TEAM › analyze
TEAM · Analyze
A portfolio-specific passage was removed from the public build.
Verdict: HOLD — conviction [6.5], unchanged. The score is the same and every input behind it moved. Business quality inflected genuinely. Price attractiveness inflected the other way, and by more.
Executive Summary
Atlassian's Q4 FY26 print was the best quarter in the company's history on the metrics the market watches and an ambiguous one on the metric this framework watches. Revenue grew 28% to $1,766M, Cloud accelerated to +31%, RPO grew +44%, GAAP operating income turned positive for the first time since FY2022, and non-GAAP EPS beat by 25%. The stock rose 38.9% in a session — the largest one-day move in its history — and has now tripled from the $56.01 March low.
The April thesis at $71.55 was that the market had priced a good business as a distressed one. That thesis was correct and it has now been paid. On growth-adjusted EV/Revenue, roughly 90% of TEAM's discount to its own history closed in a single session.
The central question from April — is reported FCF real when stock comp consumes almost all of it? — got a genuine answer this quarter, and the answer is layered:
- The ratio improved. SBC fell from 26.1% to 24.4% of revenue, and is guided to ~19% in FY27. Q4 alone was 22.3%, the best in the series.
- Buybacks exceeded SBC for the first time ever ($1,800M vs $1,607M), and the diluted share count actually shrank — −3.61% in Q4.
- But the cash claim got worse. SBC-adjusted FCF went from +$53M to −$288M, because FCF fell 6.8% to $1,319M while revenue grew 26%.
- And the FY27 SBC reduction is a reclassification, not a saving. Management is swapping equity comp for cash comp. SBC falls ~$195M; cash comp rises ~$223M. Total compensation cost rises $29M. It ends the dilution treadmill — a real governance win — but it creates no owner earnings.
Two agent findings dominated the debate round and both are load-bearing. First, the headline "non-GAAP margin falls 30%→25%" is not a deterioration: the shareholder letter discloses that FY26 was flattered ~4ppt by a one-off Data Center revenue-recognition pull-forward and FY27 carries a deliberate ~3ppt hit from the comp swap. Underlying, margin expands. Second, and cutting the other way: on a like-for-like cash basis, pre-working-capital cash generation grows only 3.5% on 13% revenue growth. Underlying operating profit is inflecting; underlying owner cash is not.
At $153.04 the stock trades slightly above a fair value range of $112–148 (central ~$128). It is marginally rich, not clearly rich — and it is nowhere near an entry.
Phase 0 — Knowledge check
Per Knowledge/INDEX.md and python .mcp/kb.py find, no live note covers TEAM specifically. Three Playbook entries applied directly and all three fired:
| Note | Result |
|---|---|
| [[pitfall-yahoo-share-count-dual-class-fpi]] | ✅ FIRED. SharesOut 159.63M is Class A only — a 37.1% understatement. See §Vendor Traps. |
| [[pitfall-stale-entry-zone-suppresses-a-name]] | ✅ FIRED. Watchlist carried "Add <$68" against a $153 spot — 123% away, the automatic-invalidation threshold. Zone re-derived from scratch below. |
| [[pattern-ai-build-inflates-earnings-while-destroying-fcf]] | ⚠️ PARTIAL. No equity marks inflate TEAM's earnings (net income < operating income — the diagnostic passes clean). But the FCF half fired: 1.41ppt of gross margin is explicitly conceded to Rovo/hosting costs. |
Knowledge/Themes/software-saas.md (fast, as-of 2026-08-04) records that SBC-versus-FCF is the single most common disqualifier in this field and that "the headline multiple is routinely misleading and the SBC adjustment is the real screen." That note governed the weighting of the final verdict.
Phase 1 — Fundamentals
All FY26 figures from the primary 8-K (accession 0001650372-26-000031) and the Q4 FY26 shareholder letter, both read directly. The FY26 10-K is not yet filed. Yahoo/fin.py/roic.ai carry no FY26 data at all.
1.1 Free cash flow and the three-year stall
| FY | OCF | Capex | FCF | FCF margin | Revenue | Basis |
|---|---|---|---|---|---|---|
| 2019 | $466.3M | $44.2M | $422.2M | 34.9% | $1,210M | IFRS ⚠ |
| 2020 | $574.2M | $35.7M | $538.5M | 33.4% | $1,614M | IFRS ⚠ |
| 2021 | $790.0M | $31.5M | $758.4M | 36.3% | $2,089M | US GAAP |
| 2022 | $821.0M | $70.6M | $750.5M | 26.8% | $2,803M | US GAAP |
| 2023 | $868.1M | $25.8M | $842.3M | 23.8% | $3,535M | US GAAP |
| 2024 | $1,448.2M | $33.1M | $1,415.0M | 32.5% | $4,359M | US GAAP |
| 2025 | $1,460.4M | $44.9M | $1,415.5M | 27.1% | $5,215M | US GAAP |
| 2026 | $1,353.1M | $34.1M | $1,319.1M | 20.1% | $6,572M | US GAAP |
⚠️ FY19–20 are IFRS (20-F filer); FY21 exists on both bases and breaks −6.1%. The clean US-GAAP series is FY21→FY26 — five years, the bottom edge of the framework's 5–8yr requirement.
