Phase Space AI

Valuation

Zscaler [ZS]

Zscaler [ZS] — Valuation

As of 2026-07-29. Spot $153.72. Framework v1.5.1.

---

1. COMPANY STATE — declared first, before any instrument is chosen

STATE C — scaling but economically observable.

Not State A — not profitable on GAAP, and the operating margin is not low-variance around a positive level. Not State B — revenue is subscription and contracted, not driven by an exogenous cycle. (The cost side has just acquired a cyclical exposure through memory/storage/processor pricing, which is noted and is a real development, but it does not make the revenue cyclical.) Not State D — revenue is $3.17bn, gross margin is stable and positive, FCF is $890m, and the reverse DCF returns solutions across the entire parameter range. Nothing is unidentified about the economics.

Consequence of State C, per valuation.md: the terminal margin must be BUILT through the opex bridge and never assumed; the instrument is the two-dimensional expectations surface {(g,m) : V(g,m) = P₀}; and all three of {required margin at underwritten growth, required growth at underwritten margin, the clearing region} must be reported. All three are below.

evidence_grade: B. Statements are audited, RPO and contract liabilities are XBRL-tagged, my billings reconstruction validates to 0.06% against the company's own last-published figure, and the FY27 growth rate is management's own. The deduction is for (i) the retirement of billings and NRR as reported metrics, (ii) ARR — now the primary metric — being unaudited and 4pp inorganically inflated, and (iii) EARNINGS_ESTIMATES returning empty, so there is no consensus anchor for the 12-month build beyond company guidance.

---

2. Terminal margin — BUILT, with the bridge, and above the trailing actual

The parameter the framework has got wrong five times. Here is the build, not an assumption.

Base case: m_EBIT,T = 15.0% GAAP, at FY2031.

m_gross,T           77.5%      (TTM 76.7%; 12-quarter range 76.1–78.5%; +0.8pp for mix shift
                                into non-seat metered usage, net of appliance-hardware cost
                                inflation and the guided capex step-up)
  − R&D             21.0%      (TTM 26.5%; the line with the most obvious scale leverage —
                                platform R&D is largely fixed against a doubling revenue base)
  − S&M             32.0%      (TTM S&M+G&A = 54.9% combined; split 43.5 / 11.4 on the FY25
                                10-K ratio. S&M carries the least leverage: new-logo
                                acquisition is the stated priority and requires more coverage,
                                not less — see the CFO's plan to add capacity in the
                                2,000–10,000-user segment)
  − G&A              9.5%      (from 11.4%; the standard scale line)
  − other            0.0%
  ─────────────────────────
  = m_EBIT,T        15.0%

Hard constraint check: 15.0% ≤ 77.5% ✓. The full expense bridge reconciles by construction above — the ceiling test is passed with 62.5pp of headroom, and the bridge, not the ceiling, is the real test per valuation.md.

Why 15.0% and not something else — the reconciliation that has to be made explicit

The whole question is SBC. ZS's non-GAAP operating margin is already 23.0% and hit an all-time high in FQ3 FY26. Getting from there to a GAAP number requires an SBC assumption:

The base case requires SBC to fall from 24.68% of revenue to 15.0% — a 9.7pp decline over five years, on a line that moved −0.20pp over the last twelve months. That is the load-bearing assumption in this memo and it should be read as such. It is not unreasonable: revenue roughly doubles over the period, so SBC dollars need only grow ~22% cumulatively. But it is not demonstrated, and ZS's own record over the last three years is of SBC holding perfectly flat as a share of revenue.

Is 15.0% above ZS's own trailing actual? Yes — by 19.7pp, from −4.73%. valuation.md requires that a terminal margin below the trailing actual be justified; this is the opposite case, and the justification for being above is: (i) the non-GAAP line already prints 23.0% and expanded, (ii) FCF margin is already 28.05%, so the cash economics of a mid-teens margin exist today, (iii) the gap is entirely a non-cash compensation line whose ratio must fall as the share base stabilises. PANW — the closest structural comparator, further along the same path — runs 13.5% GAAP operating margin today. ZS at 15.0% in five years is therefore roughly "become PANW," which is a defensible, evidenced anchor rather than an extrapolation.

terminal_margin_source: own opex bridge, base case, reconciled to the company's guided non-GAAP operating margin via an explicit SBC assumption. Not a peer median, not an industry percentile, not clamped.

What the screen used instead: 10.7%, labelled "industry median of mature profitable peers (pre-profit subject)." Per compute_criteria.py, coverage_scan._slim() stores d.get("sic") from a payload that has no sic key, so sic2 is None for all 1,433 scanned names and the "sector median" silently degraded to a universe median. The screen's terminal margin for ZS was therefore neither ZS's nor cybersecurity's. Error: −4.3pp, and — critically — in the direction that makes the required CAGR look lower, i.e. it flattered the name.

