Shopify Inc. (SHOP) — Investment Tree v1

Stage 7 final essay. Bilingual companion: tree_v1_zh.md. Date: 2026-09-06 · Anchor price: $145.09 (2026-09-04 Nasdaq close, USD) · Market cap $186.68B · EV ~$181.73B · Shares ~1,286.6M · Forward EV/gross profit ~25.6׆ Archetype: durable commerce infrastructure priced on revenue rather than gross profit — a business whose reported growth rate and its economic growth rate have been diverging for four years.

⚠️ Familiarity-bias disclosure. The owner of this repository operates a retail store on Shopify and has direct daily experience of the product. That is genuine domain knowledge and a documented bias risk. Two controls are applied and neither is optional: a full-strength premortem-steelman is published below (§ XI-b) rather than summarised, and every leaf verdict in leaves.md was written to be defensible to a reader who has never used the product. Where the two readings conflict, the one that does not depend on liking the software wins.

Price note. Two figures circulate for 2026-09-04: $145.09 (SHOP, Nasdaq, USD) and CAD $200.76 (SHOP.TO, Toronto). At ~1.38 USD/CAD they reconcile — same common share, two listings, not an ADR. This tree anchors on the USD Nasdaq line throughout.

SOURCE QUALITY: Unusually strong on the company, weak on the market. Revenue architecture, segment gross margins, GMV, Shopify Payments GMV, attach rate, balance sheet, share counts, the buyback authorisation and the agentic-commerce launch are all Tier A from 10-K, 10-Q and 8-K. Valuation multiples, peer comparisons and consensus targets are Tier C by nature — market data, not regulated disclosure. Merchant count and market share are disclosed nowhere and are correctly absent from this tree rather than estimated. All † = Tier C.

0. Company fundamentals — what Shopify is and how it earns

What it is. Commerce infrastructure sold on two very different economic models, and the split is the whole analysis.

Merchant solutions — overwhelmingly payments, plus shipping, Shopify Capital, Markets and tax bundled into one reported line — is 76% of revenue, growing twice as fast as subscriptions, at less than half the gross margin. Subscription gross margin has been rising (77.8% → 81.1%, FY22-25); merchant gross margin has been drifting down (38.8% → 37.8%). The blend is therefore falling on mix alone: 48.1% (FY2025) → 47.7% (Q2 2026).

Cash-flow anatomy.

⚠️ GAAP net income is unusable here and the tree publishes no P/E. FY2025's $1,231M includes ~$294M of negative fair-value swings on equity and other investments — leaving reported net income below operating income. Then Q2 2026 net income of $1,502M exceeded the entire prior fiscal year, on quarterly operating income of $488M. Roughly $1B of below-the-line gain in one quarter. Operating income and free cash flow are the clean reads. (Leaf L1A 1.3.)

Balance sheet — genuinely the cleanest in the peer set. Zero debt outstanding. The only debt Shopify ever carried — $920M of 0.125% convertible notes issued September 2020 — matured and was settled in Q4 2025 for $1,043M cash, above par because the stock had appreciated into conversion. Cash and marketable securities $4,947M against total liabilities of $1,716M and total assets of $14,470M. SBC + payroll tax fell 9.0% → 5.1% → 4.2% of revenue (FY2023-25). First-ever buyback authorised February 2026: up to $2B of Class A, capped at 5% of shares, no expiration; Class A count already down ~17M since December 2025.

ROIC (Tier C derivation, stated so it can be checked): invested capital ≈ total assets $14,470M − cash $4,947M − total liabilities $1,716M ≈ $7,807M; NOPAT ≈ operating income $1,468M × (1 − 20%† tax) ≈ $1,174M; ROIC ≈ 15.0%† against a WACC of 9-10%†. Above the cost of capital, comfortably, and without leverage doing any of the work.

I. One-sentence verdict

Shopify is a genuinely durable business (21/25, zero fatal flags) whose reported revenue growth has been systematically outrunning its gross-profit growth for four years, whose take-rate engine is a bounded lever that has already stopped contributing, and whose price at $145.09 — ~25.6× forward EV/gross profit, 3.3× the highest peer — pays for a favourable resolution of an agentic-commerce question that has no disclosed evidence in either direction.