| Window | FCF CAGR | Revenue CAGR | Gap |
|---|---|---|---|
| 7yr FY19→26 ⚠ | 17.7% | 27.4% | −9.7pt |
| 5yr FY21→26 | 11.7% | 25.8% | −14.1pt |
| 3yr FY23→26 | 16.1% | 23.0% | −6.9pt |
| 2yr FY24→26 | −3.5% | 22.8% | −26.3pt |
| 1yr | −6.8% | 26.0% | −32.8pt |
FCF has gone nowhere for three years — $1,415.0M → $1,415.5M → $1,319.1M — while revenue rose 50.8%. FCF margin compressed 1,620bp in five years. On every window FCF growth trails revenue growth, and the gap widens as the window shortens.
1.2 The SBC question — the central issue, resolved into two opposite truths
| FY | FCF | SBC | SBC/rev | SBC/FCF | SBC-adj FCF | Buybacks | Buyback/SBC | Diluted shares | Δ |
|---|---|---|---|---|---|---|---|---|---|
| 2021 | $758.4M | $340.8M | 16.3% | 45% | +$417.6M | $0 | 0% | — | — |
| 2022 | $750.5M | $524.8M | 18.7% | 70% | +$225.7M | $0 | 0% | 253.31M | — |
| 2023 | $842.3M | $948.1M | 26.8% | 113% | −$105.8M | $150.0M | 16% | 256.31M | +1.18% |
| 2024 | $1,415.0M | $1,081.4M | 24.8% | 76% | +$333.6M | $395.3M | 37% | 259.13M | +1.10% |
| 2025 | $1,415.5M | $1,362.2M | 26.1% | 96% | +$53.3M | $779.4M | 57% | 261.79M | +1.03% |
| 2026 | $1,319.1M | $1,606.6M | 24.4% | 122% | −$287.5M | $1,800.5M | 112% | 260.16M | −0.62% |
| Q4'26 | $474.7M | $394.5M | 22.3% | 83% | +$80.2M | $359.3M | 91% | 253.40M | −3.61% |
Is it getting better or worse? Better as a rate, worse as a cash claim — and both are true because the two ratios have different denominators.
- SBC/revenue (26.1% → 24.4%) measures dilution intensity per dollar of business. Revenue grew 26.0%, SBC grew 17.9%. Genuine, and the first sustained improvement since FY22.
- SBC-adjusted FCF (+$53M → −$288M) measures whether cash generation covers the compensation claim on it. SBC grew +$244M while FCF fell −$97M — a $341M swing.
The ratio improved on the strength of the income statement; the coverage collapsed on the weakness of the cash flow statement. The ratio is the leading indicator, the coverage is the outcome, and right now they point opposite ways.
The genuine, unambiguous win: buybacks finally exceeded SBC. FY26 was the first year buybacks fully absorbed dilution (112% of SBC), and it worked — diluted shares fell for the first time ever. Execution was excellent: 19.1M shares at a $94.27 average against $153.04 today, +62% in the money. But see §1.4 — it was not paid for out of FCF.
1.3 Why FCF fell 6.8% while revenue grew 26.0%
Exact FY25→FY26 OCF bridge from the filed statement ($M):
| Line | FY25 | FY26 | Δ |
|---|---|---|---|
| Net income (loss) | −256.7 | −53.8 | +202.9 |
| Stock-based compensation | 1,362.2 | 1,606.6 | +244.3 |
| D&A | 92.4 | 140.7 | +48.3 |
| Lease/leasehold impairment | 0.0 | 80.3 | +80.3 |
| All other non-cash | −2.2 | −44.6 | −42.4 |
| Non-working-capital subtotal | +533.6 | ||
| Accounts receivable | −150.0 | −484.5 | −334.4 |
| Deferred revenue | +366.4 | +153.3 | −213.1 |
| Prepaid, payables, accruals | +48.5 | −44.8 | −93.3 |
| Working-capital subtotal | +264.8 | −376.0 | −640.8 |
| Operating cash flow | 1,460.4 | 1,353.1 | −107.3 |
Working capital consumed $640.8M more cash than in FY25, and $547M of that is two lines. Ex-working-capital, cash generation improved $533.6M. Operating performance was not the problem; billing and collection mechanics were.
| FY25 | FY26 | Δ | |
|---|---|---|---|
| AR, net | $778.3M | $1,269.9M | +63.2% vs revenue +26.0% |
| DSO | 54.5d | 70.5d | +16.0 days |
| Total deferred revenue | $2,481.3M | $2,661.7M | +7.3% |
| Non-current deferred revenue | $254.3M | $166.3M | −34.6% |
| RPO | — | $4,817M | +44% |
The 34.6% collapse in non-current deferred revenue is the most specific corroborating evidence in the file. No collections problem and no enterprise-mix story produces that signature — an enterprise mix shift into larger multi-year deals would raise deferred revenue. Only upfront term-license recognition guts it. Contract value migrated out of deferred revenue (cash received) and into RPO (signed, not received) and AR (invoiced, not collected). Revenue quality is unchanged; cash conversion timing structurally lengthened.