---

3. The implied-path test — the Valuation Criteria

Terminal value is 100% of EV in this construction (the instrument discounts a single terminal EV), so the reverse DCF is mandatory as the primary long-horizon output. assets/reverse_dcf.py, verified inputs.

spot          $153.72     (Alpaca latest trade, 2026-07-29)
shares         168.0m     (company's own FY26 fully-diluted guidance; basic BS count 160.741m)
net cash    +$1,772.7m    (VERIFIED — the screen had $887.3m; see ZS_Research.md §5)
EV          $24,052m      EV/Sales 7.58x
revenue      $3,173.6m    TTM to 2026-04-30
years               5     WACC 10%    terminal margin 15.0%

3a. The exit multiple is NOT a free parameter — derive it, then sanity-check

EV_T/EBIT_T = (1−t)(1−g/ROIC)/(WACC−g), with t = 21%, WACC = 10%:

The identity says 11–13x for a business that has fully matured to 4–5% growth. But the exit multiple is applied at year 5 (FY2031), when ZS on the underwritten path is still growing low-double-digits and decaying toward 4%. A two-stage fade from ~12% to 4% supports roughly 18x.

Base exit multiple: 18.0x EV/EBIT. It is 1.4x the single-stage identity value, and the excess is exactly the fade period — stated, not smuggled. Implied compression from today's multiple: ZS trades at no meaningful EV/EBIT today (GAAP EBIT is negative), so the honest comparison is on sales: today 7.58x EV/Sales, and 18x EBIT on a 15.0% terminal margin is 2.70x EV/Sales at exit — a compression of 4.88 turns, or 64%. That compression is not aggressive pessimism; it is what a 2.9x revenue increase against a 25%-growth-to-12%-growth transition mechanically implies.

Growth-matched anchoring, per criteria.md. The comparator set used to sanity-check 18x is stated in ZS_analysis.json and is drawn from names whose growth brackets ZS's ~12% growth at exit: PANW (18.8% trailing / 31.1% latest, 13.5% GAAP OM), FTNT (15.5% trailing / 20.1% latest, 30.7% GAAP OM), OKTA (16.3% trailing / 11.2% latest, 5.1% GAAP OM), CRWD (29.0% / 25.6%), NET (30.5% / 33.5%), TENB (13.5% / 9.6%). Growth dispersion 9.6%–33.5% brackets 12% ✓; n = 6 ✓; matching operating model (subscription security software) ✓. The set is valid. FTNT's traded EV/EBIT of 49.6x and PANW's 180.2x are both far above 18x, which confirms the valuation.md point that the empirical multiple is not independent of the inputs and cannot be used as the anchor — it embeds the growth being paid for.

3b. Solve 1 — required 5-year revenue CAGR

Held fixed: terminal margin 15.0%, WACC 10%, 5 years, revenue base $3,173.6m, EV $24,052m. Solved for: revenue CAGR.

3c. The margin — demonstrated − required, the number the strategy ranks on

VERDICT: FAIL. Margin −18.7pp on management's guided forward growth rate.

Why the forward guide is the right comparator and the trailing CAGR is not. The criterion asks whether the implied path "sits at or below what the business has already demonstrated." Against the trailing 34.8% the answer is a knife-edge −0.4pp. But the trailing CAGR is a statement about 2023–2026, and the company has told the market in prepared remarks that 2027 will be 16–17%. Using 34.8% here would be the exact inverse of the error the brief warns about: not rejecting the future for exceeding the past, but accepting a price because the past exceeded the future. The demonstrated-capability test has to be forward-relevant or it is a momentum indicator wearing a fundamental costume.

Could this PASS WITH ARGUMENT? That requires a specific, evidenced reason the price's requirement is achievable. The evidence points the other way: management guided growth down by half, named new-logo acquisition as the underperforming motion, disclosed two sales-leadership departures, and cut FCF-margin guidance ~5pp on capex. There is no evidenced case for 35% CAGR. The argument does not exist, so this is FAIL, not PASS WITH ARGUMENT.

3d. Solve 2 — required terminal margin at each growth path (exit held at 18x)

At management's guided growth the price requires a ~31.8% terminal GAAP operating margin — above FTNT's 30.7%, which is the highest in the comparator set, and 2.4x PANW's 13.5%. Reaching it would require SBC to fall from 24.7% of revenue to roughly zero. The terminal margin cannot rescue this, which is the second dimension of the State-C surface and it fails in the same direction.

3e. The expectations surface {(g,m) : V(g,m) = P₀} — required, and shown

Combinations that justify today's price at an 18x exit multiple. The fund hurdle region (15% IRR) requires values roughly 1.5x these products.

The clearing region sits entirely outside the union of {what ZS has guided} × {what ZS's opex bridge supports}. Company history is at the bottom-left of this grid (g = 16.5% guided, m = −4.7% actual / 15.0% built). Management's target is the same cell. The valid-comparator overlay puts PANW at m = 13.5% and FTNT at m = 30.7%, both at g < 21% — and only the FTNT corner clears, on a business with structurally lower SBC.