II. Company snapshot

III. The five facts that drive everything

1. Three-quarters of revenue earns less than half the gross margin of the other quarter, and it is the fast-growing three-quarters. Merchant solutions: 76.2% of revenue, +35%, 37.8% GM. Subscription: 23.8%, +17%, 81.1% GM. Everything about how this company should be valued follows from this row.
2. The take rate rose on attach, not on price — and it has already stopped. Payments attach 53.8% → 68% drove take rate 2.84% → 3.12%. But FY2024 → FY2025 the take rate rose one basis point while revenue grew 30% and GMV grew 29.5%. Monetisation contributed ~1% of that year's growth. The engine was spent before the extrapolation stopped.
3. The balance sheet is pristine and getting better. Zero debt, $4,947M cash against $1,716M total liabilities, SBC + payroll tax down 9.0% → 4.2% of revenue, first-ever $2B buyback. No bear case on this tree is permitted to invoke financing risk.
4. The disruption axis is agentic commerce, not Amazon. Agentic Storefronts launched 2025-12-10 into ChatGPT, Perplexity and Copilot; UCP co-developed with Google, against OpenAI + Stripe's rival ACP — which substitutes precisely the leg Shopify monetises. Amazon, meanwhile, de-escalated in 2026. Nine months post-launch, no agentic GMV and no agentic take rate has been disclosed.
5. The price is 3.3× the best comparable on the metric that survives the mix. Forward EV/gross profit ~25.6× against GLBE 7.73×, TOST 7.59×, AMZN 6.08×, WIX 2.55×. Switching from revenue to gross profit does not rescue the multiple — it removes the SaaS comparison that was flattering it.

IV. The H-0 thesis

At $145.09, Shopify's business quality is not in doubt — its economics are being measured in the wrong unit. The market values a 30%-revenue-growth story on a business whose gross profit compounds several points slower and whose engine of the last four years — payments attach, 53.8% → 68% — is bounded and already decelerating. The next leg must come from GMV growth and from agentic distribution whose economics are unresolved and undisclosed in either direction. H-0 holds if and only if four things are true together: (a) the revenue-to-gross-profit wedge is real, persistent and widening; (b) the attach engine is materially spent, so take-rate expansion cannot repeat its four-year contribution; (c) the base-case return at $145.09 is below an equity hurdle despite a 21/25 durability score — the constraint is price, not business quality; and (d) the inputs a bull would need in order to check the premium — merchant count, market share, agentic GMV and its take rate, customer concentration — are not disclosed anywhere, so the multiple rests on assumptions no outside reader can verify.

What H-0 does not claim: that Shopify is a bad business (it is not — durability 21/25, Q1 business-model persistence 5/5, zero fatal flags); that the stock is a short; that agentic commerce will go badly (it claims the outcome is unknown and is being priced as if known); or that Amazon wins.

Leg (a), shown as arithmetic

Roll the disclosed FY2025 growth rates forward with segment gross margins held flat:

A ~2.6-3.0 point wedge every year, and blended margin falling ~100bps a year. Note that this projection is deliberately generous to the bull: it assumes merchant solutions compounds at 35% indefinitely, which is exactly what causes the margin to fall. The mix stops diluting only when the growth that justifies the multiple stops arriving. There is no assumption set in which Shopify gets both — and that trade-off, not any single number, is the analytical core of this tree.

h0 — computed, not asserted

h0 = Σ(tier_weight × verdict_value) ÷ Σ(tier_weight) over terminal leaves. Weights: crux ×5 · load-bearing ×3 · supporting ×1. Values: 强力支持 0.80 · 支持 0.625 · 部分支持 0.45 · 证据偏反 0.225 · 已证伪 0. 证据不足 excluded from numerator and denominator.

h0 = 24.025 ÷ 39 = 61.6% → 62% · band 55-74% = supported.

Evidence coverage: 6 of 45 tier-weight units (13.3%) excluded as 证据不足 — below the one-third threshold, so no low-coverage flag fires. But note where the thinness sits: 5 of those 6 units are the single crux leaf L1C 3.1. Coverage is fine in aggregate and thin exactly where it matters most.

Crux validation — crux is operationally defined (forcing the leaf to 已证伪 moves h0 across the 75% or 55% boundary), so it was tested rather than asserted:

Calibration check. 2 of 15 leaves are 强力支持 (13%), well under the 80% defend-or-demote threshold; 7 of 15 are non-affirmative. No leaf earned ⛔ or ✗ — that is defended, not assumed, in the calibration note at the foot of leaves.md, along with the one candidate (a stronger version of L1A 1.2) that would have earned ⛔ and was not used because it overstates the honest threshold.