The FY26 tax rate of 344% is an artifact, not an event. Consolidated pretax income of $22.1M is a small residual between profitable jurisdictions and SBC-driven losses elsewhere. Current tax actually fell 36% ($153.7M → $99.0M) while pretax income improved $121M. Note the forward implication buried in it: the non-GAAP framework assumes a 24% rate ($482M); actual provision was $75.9M. The $406M gap is the SBC windfall-deduction shield — and a smaller SBC pool shrinks it. The FY27 direction is genuinely ambiguous, since a 39% higher stock raises the deduction per share vested even as fewer shares vest.
1.4 Capital allocation — and the balance sheet nobody mentioned
| Source / use | $M |
|---|---|
| Operating cash flow | +1,353.1 |
| Capex | −34.1 |
| Free cash flow | +1,319.1 |
| Business combinations (DX + Browser Company) | −1,228.9 |
| Share repurchases | −1,800.5 |
| Debt issued / repaid | 0.0 / 0.0 |
| Dividends | 0.0 |
| Marketable securities liquidated to zero | +429.0 |
Cash + securities: $2,938.0M → $1,249.3M, −57.5% in one year. Net cash +$1,949.5M → +$249.9M, −87%.
Is the buyback self-funded? No. - FCF − buybacks = −$481.4M. The repurchase alone exceeded FCF by 36%. - FCF − buybacks − acquisitions = −$1,710.3M, funded entirely from the balance sheet. - No debt was issued, so it is not levered — but the balance sheet no longer has the capacity to repeat it. FY27 buyback capacity is roughly whatever FCF produces.
🔴 This is a new red flag and it went unmentioned in both the letter's outlook and the sell-side reaction. The April analysis named net cash as a pillar of the thesis. That pillar is 87% gone, and the interest line goes with it: FY26 net interest of +$20.3M flips to roughly −$20M in FY27 (Q4 already printed net −$4.9M), a ~$40M pretax headwind in no consensus revenue line.
Reinvestment: capex is trivial (0.5% of revenue — genuinely capital-light). The real spend is R&D $3,269M = 49.7% of revenue, up 22.5%. The two acquisitions cost $1,228.9M cash and added $998.3M of goodwill — ~81% of the price is goodwill, characteristic of talent/product tuck-ins, not acquired earnings streams.
1.5 Balance sheet and leverage — read deferred revenue correctly
| Metric | FY25 | FY26 | Note |
|---|---|---|---|
| Debt / assets | 16.35% | 16.21% | 🟢 Single $1.0B note, no near-term maturities |
| Total liabilities / assets | 77.7% | 82.7% | ❌ Misleading |
| Deferred revenue | $2,481M | $2,662M | 43.6% of assets — customer-funded float, not debt |
| Liabilities ex-deferred-revenue / assets | 36.7% | 39.0% | 🟢 The correct leverage read |
| Current ratio | 1.22 | 0.78 | ❌ Misleading |
| Current ratio ex-deferred-revenue | 3.98 | 2.60 | 🟢 The correct liquidity read |
| Net cash | +$1,949.5M | +$249.9M | 🔴 |
Deferred revenue carries no interest, no maturity, no covenant, and no refinancing risk. The headline 82.7% liabilities/assets and 0.78 current ratio are artifacts of the subscription model. Solvency is not a concern. Flexibility now is.
1.6 Per-share metrics — on the true diluted count
| FY | Diluted shares | Revenue/sh | FCF/sh | SBC-adj FCF/sh |
|---|---|---|---|---|
| 2022 | 253.31M | $11.07 | $2.96 | +$0.89 |
| 2023 | 256.31M | $13.79 | $3.29 | −$0.41 |
| 2024 | 259.13M | $16.82 | $5.46 | +$1.29 |
| 2025 | 261.79M | $19.92 | $5.41 | +$0.20 |
| 2026 | 260.16M | $25.26 | $5.07 | −$1.11 |
| CAGR FY22→26 | +0.67% | +22.9% | +14.4% | n/m (sign change) |
Revenue/share compounded 22.9% — excellent. FCF/share has declined two years running ($5.46 → $5.41 → $5.07). SBC-adjusted FCF/share is at its worst level in the series. BVPS is $4.17 against a $6.11B accumulated deficit.
