Phase Space AI

Valuation

JFrog [FROG]

JFrog Ltd [FROG] — Valuation

Spot $73.275 (2026-07-29) · EV $8,152m lease-inclusive · 14.47x TTM revenue

1. Verified inputs

Market cap $8,878m; EV $8,152m; EV/TTM revenue 14.47x. Fully diluted: EV $9,161m, 16.26x.

2. Company state: C — scaling but economically observable

All four State C tests are met with measured evidence:

• Positive and rising gross margin — 78.18% GAAP in Q1 2026, up 284bp YoY.
• Visible expense scaling — S&M 46.71% → 37.51% of revenue over thirteen quarters while revenue growth accelerated 22.0% → 25.8%. R&D 43.29% → 33.65%. G&A 19.84% → 15.42%.
• Cohort data — NRR 120%, 1,225 customers at $100k+ ARR, 80 at $1m+, $100k+ cohort = 77% of ARR.
• A formally guided long-term model — an $800m FY2027 revenue target reaffirmed by the CFO ("we are on track with that number"), plus FY2026 guidance to a specific non-GAAP operating income range of $112–116m.

Not A: GAAP loss-making at −14.09%; margin is not low-variance (−32.1% → −14.1% over 3.25 years). Not B: no cyclicality — sequential revenue has risen in each of the last 23 quarters. Not D: revenue is $563.4m, not negligible; gross margin is 77–78% and stable; financing does not dominate ($725.5m net cash, no debt, a $300m buyback); and the reverse DCF solves across the entire 10x–30x exit range, so State D's "no solution in range" test is not met.

State C mandates that the terminal margin be BUILT, never assumed.

3. Terminal margin: 13.0% — built, and a Tier-2 judgement

m_EBIT,T = m_gross,T − R&D − S&M − G&A
   13.0% =   78.0%   − 25.0% − 30.0% − 10.0%

Hard constraint satisfied: m_EBIT,T (13.0%) ≤ m_gross,T (78.0%), with 65.0pp of headroom. The full expense bridge — the real test, per the record where one name carried a 14.4% terminal margin against an 11.9% gross margin — is satisfied by construction, since the terminal margin is the residual of the bridge rather than an input to it.

It sits 27.1pp ABOVE FROG's own trailing actual (−14.09%), which is what the thesis requires. This is the opposite of the corpus defect where 21 of 84 names carried terminal margins below their own trailing actuals (XZO −26.9pp, AVGO −24.3pp, MSFT −23.7pp).

Opex lines are from the 10-K, with S&M as the residual. AV's sellingGeneralAndAdministrative reports G&A only — $81.151m for FY2025, excluding $222.075m of S&M. Building this bridge from AV's fields alone gives 76.79 − 36.68 − 15.26 = +24.85% for a year in which FROG printed −16.91%: a 41.8pp error, in the direction that manufactures a false PASS.

The named mechanism (required, since FROG is GAAP loss-making)

1. S&M leverage from a product-led motion. −9.20pp of revenue across thirteen quarters while growth accelerated. Growth is not being bought.
2. Expansion-led growth. NRR 120%; the $1m+ cohort compounding +48.1% YoY on 80 accounts. Expansion inside an installed account carries near-zero incremental selling cost — the arithmetic source of (1).
3. Cloud gross margin scaling. Blended GM +284bp while SaaS mix +8pp. Mix shift has turned from dilutive to accretive, which is what permits holding terminal gross margin at the current level.
4. G&A leverage on a fixed public-company cost base: 19.84% → 15.42% on 1.9x the revenue.

The stated weakness — the non-GAAP leg is not demonstrated

13.0% GAAP at terminal implies a non-GAAP operating margin of roughly 28% (13.0 + ~14% terminal SBC + ~1% amortisation) against a ~16.7% TTM actual and an 18.1% FY2026 guide — i.e. +2.3pp a year of non-GAAP expansion.

FROG's demonstrated non-GAAP expansion has been approximately zero for three years: 16.6% (Q4 2023), 18.0% (Q4 2024), 18.1% (guided FY2026). The GAAP margin expands because SBC/revenue is decaying (31.2% → 28.3% → ~26.7%), not because non-GAAP operating leverage is landing.