V. Investment tree (condensed)

H-0  Durable business, wrong unit of account, bounded engine, price pays for
     an unresolved question                                          h0 = 62%
│
├── L1A  Are we measuring profit or revenue?                    ✅ 支持
│   │    Unit economics × segment profit pool
│   ├── 1.1  CRUX  Revenue growth structurally overstates GP growth      ✅A 强力支持
│   ├── 1.2        Subscription too slow to defend the blend             ⚠️A 部分支持
│   └── 1.3        GAAP net income unusable as a valuation base          ✅A 强力支持
│
├── L1B  Is the engine that got us here spent?                   ✅ 支持
│   │    S-curve penetration × monetisation-vs-volume
│   ├── 2.1  CRUX  Take-rate rise is attach, and attach is bounded       ✅A 支持
│   ├── 2.2        Practical ceiling ~78-85%, not 100%  [NOT VERIFIED]   ⚠️A 部分支持
│   └── 2.3        GMV must carry revenue growth from here               ✅A 支持
│
├── L1C  Agentic: preserved rake or commoditised?               ⚠️ 部分支持
│   │    Christensen disintermediation × Jobs-to-be-Done
│   ├── 3.1  CRUX  Agentic economics unresolved, undisclosed both ways   ⊗ 证据不足
│   ├── 3.2        Shopify's own behaviour treats the threat as live     ✅A 支持
│   └── 3.3        Best GMV exposed; insulation asserted, unverifiable   ⚠️B 部分支持
│
├── L1D  Who can take it, and can we see them?                  ⚠️ 部分支持
│   │    Porter (rivalry + buyer power) × moat-measurability audit
│   ├── 4.1        Moat's key inputs withdrawn from disclosure           ✅A 支持
│   ├── 4.2        Axis moved from Amazon to channel intermediation      ⚠️B 部分支持
│   └── 4.3        GMV concentration cannot be bounded                   ⊗ 证据不足
│
└── L1E  What does $145.09 require?                              ✅ 支持
    │    Implied expectations × capital-allocation record
    ├── 5.1  CRUX  Valuation, not quality, is the binding constraint     ✅C 支持
    ├── 5.2        Balance sheet removes the financing leg entirely      ✅A 强力支持
    └── 5.3        One large, recent, self-inflicted adjacency error     ⚠️B 部分支持

Tally: 8 ✅ · 5 ⚠️ · 0 ✗ · 2 ⊗          Shared driver: merchant-solutions mix
                                          shift drives L1A 1.1 / 1.2 / L1B 2.3

VI. Business model taxonomy

Five engines. Disclosure separates exactly one of them.

Assignment. archetype_category: compounder-under-review — not incumbent-under-threat, because Shopify is not being displaced from its category; it is co-authoring the standard of the channel that could displace it. The open question is the rake, not survival. Not speculative-fragile-thesis, because nothing with zero debt, $4,947M of cash and $2B of FCF is fragile. The "review" is whether the compounding metric being quoted is the right one, and whether the price already pays for the answer.

cycle_exposure: ai-capex-low. Shopify sells nothing into hyperscaler capex; its revenue is a function of consumer e-commerce GMV. The AI link is genuinely second-order — agentic distribution affects the channel and the multiple, not the capex cycle — which is what ai-capex-low denotes, as against uncorrelated (first-order independent). Recorded deliberately: SHOP would not be protected by an AI-capex de-rating, but it would not be a first-order casualty of one either.

sovereign_exposure: deliberately omitted, and here is why. The band exists to carry ADR and foreign-issuer translation risk — the wedge between a company's home-currency economics and a US holder's realised return. Shopify is Canadian-domiciled, but it files 10-K/10-Q/8-K as a voluntary US domestic filer, reports in USD, and the security analysed here is a US-listed common share, not a depositary receipt. There is no translation wedge to carry, no depositary ratio, no custody or cancellation fee, no home-country withholding on a dividend that does not exist. The CAD line (SHOP.TO) is a secondary listing of the identical share — hence the exact reconciliation at ~1.38 USD/CAD — not an ADR. Stamping a sovereign band here would import a risk category that does not apply and would degrade the field's meaning for the tickers that genuinely need it.

VII. Valuation scenarios

Horizon FY2029 (ending 2029-12-31) ≈ 3.3 years. Method, stated once: FY2029E revenue → × blended gross margin → gross profit → × target EV/gross-profit multiple → EV → + net cash → ÷ ~1,270M† shares. All † Tier C. Full derivation in scenarios.md.

Common base: FY2026E revenue $15.0B† (back-solved from the ~11.9-12.3× forward EV/Revenue band); FY2026E gross profit ~$7.10B†; forward EV/gross profit today = $181.73B ÷ $7.10B = 25.6×.

Scenario 1 — Bear: the rake gets intermediated and the multiple joins the peer band ($65†, −55%, 28%)

Agentic intermediation resolves the wrong way (L1C 3.1 → ✗); GMV growth decelerates to ~12%; attach stalls near 72%; mix dilution continues unoffset.

15.0 × 1.12³ = $21.1B revenue → × 44.5% = $9.39B GP → × 8× = $75.1B EV → + $8B net cash = $83.1B → ÷ 1,270M = $65†

Not a distress scenario. Zero debt and $4,947M of cash mean this is a de-rating, not an insolvency — and even the 8× target is above every peer in the table. 28% because it needs two things to go wrong together; not lower because the multiple half needs no help at all.

Scenario 2 — Base: it works operationally, and the multiple normalises anyway ($165†, +14%, 50%)

Agentic is roughly rake-neutral; GMV compounds ~20%; attach reaches ~76%; the multiple decays 25.6× → 16× as gross-profit growth slows to ~19%.