1.7 Revenue mix and the FY27 Data Center cliff
| Line | FY25 | FY26 | YoY | FY27E (guide) | YoY | Δ$ |
|---|---|---|---|---|---|---|
| Cloud | $3,447.4M | $4,410.6M | +27.9% | $5,535.3M | +25.5% | +$1,124.7M |
| Data Center | $1,467.2M | $1,830.9M | +24.8% | $1,519.7M | −17.0% | −$311.3M |
| Marketplace & other | $300.7M | $330.7M | +10.0% | $370.4M | +12.0% | +$39.7M |
| Total | $5,215.3M | $6,572.3M | +26.0% | $7,426.7M | +13.0% | +$854.4M |
The Data Center swing is −$675.1M of growth contribution — it added $364M in FY26 and is guided to subtract $311M in FY27. Ex-Data-Center, FY27 is guided at ~24.6% growth ($4,741M → $5,906M). The entire 26%→13% deceleration reconciles to one line.
Three risks inside that, in order of importance:
- The DC decline is back-loaded. Q1 FY27 is guided at −4.0% against a full year of −17.0%, requiring roughly −22% to −25% in H2. The hard quarters are Q3 and Q4 FY27, and Cloud's offset is front-loaded (+28.5% Q1 vs +25.5% FY) exactly where DC pressure is lightest.
- ~6–8pts of Cloud growth is migration, not new business. The letter states migrations contribute "mid-to-high single-digits of Cloud revenue growth." Cloud ex-migration is ~17.5–19.5%, not 25.5%. This does not double-count at the total-company level, but "Cloud +25.5%" overstates Cloud's organic momentum.
- ~1.5–2.0pts of total growth may be inorganic. The letter concedes 2H faces "lapping the impact of the DX acquisition" and Q1 Cloud is guided +28.5% falling to ~24.5% thereafter — a ~4ppt step-down at exactly the lapping point. If so, FY26 organic was ~24–24.5% and FY27 organic is ~11.25%, not 13.0%. DX's revenue contribution has never been disclosed.
1.8 Scorecard
A portfolio-specific passage was removed from the public build.
Composite: 🟡, and the split is the story. The income statement inflected genuinely. The cash flow statement moved the other way.
Phase 2 — Moat & Competitive Advantage
Rating: 6.5 / 10 — wide at the suite, thin at the product. Widening in enterprise, eroding in developer preference, simultaneously.
The moat is not Jira, and it is not a network effect. It is a switching-cost moat around a suite plus a compliance perimeter — currently defended in part by a competitor's forbearance, and financed by dismantling the company's original cost advantage.
2.1 Quantitative base — and why the standard test does not apply
ROIC is not a usable moat signal on this name, and fabricating one would be worse than saying so. Negative GAAP earnings, ROE −19%, and BVPS of $4.17 make every return-on-capital denominator meaningless. Any clean ROIC quoted for TEAM has laundered out stock comp.
What the gross margin does say is more useful: 84.8% GAAP, stable at 82–85% for five years, straight through an AI inference cost shock, with no discounting visible. That is the cleanest moat evidence in the file. But note the devil's-advocate framing: flat is not widening. A deepening moat shows rising gross margin. TEAM's is stable — consistent with a strong defensive position, not with growing pricing power. And FY27 guides it down 1.41ppt on Rovo/hosting costs.
2.2 Adversarial stress-test
(a) Linear / Notion attacking Jira from below on UX — contained, permanently lossy at the margin. Linear went $8.4M → $100M ARR in two years on ~$35K of marketing spend. A number like that only appears when the incumbent is actively disliked, and Jira is. But ~18% of switchers return, and never because they missed Jira — they cite Confluence, JSM, compliance, and 6,000+ Marketplace apps against Linear's ~50. The structural insight: the things developers hate about Jira — configurability, permission depth, admin surface — are byproducts of the exact feature set enterprises pay for. You cannot ship opinionated simplicity and enterprise configurability in one product.
The honest counter: Atlassian's answer is to concede the developer-preference battle and move the buying decision to the CIO. Re-basing a moat from product love onto procurement is a real strategic retreat, and it lands them on ServiceNow's turf. Strength: 7/10 at suite level, 4/10 for Jira standalone.
(b) Microsoft/GitHub bundling — the attack that should be lethal is not being run, and that distinction is the finding. Entry difficulty is zero; GitHub Issues and Projects are free inside an E5 estate. It already worked on one flank — Bitbucket lost to GitHub/GitLab and Atlassian has effectively conceded SCM. Yet in 2026 Microsoft built a bridge, not a replacement: GitHub Copilot for Jira went GA 2026-06-25, and the Copilot coding agent takes assignments from Jira.
Steelmanning Microsoft: its 2026 monetization unit is Copilot seats and Azure inference, not PM SKUs. Capturing the category is worth a few billion of ARR and a displacement war; capturing the agent runtime is worth an order of magnitude more, and Jira is the distribution into 85% of the Fortune 500. But this is rented safety, not a moat — a reversible strategic choice by a rival with infinite resources, and the connector is a plausible Trojan horse. Atlassian's genuine defense is narrower than it looks and worth naming: 44% of MCP users are not on software teams. That surface is not one GitHub serves. Strength: 5/10 — the weakest link, because it is contingent on a rival's choice.