3f. Sensitivity — over the exit multiple, never over scenario probabilities

Already given in §3b: required CAGR ranges 22.1% (30x) to 49.2% (11x). The verdict is FAIL at every exit multiple from 11x to 30x when measured against the 16.5% guide. The flip point — where required CAGR falls to 16.5% — is an exit multiple of ≈41x EV/EBIT, which on a 15.0% terminal margin is 6.2x EV/Sales at exit, i.e. ZS holding 82% of today's sales multiple after five more years of decelerating growth. That is the assumption the current price makes.

---

4. The 12-month target

4a. Own-multiple history and the regime test — run explicitly, because this name required it

Daily EV/Sales series, 2019-01-02 → 2026-07-28, n = 1,902. TTM revenue as known at each date (lagged 32 days from period end, so the series is not forward-looking); shares and net cash held at current verified values so the series measures multiple movement.

Regime test, run as a distributional-overlap test rather than asserted:

Only 12.7% of 2026 observations fall inside the 2023-H2–2025 range. The distributions are effectively disjoint.

Declarations, per valuation.md: - The full-history percentile (3.8th) is UNIDENTIFIED and must not be used. It spans the 2019–2021 ZIRP regime at 35–63x — a rate environment, not a business. Using it would produce a target of $600+. - The 5-year percentile (5.7th) is UNIDENTIFIED for the same reason; it still contains 2021. - The 3-year percentile (9.6th) is UNIDENTIFIED, and this is the non-obvious one: the 3-year window looks clean (2023 15.3x, 2024 14.6x, 2025 16.5x) but is broken by 2026 itself. A name at the 9.6th percentile of a window whose bottom decile is entirely one recent year is not cheap; it has re-rated, and the percentile is reporting the re-rating. - The identified anchor is the 2026 window alone, n = 142 trading days. Stated as thin, and it is the reason evidence_grade is B rather than A.

4b. The finding that reverses the screen's intuition

On the corrected basis ZS trades at 7.58x EV/Sales, which is the 51st percentile of the window that currently applies. ZS is at the median of its own current regime. It is not cheap within the regime it is actually trading in; it looks cheap only against a regime that ended.

(Note: the percentile is computed with net cash at the corrected $1,772.7m. On the screen's erroneous $887.3m the current multiple reads 7.45x — a 1.7% difference that does not move the percentile.)

4c. Build

NTM revenue. FY26 actual-plus-guide: Q1 $788.1m + Q2 $815.8m + Q3 $850.5m + Q4 guide $876.5m (midpoint of $875–878m) = $3,330.9m. NTM as of 2026-07-29 is FY27 (Aug-2026 → Jul-2027): $3,330.9m × 1.165 = $3,880.5m, using management's own 16–17% guide midpoint.

(No consensus cross-check is available: AV EARNINGS_ESTIMATES returns an empty array for ZS. The base is therefore company guidance alone, which is a real weakness in this build and is stated.)

Named product-cycle events inside 12 months, each dated in ZS_Catalyst_Calendar.md: FQ4 FY26 results and formal FY27 guidance (early Sept 2026 — this is where 16–17% becomes a commitment or gets revised); the second sales-leadership appointment; Z-Flex TCV trajectory after $480m in FQ3; the Red Canary integrated SecOps rollout; and the appliance price increase flowing through gross margin.

Multiple. Anchored at 7.47x forward EV/Sales — the 50th percentile of the 2026 regime. This is the neutral assumption by construction: it holds ZS's current multiple onto guided revenue, so all of the return comes from revenue growth and none from re-rating.

12-month target: $182, +18.4% to spot, at 7.47x forward EV/Sales — the 50th percentile of ZS's 2026 regime.

Sanity band vs external targets. Not available from AV (EARNINGS_ESTIMATES empty). No external professional target is on file for ZS in this project, so no gap can be reported. I am not substituting a peer median in its place.

4d. Why I issue a target here when PANW and CRWD were declared UNIDENTIFIED — the explicit contradiction

Both PANW and CRWD sat at the 90th–100th percentile of their own multiple history, and both memos correctly reasoned that "flat multiple" is not a neutral assumption for a name at a near-record multiple — it is an active bet that a record holds. Declaring UNIDENTIFIED was right there.

ZS is the mirror case. It sits at the 51st percentile of the identified window, so the flat-multiple assumption is neutral, and the regime break is handled by narrowing the window and stating n = 142 rather than by abandoning the output. The general rule I am proposing from this: a regime break makes the long windows unidentified; it does not make the output unidentified if a stationary sub-window exists and its size is disclosed. Filed to CALIBRATION_WATCH.md.

The two outputs disagree, and that is the point of having two. The 12-month target is +18.4% and the 5-year implied-path test is FAIL at −18.7pp. There is no contradiction: over 12 months ZS grows revenue 16–17% at a held multiple; over 5 years today's price requires 35% CAGR it will not deliver. valuation.md exists because these are different questions, and reporting only one would be the defect.

---

5. Cross-checks