15.0 × 1.21³ = $26.6B revenue → × 46.5% = $12.36B GP → × 16× = $197.8B EV → + $12B = $209.8B → ÷ 1,270M = $165†

The level is not the point; the rate is. +14% over 3.3 years is ≈ +4.0% per year, from a scenario where essentially everything goes right operationally. You can be right about Shopify's business and still not be paid for owning it at $145.09.

Sensitivity to the one assumption doing the work: 12× → $126† (−13%) · 16× → $165† (+14%) · 20× → $204† (+41%). Anyone who disagrees with this tree should disagree here — but 20× requires believing SHOP still commands 2.6× the highest peer multiple in 2029, after three more years of mix dilution.

Scenario 3 — Bull: Shopify becomes the catalogue layer and keeps the rake ($288†, +99%, 22%)

L1C 3.1 resolves ✅: UCP wins meaningful adoption, agentic GMV monetises at or above blended take rate and expands the addressable pie; attach reaches ~82%; new higher-margin lines (agentic placement, advertising, financial products) hold the blend near 47.5%.

15.0 × 1.27³ = $30.7B revenue → × 47.5% = $14.59B GP → × 24× = $350.2B EV → + $16B = $366.2B → ÷ 1,270M = $288†

The bull requires three things, not one: a favourable resolution of a question with no evidence today; new adjacencies delivering at scale from a company whose last large adjacency (Deliverr) had to be unwound at a loss; and the market still paying ~24× forward gross profit in 2029. Each is plausible. 22% prices the conjunction, not any one of them.

Expected value, and the discounting that decides the case

The two columns disagree and the right-hand one governs. Comparing a 2029 value to a 2026 price without discounting manufactures upside out of arithmetic — a property of the comparison, not a fact about Shopify. At a 10% required return, $145.09 sits ~17% above probability-weighted fair value and the asymmetry is worse than one-to-one: $65 of discounted upside against $98 of discounted downside.

Reverse-engineered market expectations (holding base at 50%, discounted): $145.09 implies bull 37.5% / base 50% / bear 12.5%, against this tree's 22 / 50 / 28. The market is assigning ~37% probability to an outcome nobody can currently observe. The finding is robust to the discount rate — the market implies 32.3% bull at an 8% hurdle and 43.0% at 12%; it only inverts at a required return of roughly zero. Full working in implied_prob.md.

VIII. Competitive analysis

Peer calibration — on gross profit, because revenue misleads in both directions

Shopify's 48% blended margin sits between the archetypes, which is why EV/Revenue misleads twice over — flattering SHOP against SaaS peers whose 70-77% margins inflate their own revenue multiples, and penalising it against payments peers whose thin margins deflate theirs. On gross profit the premium is 3.3× the highest peer and 4.2× the nearest structural analogue.

⚠ Two Tier C inconsistencies carried unreconciled rather than smoothed: the implied-margin column disagrees with the primer's prose for TOST (31.8% vs 26.7%) and WIX (66.7% vs ~77%). NOT RECONCILED. Neither changes the conclusion — at the prose margins, TOST's EV/GP is 9.0× (SHOP still 2.8×) and WIX's is 2.2× (SHOP still 11.6×). CMRC's implied 77.1% matches its prose figure exactly.

The gap this table cannot close, stated rather than hidden: peer growth rates are not in the evidence base. A 3.3× premium could be rational if SHOP's gross profit compounds at 2-3× the peer set's, and this tree cannot verify that. It is why leaf L1E 5.1 carries the C tier suffix and why Section XII Q12 scores ⚠️ rather than ✗. This is the single largest closable evidence gap in the tree.

Competitive position by axis

These are deliberately not netted into one "competitive position" verdict. Netting them would destroy the most important finding in the branch — that the axis changed. A tree that reported "competition: stable" because a de-escalation on one axis offset an emerging threat on another would be worse than useless.

IX. Forward triggers and red flags

Full register with 应对 playbooks and node coverage audit in triggers_redflags.md. Every entry resolves to a live node ID — no orphans.

Triggers (would move a verdict):

Red flags (would damage the thesis or the position):

Catalyst proximity: nothing in 0-4 weeks. Q3 2026 results (4-12 weeks) give the first read on RF2/RF3/RF4. The FY2026 10-K in early 2027 is the earliest realistic window for T1. Falsification proximity: furthest from a break on the balance sheet; closest on the crux — RF1 could fire at a single earnings call.

X. Long-term durability assessment

Result: 21/25 — Medium durability, at the top of the 17-21 "selective hold" band. Zero fatal flags. Full working in durability_test.md.

Long-term holdability verdict

Shopify is holdable for 5-10 years. It is not buyable at $145.09. Those are separate findings and conflating them is precisely the error familiarity bias produces on this ticker.

The durability test's own closing line is the correct summary and this essay does not improve on it: a 21/25 score says the business will very likely still be compounding in ten years; it says nothing about whether $145.09 is a sensible entry. The 4/5 scores on Q2, Q4 and Q5 are not complaints about the business — each is a measurement limit (the moat's width, the agentic economics, the attach ceiling), and each corresponds to a leaf in this tree carrying ⚠️ or ⊗ for the same reason. The business is better understood than it is measured, and the price is set on the measurement.