(c) The AI-agent attack on the data layer — steelmanned, then broken. The bull case: 150B+ connections, ~20 years of institutional memory; agents grounded in the graph get 44% more accurate answers using 48% fewer tokens; 1M+ MAU doubling in a quarter, 5M+ tool calls per working day, and ~33% of those calls are WRITES. External agents are not merely reading the graph — they are depositing structured work into it. The attackers' agents build the incumbent's moat.
Five ways it breaks: 1. The MCP server is the commoditization vector, not the defense. An open read/write API over your data is a database with permissions. If the human works in Claude or Copilot, Atlassian has demoted itself from interface to backend — and it still sells seats for the interface. 2. Management's own words break the strongest form. From the Q4 call: "most organizations will have three to five large-scale knowledge graphs, and we intend to be one." That is an explicit concession that this is not a winner-take-all context monopoly. Value then accrues to whoever federates them — a layer above Atlassian. 3. The data is the customer's, not Atlassian's. Institutional memory is a customer asset held in Atlassian's schema. The lock-in is schema, integration and inertia — not ownership. 4. The write share cuts both ways. If a third of MCP traffic is agents writing, the human seat degrades from operator to reviewer — while pricing stays per human. 5. The 44%/48% edge is denominated in a decaying metric. That is a 2026 RAG-quality delta; retrieval and context windows improve every quarter.
🔍 Taxonomy correction, and it matters for the multiple: the Teamwork Graph is NOT a network effect. My data does not improve because another company uses Jira — there is no cross-tenant effect. It is a per-tenant data asset, i.e. a switching cost. Mislabeling it as a network effect is the most common analytical error on this name and it inflates what people will pay. Strength: 6.5/10 — genuine and compounding, undefended against orchestration moving up a layer.
2.3 Revenue-stream map, and two self-inflicted switching-cost resets
| Stream | Scale / trend | Durability | Exposure |
|---|---|---|---|
| Cloud subscription | $1,213M Q4, +31% accelerating (26→26→29→31) | High | Human seat count |
| Data Center | Guided −17%; price +15% Feb 2026; EOL March 2029 | Terminal by design | Migration = a switching-cost reset |
| Marketplace | 6,000+ apps; take rate rising | Medium, self-eroding | Forge migration; Rovo subsumption |
| Rovo / AI | Credits; overages not yet billed | Unproven | No disclosed revenue |
| DX (~$1B, closed Nov 2025) | ~1pt of FY26 ARR growth | Medium | Competitive category |
🚩 The most underrated risk in the story: the moment a customer re-platforms is the only moment their switching cost goes to zero. Raising DC prices 15% to force migration is rational harvesting, but if you are re-platforming anyway, "why Atlassian Cloud rather than Linear, GitHub or ServiceNow?" becomes a live question for the first time in a decade. And a second reset runs concurrently: the Marketplace's Connect framework reaches end-of-support in Q4 2026, forcing every app onto Forge, with the take rate spread widened to coerce it (Connect 20%→25%, Forge 16%→17%). Long-tail apps that don't migrate simply die — weakening the "6,000 apps" retention wall exactly when Linear and GitHub are attacking, and Rovo subsumes the thin-app tier from the other side.
Atlassian is taxing and disintermediating its ecosystem at the moment it most needs that ecosystem as a defensive perimeter. Almost nobody is modeling this.
On the Marketplace itself: it is a genuine two-sided network effect, but asymmetric. App density is the most-cited reason leavers return — supply-side density produces demonstrable demand-side retention. But nobody buys Jira because of the Marketplace; they fail to leave because of it. It is a retention moat, not an acquisition moat, worth roughly half of a symmetric one.
2.4 The seat-based pricing question — the defining 5-year risk
What they have actually shipped: Rovo Credits pooled at org level (25/70/150 per user per month on Standard), overages not currently billed (90 days' notice plus opt-in required), Rovo standalone still $20/user/month, and Collections as the monetization vehicle — "Teamwork Collection with approximately 10 times as many Rovo credits included is a great motivator, inducement for upgrade."
Honest reading: the answer to the agentic threat is to defend the seat, raise its price, and bolt a meter on top. There is no agent SKU. Bundling 10× credits into a pricier Collection is ARPU uplift per human seat dressed as AI monetization.
The tailwind mechanism is currently working — Cloud +31% accelerating on seat expansion and Collections cross-sell, and RPO +44% says customers are committing longer and larger while the AI question is at its loudest. That is the opposite of what an existential threat produces.
🚩 But the sharpest fact in this report cuts the other way: in March 2026 Atlassian cut 1,600 employees — 10% of its workforce — explicitly to self-fund AI, and replaced its CTO. A company whose entire revenue model is priced per knowledge worker just demonstrated on itself that AI lets you run with 10% fewer knowledge workers. The org-level credit pool is the correct escape hatch and it is well designed — but they have not switched it on. Until overages bill, ~100% of revenue remains denominated in human seats. The escape hatch is built, and untested.