Therefore: 13.0% is recorded as TIER2_JUDGEMENT, not as a demonstrated fact — and not as INDETERMINATE. The brief permits a terminal margin for a GAAP loss-maker where a named, evidenced mechanism exists; four are named above, each measured from primary filings. The uncertainty is carried by this flag and by the sensitivity in §4 rather than by refusing to produce a number. Per rule 5 of valuation.md, uncertainty reduces position size, never the operating assumption.

4. The implied-path test — FAIL

Terminal value is 83.2% of EV, above the 60% threshold, so the reverse DCF is the primary long-horizon output and any forward DCF is supporting evidence only.

Exit multiple derived, not asserted. The identity EV_T/EBIT_T = (1−t)(1−g/ROIC)/(WACC−g) at t = 21%, g = 6%, ROIC = 30%, WACC = 10% gives 15.8x; at g = 5%/ROIC 30%, 13.2x; at g = 6%/ROIC 25%, 15.0x. Base set at 15.0x. No peer multiple is load-bearing anywhere in this analysis — see §6.

Held fixed: terminal margin 13.0%, exit multiple 15.0x EV/EBIT, WACC 10.0%, horizon 5 years, FCF margin 26.9% (FROG's own demonstrated TTM), shares 121.157m, net cash $725.508m.

Required revenue CAGR: 53.7%

Demonstrated: 25.0%

Margin (demonstrated − required): −28.7pp

Implied compression: today's traded multiple is 14.47x EV/TTM sales. The solved path exits at 15.0x EV/EBIT on a 13.0% margin = 1.95x EV/Sales. Required compression 14.47x → 1.95x, a factor of 7.4x, or −86.5%.

Sensitivity over the exit multiple — the parameter that can change the answer

The FAIL is robust and is not an exit-multiple artifact. The required CAGR exceeds the demonstrated 25.0% in every cell of the grid. The most favourable cell available — 30x exit, FROG's own 26.9% FCF margin — still requires 37.1%, i.e. +12.1pp above demonstrated. This is the opposite of NOW, where the exit multiple was the unexamined parameter that determined the verdict.

Two caveats on the FCF-margin column, both stated rather than resolved silently. FROG's reported 26.9% FCF margin is funded by SBC of 28.3% of revenue; economic FCF net of SBC is approximately −1.4% of revenue, which is the FCF 0% column. The reported figure is the charitable assumption and it is the one used in the headline.

Inverted solves

• Required terminal margin at the demonstrated 25.0% CAGR and 15x exit: 43.8% — versus a built 13.0%, or 3.4x the bridge. It passes the literal m_EBIT ≤ m_gross constraint (43.8% < 78.0%), but the bridge would then have to fit all of R&D, S&M and G&A into 34.2% of revenue against 76.6% today. At a 30x exit it falls to 21.9% — still 8.9pp above the build.
• Required exit multiple at 25.0% CAGR and a 13.0% terminal margin: 50.6x EV/EBIT — versus 15.8x warranted by the identity, 3.2x.
• Fully diluted (134.929m shares): required CAGR 57.6%, +3.9pp. Dilution worsens it.

EV/Sales cross-check — the most charitable honest framing

Because FROG's EBIT is negative today, an EV/EBIT exit is highly sensitive to the terminal margin. On a sales exit (FCF 10%):

The single most favourable defensible reading: the price works if FROG still trades at 8x EV/Sales in 2031, at which point the required 22.6% falls just below the demonstrated 25.0%. And 8x is roughly FROG's own post-ZIRP median (8.96x). So today's price pays for five more years of ~25% growth with no multiple expansion and no reversion below the median. That is a fair price, not a mispricing — which is precisely what a FAIL on this criterion means. It is not a claim that the business is bad.

5. Twelve-month target: $74.24, +1.3% to spot

NTM revenue

Time-weighted across the fiscal boundary: at 2026-07-29, five months of FY2026 remain plus seven of FY2027. Base = (5/12)(630) + (7/12)(760) = $705.8m.

• FY2026 = $630m — the midpoint of the company's own $628–632m guidance from the Q1 2026 call (+18.5% YoY at the midpoint).
• FY2027 = $760m — the FY2026 guide grown at +20.6%, continuing the guided rate. ⚠ FLAGGED: no Street FY2027 figure was obtainable. EARNINGS_ESTIMATES returned {"estimates": []} at HTTP 200 (the documented defect, twelfth name). The FY2027 leg is a house extrapolation off company guidance, not consensus, and is labelled as such rather than invented. The company's own $800m target is used in the bull case only.