XI. Tracking dashboard (summary)

Full version in dashboard.md.

Secondary: subscription growth (+17%), merchant growth (+35%), SBC + payroll tax (4.2%), cash ($4,947M), buyback execution, take rate (3.12%), Shop Pay GMV growth (+62%), merchant count (withdrawn).

h0 62% · xii_score 65% · durability 21/25 · fatal flags 0. The h0/xii_score gap is only 3 points, which is unusual — on most trees structural quality runs well ahead of thesis confidence. Here they nearly coincide because the thesis is about structure: the same mix arithmetic that limits confidence is what caps quality. The headroom that gap normally names is, on SHOP, almost absent.

XI-b. Premortem — the strongest case that this loses money over five years

Written to win, not to balance. Required by the familiarity-bias control, and not softened. It is 2031. The position is down substantially. Here is why.

1. The multiple did the killing, and it did not need any help from the business.

This is the part that should be read twice. Suppose gross profit doubles — $7.1B (FY2026E) to $14.3B by 2031, a 15% CAGR sustained for five years, which is a genuinely good outcome. Suppose the multiple settles at 12× forward EV/gross profit, still a 55% premium to GLBE, the most expensive peer on the board today. Then: EV = $172B; add ~$20B of accumulated cash; equity ≈ $192B; ÷ ~1,270M shares ≈ $151 per share.

Five years. Gross profit doubled. The stock went from $145 to $151. That is not a bear case. That is the base case for what happens when you buy a durable compounder at 25.6× gross profit — you spend the entire holding period paying back the multiple, and the business's excellent performance is the thing that funds the repayment rather than the thing that pays you. Nothing in the business had to go wrong.

2. The mix did the second killing, and it was a re-classification, not a mood.

Merchant solutions goes 76% → 82% of revenue. At some point an analyst writes the sentence that has been true for years and stops being ignorable: this is a payments company with a software attachment. Payments companies trade at 3.7-7.6× gross profit (XYZ 3.74×, TOST 7.59×). The de-rating is not sentiment reverting; it is the comparable set being corrected, and comparable-set corrections do not mean-revert. Every quarter of "+34% revenue growth" in the interim made the correction larger, because every one of those quarters shifted the mix further toward the low-margin side.

3. The engine was spent, and everyone could see the number that proved it.

FY2024 → FY2025 the take rate rose one basis point. It was in the filings the whole time. Attach reached the high 70s around 2029 and stopped, exactly as the deceleration series (+4.3 → +3.8 → +3.7pp) said it would. From then on revenue growth was GMV growth, GMV growth converged on global e-commerce growth, and the premium multiple was being paid for a business growing at high single digits to low teens. The company did not disappoint. The extrapolation did.

4. Agentic commerce went the other way, and it took the exact leg that mattered.

ACP won. Stripe owned the payment inside ChatGPT's checkout, and Shopify's role settled into supplying the catalogue and the fulfilment. Shopify kept the GMV — GMV actually grew, which is why it took years for anyone to admit what had happened — and lost the rake, because payments attach is the entire monetisation engine and agentic checkout is precisely what substitutes it. Shopify became a catalogue-and-fulfilment API with excellent uptime. Catalogue APIs are valued as infrastructure, and infrastructure trades at CMRC's 0.83×, not GLBE's 7.73×. The most painful detail: Shopify saw this coming, which is why it built Agentic Storefronts and co-authored UCP — and building the on-ramp to your own disintermediation is what a company does when it has correctly concluded it cannot stop it.

5. The adjacency record was on the table the entire time.

The bull required Shopify to monetise agentic placement, or advertising, or financial products — new adjacencies. The evidence available in 2026 for how this management executes new adjacencies was a single large recent data point: Deliverr, acquired and unwound at a substantial loss in 2023. One observation is a small sample, and the correction was decisive to management's credit. But it was the only observation there was, and the bull case treated it as if the decisiveness of the exit were evidence about the quality of the next entry. It was not.

6. And the familiarity did the quiet damage underneath all of it.

The owner runs a store on Shopify. The software is genuinely good; the merchant experience is genuinely better than the alternatives; the enthusiasm was earned. All of that is evidence about Section XII Q1 — and Q1 was already scored ✅A at maximum weight. Product quality contributed nothing to Q10 (is it undervalued?), Q11 (is there a catalyst?) or Q12 (is the multiple appropriate?) — the three questions that scored 0, 0 and 0.5, and the three questions where the money was actually lost. Every merchant who loved Shopify in 2026 was right about Shopify and wrong about the stock, and there was never any contradiction between those two things.

What the premortem does not say, because the evidence does not support it: Shopify did not go bankrupt, did not dilute, did not need to refinance, and did not lose its merchants. It remained an excellent business throughout. That is the whole point. The loss came from the price, the mix and one unresolved question — every one of which was visible, quantified and in the filings on 2026-09-06.