(One widely circulated claim I am discarding: that Atlassian reported a "first-ever decline in enterprise seat counts." It is uncorroborated and contradicted by the Q4 disclosure, which names seat expansion as a Cloud driver. Likewise the "Linear took 30% of Jira's market" figure traces to an SEO content farm with no methodology — do not carry it forward. No named enterprise customer loss to any competitor was found in the last three months.)
2.5 R&D at 49.7% of revenue — moat or treadmill?
| Company | R&D / revenue |
|---|---|
| Atlassian | ~50% |
| ServiceNow | ~20% |
| Adobe | ~17% |
| Microsoft | ~12% |
2.5–4× peer intensity, above 50% for three years. The treadmill case is stronger than consensus allows: Atlassian competes in ~8 categories at once against a focused best-of-breed in each (Linear, Notion, ServiceNow, GitHub, Glean, Jellyfish, Backstage). A 50% R&D line is the price of being a suite whose individual products are not the best in their category — structural, not transitional. The moat case is that Forge, the Teamwork Graph, MCP and Rovo are shared infrastructure spent once and used eight times; if true, intensity should fall as the platform matures.
🔍 The real tell is not R&D at all. It is that Atlassian is converting PLG dollars into salesforce dollars. S&M grew +35.8% in FY26 against revenue +26.0%, and FY27 funds "the largest enterprise sales team in the company's history." The original moat source was a cost advantage — near-zero CAC, a product that sold itself. They are dismantling the cost advantage to purchase switching costs. That trade may be revenue-accretive; it is unambiguously moat-quality-dilutive, and switching-cost moats cost far more to maintain than cost-advantage moats.
2.6 Evergreen assessment — the test already ran
−78% from ~$326 (Jan 2025) to $56.01, then a triple to $153 on the sequential direction of one segment's growth rate. A business valued that way is not being priced as an evergreen compounder; it is being priced as a binary referendum on whether per-seat pricing survives agents. A genuine evergreen does not fall 78% on a business-model question, because there is no business-model question.
Read honestly in both directions: revenue compounded 25%+ the entire way down, gross margin never moved, customer count never fell. That was multiple compression, not business impairment.
Verdict: not evergreen — but the failure is in the pricing model, not the asset. The asset is close to evergreen: 350,000 customers, 85% of the Fortune 500, 20 years of institutional memory. The mapping from that asset to revenue is not durable, because it runs through a headcount-denominated meter that management itself just proved is shrinkable. Evergreen requires both. This has one.
2.7 Moat sources, rated
| Source | Strength | Note |
|---|---|---|
| Switching costs | 7.5/10 | The primary moat. Suite-level, not product-level. Two self-inflicted resets in flight. |
| Efficient scale (compliance) | 7/10 | FedRAMP, HIPAA, residency, Isolated Cloud. No newcomer holds this within 5 years. |
| Intangibles (data/graph) | 6.5/10 | Per-tenant, compounding via agent writes. Not a network effect. |
| Network effects (Marketplace only) | 5/10 | Real but asymmetric — and being taxed to 25% and force-migrated. |
| Cost advantage | 3/10 ↓ | Being dismantled deliberately. |
| Brand | 3/10 | Weak-to-negative among primary users. Unusual and material for a switching-cost moat. |
What moves the rating up: overage billing switched on with visible consumption revenue · an explicit agent/non-human SKU · Cloud NRR holding while seat growth flattens · Connect→Forge completing without material app attrition. What moves it down: Microsoft reversing into a GitHub Projects displacement play · Marketplace app count falling through the Forge cutover · DC migration cohorts churning to third parties · Cloud below the 25.5% guide.
Phase 3 — Debate Round
Two tensions were live after Phase 1. Both were resolved against the analyst who raised them, and the resolutions are the most important content in this report.
3.1 Tension 1 — "Non-GAAP operating income is guided to shrink 7%"
Fundamentals flagged this 🔴 as the headline finding. Sentiment found the answer in the shareholder letter. I verified page 20 myself against the PDF — it is verbatim:
"Non-GAAP operating margins in FY26 benefited by approximately four ppts from the impact of the DC EOL announcement on the timing of Data Center revenue recognition. In FY27, as part of our ongoing effort to reduce stock-based compensation as a percentage of revenue, we will change the compensation mix to include a greater proportion of cash, and a lower proportion of equity, which we expect to reduce non-GAAP operating margin in FY27 by approximately three ppts. After adjusting for these impacts, we anticipate our non-GAAP operating margin in FY27 will reflect an increase as compared to FY26."
| Bridge (ppt) | Management's 4.0 | Fundamentals' tested 3.27 |
|---|---|---|
| FY26 reported non-GAAP margin | 30.38% | 30.38% |
| less DC EOL one-off | −4.00 | −3.27 |
| FY26 underlying | 26.38% | 27.11% |
| FY27 guided | 25.00% | 25.00% |
| add back comp-mix transfer | +3.00 | +3.00 |
| FY27 comparable | 28.00% | 28.00% |
| Genuine change | +1.62ppt | +0.89ppt |
Fundamentals conceded and withdrew the flag, and also withdrew its "~350bp of opex deleverage" claim — underlying opex leverage is +2.3ppt, the opposite sign. Underlying operating profit grows +22.2% ($1,701.5M → $2,079.7M), and GAAP operating income is guided $10.4M → $334.2M — a 32× increase, on a measure that already charges SBC in full.