The multiple — FROG's own history, with the regime check

Current 14.47x EV/TTM sales.

Regime-overlap check — the full history must NOT be used. Two regimes break at the same point:

1. Rate regime. 2020 median EV/S 38.29x, 2021 median 21.20x under ZIRP, against a 4.91x–19.63x range from 2022 onward.
2. Growth regime. FROG grew 40–50% in 2020–21 against 25% now, so the multiple/growth pairing also differs — a multiple percentile drawn across both is not measuring the same instrument.

The applicable window is 2023-07-29 onward, which contains only the current rate regime and the current 23–26% growth regime. Reporting the 78.9th full-history percentile instead of the 93.8th would understate the valuation by 15.1 percentile points on a pure regime artifact.

History is not declared UNIDENTIFIED: 752 trading days within a single, correctly identified regime is ample support for a percentile, and no peer median is substituted anywhere.

Scenarios

Target: $74.24, +1.3%. The upside is small, and the reason is mechanical rather than editorial: the name sits at the 93.8th percentile of its own applicable multiple history. Per item B16, a process whose every target sits below spot is expressing a house market view — this one is above spot, but only just, and the stated cause is the percentile.

Sanity band against the external target

AV OVERVIEW reports $92.62 across 22 rating analysts (5 strong buy, 16 buy, 1 hold, 0 sell) — +26.4% to spot, a 25.1pp gap to the house base.

The entire gap is one assumption: whether a 93.8th-percentile multiple holds for twelve months. The Street's $92.62 is my bull case almost exactly ($93.07) — it implicitly holds the multiple and takes the $800m FY2027 target at face value. I mean-revert half way and use a guide-consistent $760m.

Two caveats on the external number itself: AV's OVERVIEW block is computed on stale prices (its $9,257m market cap against 121.157m shares implies $76.40 versus an actual $73.275 — 4.3% stale), and spot sits 11.6% below the 50-day average of $82.87, so the published targets may predate the move.

Named events inside twelve months

The two that move estimates: the Q2 2026 print — the FY2026 guide implies near-flat sequential revenue for three quarters, which has not happened in 23 quarters — and any revision or retirement of the $800m FY2027 target, which management said it would "revisit". Both are dated in FROG_Catalyst_Calendar.md, with date confidence labelled.

6. Comparator set: UNIDENTIFIED, deliberately

Named peer: GitLab [GTLB] — the closest same-end-market comparator (integrated DevOps platform, overlapping buyer, competing for the same CI/CD and artifact-management budget).

Rule 6 of valuation.md requires n ≥ 5 mature profitable firms, a matching operating model rather than a matching SIC code, and dispersion in the matched dimension, before any peer figure may be cited. The DevOps-platform reference class has no mature profitable member — GTLB is also GAAP loss-making. The class therefore cannot supply an exit multiple.

Declared UNIDENTIFIED rather than backfilled with a software-sector median. This is exactly why the exit multiple in §4 was derived from the (1−t)(1−g/ROIC)/(WACC−g) identity and cross-checked against FROG's own EV/Sales history. No peer multiple is load-bearing anywhere in this analysis.

7. Criteria summary

Is the Quality PASS perverse?

No — but it is generous, and the window matters. The operating-margin leg reads differently on different windows and all of them must be stated:

The TTM window is flattered by Q3 2024 at −27.44%, an outlier quarter carrying Qwak deal and holdback costs. And roughly 60% of the GAAP expansion is SBC/revenue decay rather than non-GAAP operating leverage, which has been flat for three years.

Diagnosis: REAL, not artifact — but not extrapolable. A falling SBC ratio is genuine per-share value creation, and the February 2026 $300m buyback is the first mechanism to convert it into a falling share count. But SBC/revenue has a floor around 10–12%, so the +7.53pp/yr rate cannot run for five years. Gross margin, growth acceleration and the S&M/G&A line-item trajectories are all independently verified from primary filings, so the PASS is earned — it is simply earned on a narrower basis than the headline number implies. This is the diagnosis the brief asks for: artifact in magnitude, real in direction.