XII. Investment Scorecard

Format B (15 questions — long-term hold). Per MANUAL_en.md Part K.6. Verdict values: ✅ = 1.0 · ⚠️ = 0.5 · ✗ = 0. Weight tiers: Critical (5×) · Load-bearing (3×) · Important (2×) · Confirming (1×). Markers carry an M1 tier suffix (A/B/C) = the tier of the weakest evidence the answer rests on.

Section A — Business quality (Q1-Q5)

Q1 [Critical 5×]: Does the company have a durable competitive advantage? ✅A Yes. Payments attach rose +14.2 percentage points in four years inside the existing merchant base — that is not a customer-acquisition story, it is existing merchants routing more volume through Shopify, which is what a deepening advantage looks like in numbers rather than adjectives. Switching costs on a live storefront are real; the app ecosystem compounds them; the 10-K names no competitors at all. Caveat recorded rather than scored here: the moat's width is unmeasured (no Tier A/B share data in either direction — leaf L1D 4.1), which is why durability Q2 scored 4/5 and why the width question is carried in Q10 and Q12 instead of being double-counted here. The question asks whether a durable advantage exists; it does. Score: 1.0 × 5 = 5.0

Q2 [Critical 5×]: Is the business model persistently cash-generative at high returns on capital? ✅A Yes. FCF $2,007M (17% margin, FY2025) and $654M (18%, Q2 2026) — a rising margin, not a rescued one. Zero debt; $4,947M cash. ROIC ≈ 15.0%† (NOPAT ~$1,174M† on invested capital ~$7,807M) against a 9-10%† WACC — comfortably above the cost of capital with no leverage doing the work. Asset-light: the merchant funds the inventory. Score: 1.0 × 5 = 5.0

Q3 [Load-bearing 3×]: Is revenue growth durable for 3+ years? ✅A Yes. GMV compounded $197.2B → $378.4B (FY2022-25); Q2 2026 revenue accelerated to +34% from FY2025's +30%; international revenue grew +36% against +30% total with the US still only 44% of GMV; attach still has room. Growth durability is the least contested thing about this company. Score: 1.0 × 3 = 3.0

Q4 [Load-bearing 3×]: Is management capital allocation trustworthy and ROIC-maximising? ⚠️B Partially. The discipline is real, measurable and improving: SBC + payroll tax 9.0% → 5.1% → 4.2% of revenue, zero debt after retiring the converts in cash above par, first-ever $2B buyback. Against that: Deliverr was real capital destruction — acquired and divested at a substantial loss in 2023, exactly the adjacency error this question exists to catch. The correction was decisive, which is to management's credit; one large recent failure with one decisive correction is a genuinely mixed record, not a good one. Score: 0.5 × 3 = 1.5

Q5 [Important 2×]: Does the company have pricing power? ⚠️A Limited demonstrated evidence, and this is a more interesting answer than it looks. The take rate rose 2.84% → 3.12% over four years without a disclosed price increase — the entire move came from attach. Subscription gross margin rising to 81.1% shows cost leverage, not pricing. So Shopify has demonstrated monetisation power (capturing more of a merchant's existing spend) and has not demonstrated pricing power (charging more for the same thing). Those are different, and only the second survives when the first hits its ceiling. Score: 0.5 × 2 = 1.0

Section B — Balance sheet and capital allocation (Q6-Q9)

Q6 [Important 2×]: Is the balance sheet survivable through a bad scenario? ✅A Yes, unambiguously. Zero debt. Cash and marketable securities $4,947M against total liabilities of $1,716M. The only debt ever carried was settled in cash, above par, without refinancing. Even the bear scenario ($65†) is a de-rating, not a solvency event — leaf L1E 5.2 removes the financing leg from every bear case on this tree. Score: 1.0 × 2 = 2.0

Q7 [Important 2×]: Is the company buying back shares at a value-accretive price? ⚠️C Ambiguous, and this is where a first buyback deserves scrutiny rather than applause. At 79.4× P/FCF and 25.6× forward EV/gross profit, with this tree's discounted fair value at ~$120† against a $145.09 price, repurchases at current levels are value-neutral at best and mildly value-destructive on my own numbers. The $2B authorisation is small relative to a $186.68B market cap (~1.1%), so the damage is bounded either way. Credit for the regime change from land-grab to return of capital; no credit for the entry price. Score: 0.5 × 2 = 1.0

Q8 [Confirming 1×]: Is FCF conversion high (FCF / net income > 80%)? ✅A Yes, on any base. FY2025 FCF $2,007M against GAAP net income $1,231M = 163%; against operating income $1,468M = 137%. The GAAP denominator is distorted by equity marks (leaf L1A 1.3), which is precisely why the operating-income comparison is given alongside it — both clear the threshold comfortably. Score: 1.0 × 1 = 1.0