But it stress-tested management's 4ppt and found it ~15–25% generous. Solving the margin identity properly (the pull-forward inflates the denominator too), 4ppt requires $357M of pulled-forward revenue, not the $263M a naive reading gives. Four independent tests — excess DSO ($288M), the billings gap ($287M), the FY27 DC bridge ($212–310M) and the margin claim itself ($357M) — converge at $285–310M. Working estimate ΔR ≈ $295M ≈ 3.27ppt. Management's version would imply FY27 Data Center growing 3.1% in the year they guide it down 17%, which is not credible.
✅ Directionally management is telling the truth, and the DC line corroborates it: quarterly DC growth ran 11% → 20% → +44% → 21% across FY26 against a ~12% trend, with the EOL announced September 2025. The +44% Q3 spike is the pull-forward, visible in the data.
3.2 Tension 2 — the comp swap is not what the bull case thinks
| Line | FY26 | FY27 guided | Δ |
|---|---|---|---|
| SBC | $1,605.1M (24.42%) | $1,411.1M (19.0%) | −$194.0M |
| Incremental cash comp (3.0ppt × $7,426.7M) | — | +$223.0M | +$223.0M |
| Total compensation cost | $1,605.1M | $1,634.1M | +$29.0M (+1.8%) |
Zero of the $194M SBC decline is a reduction in total compensation cost. More than 100% of it is a transfer into cash opex. Mathematically, (FCF − X) − (SBC − X) = FCF − SBC — SBC-adjusted FCF is identically unchanged by a swap. The $29M excess is the tell: cash is paid 100% in FY27 while the SBC saving is only the FY27 slice of grants never made on a ~4-year vest. Owner earnings are $29M worse, not neutral.
The improvement that is real (total comp 24.4% → 22.0% of revenue) is bought entirely with revenue growth, not cost discipline. "Reduce SBC as a percentage of revenue" is a target the denominator satisfies.
✅ What it is genuinely worth: it ends the dilution treadmill. In FY26 TEAM spent $1,800M on buybacks against $1,319M of FCF — 136% of FCF, balance-sheet-funded — to offset dilution. Paying cash directly is cheaper and more honest than issuing stock and buying it back. That is a governance improvement, not an earnings one — and per the field's own record it is the opposite of the SBC behavior that disqualified ZS, HUBS, MDB and PD in the March 2026 screens.
3.3 Tension 3 — does the sign flip in owner earnings survive?
Fundamentals revised FY27 FCF up 19% (base $1,770M) and FY27 SBC-adjusted FCF to +$357M, arguing the working-capital reversal sets the magnitude but not the sign, because even a no-reversal case yields +$240M. Valuation adjudicated: conclusion right, mechanism wrong, and the correction matters more.
| Driver of the +$528M swing to +$240M | $M | % | Recurs? |
|---|---|---|---|
| Working-capital flow deceleration | +375.7 | 71% | Once |
| Restructuring non-repeat | +198.0 | 38% | Once, permanently |
| Underlying operating profit growth | +83.2 | 16% | Yes |
| Cash taxes / net interest / comp swap / capex | −124.7 | −24% | Yes |
"No reversal" is not "no working-capital benefit." FY26's drag was caused by the change in DSO (+16 days), not the level. Holding DSO flat at 70.5d means AR grows only with revenue — a −$165M use instead of −$492M. That is a +$376M swing booked as neutral. The no-reversal case is a normalization case wearing a bear case's label.
The conclusion survives anyway, because once DSO stabilizes at any level the step-change is behind you. The FY28 run-rate test confirms it: revenue $8,466M, margin 25.5%, DSO held → FCF ~$1,868M, SBC ~$1,524M → SBC-adj FCF +$344M = 4.1% of revenue. The sign holds and persists.
The like-for-like cash test is the number to remember:
| $M | |
|---|---|
| FY26 pre-WC FCF (rebuilt) | 1,739.8 |
| ...ex-restructuring, ex-$295M pull-forward | 1,642.8 |
| FY27 pre-WC FCF | 1,700.7 |
| Underlying pre-working-capital cash growth | +3.5% |
On 13% guided (≈11.25% organic) revenue growth, underlying pre-working-capital cash generation grows 3.5%. That is negative operating leverage on cash.
