §3c. The flip point is an exit multiple of 3.55x EV/Sales — the level at which required CAGR rises to 19.9% and the margin against consensus goes to zero. 3.55x is the ~13th percentile of S's 2026 window and above its 3.21x minimum. So the PASS survives a de-rating to roughly the 13th percentile of the current regime, and fails below it. That is the single most useful number in this section.

3f. The expectations surface {(g,m) : V(g,m) = P₀} — the STATE C requirement, all three outputs

Because the EBIT basis is UNIDENTIFIED, the surface is presented on the growth × exit-multiple plane, which is the identified pair for this name. ✓ = clears today's price; ✓✓ = clears a 15% fund hurdle.

Required margin at the underwritten growth: at g = 19.9% and an 18x EBIT exit, required terminal margin is ~18.7% vs built 5.5% — the EBIT basis fails, as declared. Required growth at the underwritten margin: on the identified sales basis at a 4.0x exit, 15.8% vs demonstrated 20.8%. Region clearing the fund hurdle: g ≥ ~20% at an exit ≥ 5.0x, or g ≥ ~23% at 4.0x. Company history sits at g = 20.8% and falling; consensus at 19.9%. So today's price is cleared comfortably but the 15% fund hurdle is only cleared if the multiple holds at or above 5.0x — i.e. essentially where it trades now, at the 90th percentile of its own current regime. That is the tension in this name, stated rather than hidden inside a point estimate.

---

4. The 12-month target

4a. Own-multiple history and the regime test

Daily EV/Sales series, 2021-06-30 (IPO) → 2026-07-28, n = 1,274. Construction as for ZS.

S's multiple has de-rated in every single year since IPO. There is no stationary window at all.

143x → 32.6x → 10.3x → 10.8x → 6.2x → 4.25x. The one non-monotone step (2023 → 2024) is +0.5x. This is not a mean-reverting series; it is a trend. Only 40.8% of 2026 observations fall inside the 2023-H2–2025 range.

Declarations, per valuation.md: - Full-history percentile (11.9th): UNIDENTIFIED. It contains a 143x post-IPO multiple. Using it produces a target measured in hundreds of dollars. - 5-year percentile (12.1st): UNIDENTIFIED — still contains 2021–22 at 32–143x. - 3-year percentile (20.2nd): UNIDENTIFIED. The window covers 2023-H2 through 2026, across which the median fell from 10.3x to 4.11x — a 60% de-rating inside the anchor window. A percentile drawn from a monotonically de-rating series is reporting the trend, not the position. - The identified anchor is the 2026 window alone, n = 142 trading days. Thin, and stated as such.

4b. The finding: S trades at the 90th percentile of the window that applies

Corrected current multiple 5.18x EV/Sales = the 90th percentile of the 2026 regime (median 4.11x, max 5.63x).

This is the opposite of the screen's framing, and it is the opposite of ZS. ZS sits at the 51st percentile of its 2026 window; S sits at the 90th. On the metric that actually governs a 12-month horizon — position within the regime you are trading in — ZS is the neutrally-priced one and S is the extended one, which is the exact reverse of the EV/Sales headline (7.58x vs 5.18x) and of the 5-year implied-path verdicts. The two horizons disagree on both names, in opposite directions. That is not a defect in the method; it is why valuation.md requires both outputs.

4c. Build

NTM revenue. From AV EARNINGS_ESTIMATES (available for S, unlike ZS):

NTM (Aug-2026 → Jul-2027) spans FY27 H2 and FY28 H1 ≈ $1,306.3m.

Revision history, built in (7/30/60/90-day):

Near-year down, out-year up. The Street is deferring the profitability inflection by roughly a year rather than abandoning it — consistent with FCF margin having gone 8.77% → 4.24% (S_Research.md §2d). Net revision direction: mildly negative inside 12 months, positive beyond. That argues against paying for the near-term multiple.

Named product-cycle events inside 12 months (dated in S_Catalyst_Calendar.md): FY27 quarterly reports; the Prompt Security close and its ~80bp margin drag; the first full year of non-GAAP operating profit landing or missing; Purple AI / AI-SIEM attach disclosures.

Multiple. 4.63x forward EV/Sales — the 75th percentile of the 2026 regime. This assumes partial reversion from today's 90th percentile toward the median, not a hold and not a full reversion. Rationale: a flat multiple is not neutral for a name at the 90th percentile — that is precisely the PANW/CRWD reasoning, applied rather than contradicted — and a full reversion to 4.11x would ignore that consensus has S growing 19.9%.

12-month target: $20.36, +9.9% to spot, at 4.63x forward EV/Sales — the 75th percentile of S's 2026 regime.

Sanity band vs external targets. No external professional target for S is on file in this project, so no gap can be reported. I am not substituting a peer median.

Share-count caveat repeated because it is load-bearing: the target is computed on 337.0m basic shares. S is dual-class and the diluted count is not determinable (all three EDGAR concepts 404). With SBC at 28.96% of revenue, the diluted figure is materially higher and the target on a diluted basis is correspondingly lower. The target should be read as an upper bound on the share-count dimension.

4d. Why an identified target here, when CRWD and PANW were declared UNIDENTIFIED

Same defence as for ZS — a stationary sub-window with its size disclosed — but weaker for S, and I grade it accordingly: - S sits at the 90th percentile of the identified window, so I do not use a flat multiple. I assume partial reversion to the 75th. That is the PANW/CRWD logic applied. - The de-rating is monotone across all five years, so even the 2026 window may not be the end of the trend. The bear case (3.36x, 1st percentile) is the continuation case and carries a 0.30 probability, higher than ZS's 0.20. - The defensible range is −16.7% to +30.8%, centred on +9.9%, and the width is the honest expression of a multiple that has fallen every year since IPO.

The two outputs, side by side: 5-year implied-path PASS at +4.1pp; 12-month target +9.9% with a wide band. Both are modestly constructive, and the 12-month figure is the weaker of the two — which is the mirror image of ZS, where the 12-month figure (+18.4%) is the stronger and the 5-year test fails.

---

5. Cross-checks