Q9 [Confirming 1×]: Is the dividend sustainable (payout ratio < 50%)? ✅A Trivially. Shopify pays no dividend — the capital-return record consists solely of the February 2026 $2B buyback authorisation. A 0% payout ratio is unsustainable only in the sense that there is nothing to sustain. This row carries no discriminating information for SHOP and is scored ✅ per corpus convention rather than because it tells a reader anything; a scorecard designed for dividend-payers has one question that idles here. Flagged so the 1 point is not mistaken for a finding. Score: 1.0 × 1 = 1.0

Section C — Valuation (Q10-Q12)

Q10 [Critical 5×]: Is the stock undervalued relative to intrinsic value? ✗C No. Probability-weighted FY2029 value discounted at a 10% required return is ~$120† against a price of $145.09 — the stock trades roughly 17% above discounted fair value, with discounted asymmetry of 0.67× ($65 of upside against $98 of downside). The market at $145.09 implies 37.5% bull probability against this tree's 22%, and that gap is robust across 8-12% hurdles (32.3% to 43.0%). Undiscounted the picture flatters (+13.1%), and that flattery is arithmetic, not information. Score: 0.0 × 5 = 0.0

Q11 [Load-bearing 3×]: Is there a clear catalyst or timeline for the discount to close? ✗C No — and scoring this zero is not double-counting Q10. Two independent failures. First, on this tree's numbers there is no discount to close. Second, and this would bind even for a bull: the mispricing here is structural blindness with no forced-disclosure catalyst. Shopify is not obliged to report gross-profit growth as a headline, agentic GMV, agentic take rate, payments revenue, merchant count or market share. The correcting event is a growth deceleration — slow, ambiguous, arguable — not a dated revelation. A thesis whose resolution mechanism is "eventually, arithmetic" cannot be sized as though it had a date. Score: 0.0 × 3 = 0.0

Q12 [Important 2×]: Is the valuation multiple appropriate relative to growth and quality? ⚠️C Measurably extreme, not fully adjudicable. 25.6× forward EV/gross profit against a peer set topping out at GLBE 7.73× is a 3.3× premium to the best comparable and 4.2× to the nearest structural analogue (AMZN 6.08×) — far too large to be Tier C noise. But peer growth rates are not in this tree's evidence base, so the premium can be measured and cannot be fully justified or refuted. Scored ⚠️ rather than ✗ for exactly that reason: the honest state is "extreme premium, incomplete comparison," and this is the single largest closable evidence gap on the ticker. Score: 0.5 × 2 = 1.0

Section D — Long-term hold assessment (Q13-Q15)

Q13 [Important 2×]: Does the business earn ROIC > WACC chronically? ✅B Yes. ROIC ≈ 15.0%† against WACC 9-10%†, with zero debt and $2B+ of annual FCF. Durability Q3 scored 4/5 and the fatal flag explicitly did not fire: Deliverr was one episode, not a chronic pattern of value destruction. Score: 1.0 × 2 = 2.0

Q14 [Important 2×]: Is there a reinvestment runway at high incremental ROIC for 5+ years? ⚠️A Partially, and the caveat is the thesis. Three genuine vectors exist — remaining attach headroom, international outgrowing the core (+36% vs +30%), and enterprise/Plus already generating the majority of GMV. But the primary historical engine is bounded by construction: attach cannot deliver another 14 points twice, its increments are already decelerating, and the practical ceiling (78-85%†) is NOT VERIFIED. A long runway whose main lane is closing. Score: 0.5 × 2 = 1.0

Q15 [Confirming 1×]: Is there meaningful unpriced optionality? ✅A Yes, and it is genuinely good. Shop Pay GMV grew +62% against total GMV +29% — the consumer-side asset is compounding at more than twice the platform it sits on. Add Shopify Capital and Credit (underwritten on transaction data no outside lender can see), POS/offline, and the UCP standards seat alongside Google. None is priced as a core driver; each is a real call option. Score: 1.0 × 1 = 1.0

K.3.5 weighted score derivation

Tier assignment: Critical (5×) = Q1, Q2, Q10 · Load-bearing (3×) = Q3, Q4, Q11 · Important (2×) = Q5, Q6, Q7, Q12, Q13, Q14 · Confirming (1×) = Q8, Q9, Q15. Theoretical max: 39 = 3×5 + 3×3 + 6×2 + 3×1 = 15 + 9 + 12 + 3.

xii_score: 65% — at the lower boundary of the "moderate buy with sizing" band (≥65%), one point of weighted score above "wait or skip." That boundary position is the honest summary of the whole ticker: an excellent business (Sections A and B score 19.5 of a possible 23) dragged to the edge by valuation (Section C scores 1.0 of a possible 9).

Scorecard summary

xii_score 65% · h0 62% · durability 21/25 · fatal flags 0.

Final verdict

Hold-with-sizing / WAIT at $145.09 — 0% today.

Shopify is a high-quality, genuinely durable commerce-infrastructure business with the cleanest balance sheet in its peer set and a real, measurably deepening moat. Sections A and B of the scorecard score 19.5 of 23. Section C scores 1.0 of 9. Nothing about this verdict is a criticism of the company; the entire deduction comes from the price, and no improvement in business quality can close a valuation gap of this size.

The action word is wait, not watch, and the distinction is deliberate: this tree's own numbers say do not initiate at this price, and the decision journal records that so it cannot be read as permission.

Action: 0% at $145.09. Three gates, any one of which opens a starter position:

G3 matters more than G1 or G2. That single disclosure would return ~5 weight units to h0, close the largest evidence gap in the tree, and legitimately move the bull probability toward the market-implied 37.5%. The thesis is more sensitive to one disclosure than to any price move — which is the correct summary of what is actually unresolved here.

Maximum size 3%, capped by the crux sitting at ⊗ and by a Medium-band (21/25) durability score. Exits: RF1 (agentic take rate ≥50bps below blended), RF2 (GMV growth < 15% ×2 quarters), RF8 (any debt or equity issuance).

Sizing is expressed in percentage bands only. This is a framework output with an audit trail, not advice to any person.

2-minute pitch

Shopify's revenue is 76% payments at a 37.8% gross margin and 24% software at 81.1%, so its reported 30-34% revenue growth is running two to three points ahead of the growth that actually matters, and the blended margin has already started falling — 48.1% to 47.7%. The engine that lifted the take rate from 2.84% to 3.12% was payments attach going 53.8% to 68%, not pricing, and the tell is that between FY2024 and FY2025 the take rate rose one basis point while revenue grew 30%: the engine had already stopped contributing. From here revenue growth is GMV growth, GMV growth converges on e-commerce growth plus a share nobody can measure, and the next leg depends on agentic commerce — where Shopify has launched into ChatGPT, Perplexity and Copilot and co-authored a protocol with Google, and where nine months later there is still no disclosed agentic GMV and no disclosed take rate in either direction. For that unresolved question the market pays 25.6× forward gross profit, 3.3× the most expensive peer on the board. Discount the scenarios at 10% and fair value is ~$120† against $145.09, with worse-than-one-to-one asymmetry. The business is a hold for ten years; the stock is a wait. Entry at $120†, or on the one disclosure that resolves the crux.

Relevant risk types (per K.4)

• Valuation / multiple risk — dominant. 25.6× forward EV/gross profit against a 7.73× peer ceiling. Five years of doubling gross profit into a 12× exit multiple returns approximately nothing (§ XI-b).
• Business-model mix risk. Structural, mechanical and already visible: the profitable quarter of revenue grows at half the rate of the unprofitable three-quarters.
• Disruption / disintermediation risk. Agentic commerce, with the specific mechanism that ACP substitutes the payments leg Shopify monetises. Unresolved in both directions.
• Measurement risk (unusual, and material here). Merchant count withdrawn, no market-share series, no agentic disclosure, no concentration figure, no payments breakout. Multiple bull arguments are unfalsifiable with public data.
• Correlated exposure. Consumer e-commerce GMV, and high-multiple commerce/payments platforms generally (AMZN, GLBE, TOST, XYZ, MELI, SE). Not the AI-capex complex — cycle_exposure: ai-capex-low.
• Behavioural risk, disclosed. Owner familiarity with the product. Controlled by § XI-b and by the leaf-authoring standard, not eliminated.
• Not applicable: solvency, refinancing, dilution, sovereign/FX translation.

When NOT to buy (anti-pattern check, per K.5)

• ✗ Do NOT buy at $145.09 for the quality. The quality is real and it is already scored — Sections A and B take 19.5 of 23. Quality is not a reason to pay any price, and Section C's 1.0 of 9 is the entire finding.
• ✗ Do NOT buy on the accelerating revenue line. Revenue accelerating from +30% to +34% while blended gross margin falls is the mechanism of the thesis, not a refutation of it. Each of those quarters shifts the mix further toward the low-margin side.
• ✗ Do NOT buy expecting the mispricing to close on a date. It is structural blindness with no forced-disclosure catalyst (Q11 = 0). The correcting event is a slow, arguable deceleration. Sizing this as though it had a date is the error.
• ✗ Do NOT buy the agentic story before the agentic number. The market already implies ~37.5% probability that it resolves favourably. Paying that when the disclosure is ⊗ is buying an unknown at the price of a probable — and G3 exists precisely so the disclosure can be waited for rather than assumed.
• ✗ Do NOT let product familiarity underwrite the position. Loving the software is evidence about Q1, which already scored maximum. It contributes nothing to Q10, Q11 or Q12 — the three questions that scored 0, 0 and 0.5.

End of essay. Bilingual companion: tree_v1_zh.md. Leaves: leaves.md. Scenarios: scenarios.md. Implied probability: implied_prob.md. Triggers: triggers_redflags.md. Durability: durability_test.md. Dashboard: dashboard.md.