The two figures reconcile exactly: +3.5% + the $223M comp swap and $91M of tax/interest on a $1,643M base (19.1ppt) ≈ +22.6% ≈ Fundamentals' +22.2%. Both are correct. +22.2% is the right measure of the business's operating trajectory; +3.5% is the right measure of what reaches the owner. The valuation runs on the second.
Phase 2b — Valuation
Models applied conditionally per §3 of the framework.
❌ Graham's Intrinsic Value is VOID. GAAP EPS is −$0.21; the formula requires positive earnings. Substituting non-GAAP EPS returns $23.43 against a BVPS of $4.17 hollowed out by a $6.11B accumulated deficit (P/B 36.7×). Same disposition as APP. Nothing below anchors to it. ❌ DDM and DYT: N/A. No dividend, none guided, none plausible while SBC-adjusted FCF is near zero. ✅ Weight-bearing: reverse-DCF, owner-earnings multiples, peer EV/Revenue and EV/FCF, Bogle.
4.1 Discount rate
| Input | Value |
|---|---|
10yr UST (DGS10, FRED, 2026-08-05) |
4.63% |
| 2yr UST 4.18% vs Fed funds 3.63% | The curve is pricing hikes, not cuts |
| ERP / beta | 5.0% / 1.04 |
| CAPM cost of equity | 9.83% |
| WACC used | 10.5% |
The 70bp add-on: beta 1.04 is a lie about an asset that printed −36% in one session and +38.9% in another; zero dividend means 100% of return is terminal price; 57–61% of the DCF sits in terminal value, so the discount rate is the model; and management has revised guidance three times in twelve months, twice sharply downward.
4.2 Reverse-DCF on reported FCF — the hurdle is reachable
Implied 10-year FCF CAGR to justify EV $38,586M at 3% terminal growth:
| FY27 FCF base | 9.5% | 10.5% | 11.5% | TV share |
|---|---|---|---|---|
| $1,350M (bear) | 10.83% | 13.03% | 15.09% | 60% |
| $1,490M | 9.53% | 11.69% | 13.71% | 59% |
| $1,770M (base) | 7.24% | 9.34% | 11.30% | 57% |
| $2,200M (bull) | 4.35% | 6.37% | 8.25% | 53% |
At the revised base the required hurdle is 9.34%. That is below the 5yr revenue CAGR, below the 5yr reported-FCF CAGR of 11.7%, and roughly equal to nominal revenue growth once the DC drag laps — it requires no FCF margin expansion at all, only that FCF grow with the business. That is a materially different and more achievable claim than the 11.69% implied by the lower base.
| 10yr FCF CAGR | 4% | 6% | 8% | 10% | 12% |
|---|---|---|---|---|---|
| $/share | $103.96 | $120.08 | $138.81 | $160.56 | $185.77 |
4.3 Reverse-DCF on SBC-adjusted FCF — where the case does not clear
The base is negative (FY26 −$287.5M), so a CAGR is undefined and none was fabricated. Instead: hold revenue on an explicit path, ramp SBC-adjusted FCF as a percentage of revenue, and solve for the terminal margin the price requires.
| Revenue path | Base $79M | Base $240M | Base $357M |
|---|---|---|---|
| Management's re-acceleration | 28.4% | 27.8% | 27.4% |
| No re-acceleration | 33.5% | 32.9% | 32.4% |
| Organic-adjusted (11.25% start) | 31.7% | 31.1% | 30.6% |
A 352% improvement in the FY27 base — $79M to $357M — moves the required terminal outcome by 1.0 percentage point. In this frame the near-year base is nearly irrelevant; the terminal margin is everything.
| Benchmark: SBC-adj FCF margin | |
|---|---|
| ADBE | 33.3% |
| CRM | 26.2% |
| NOW | 19.4% |
| WDAY | 12.0% |
| MNDY | 10.8% |
| TEAM's own FY21 peak | 15.0% |
| TEAM FY27E | 4.8% |
| TEAM required by FY36 | 27.4% – 32.4% |
The identity makes it concrete. At working-capital-neutral steady state, SBC-adj margin = FCF margin − SBC/revenue:
| FCF margin | − SBC/rev | = SBC-adj | |
|---|---|---|---|
| TEAM FY27E | 23.8% | 19.0% | 4.8% |
| ADBE actual | 41.4% | 8.2% | 33.2% |
| TEAM required | ~36% | ~8.6% | 27.4% |
To clear the hurdle Atlassian must simultaneously lift FCF margin from 23.8% to roughly 36% — Adobe territory — and cut SBC from 19.0% to about 8% of revenue. Not one or the other. Both. Neither is in any guide.
| Terminal SBC-adj margin | Re-accel | No re-accel |
|---|---|---|
| 12% (≈WDAY) | $71.50 | $61.36 |
| 15% (TEAM's own peak) | $87.38 | $74.84 |
| 19% (≈NOW) | $108.55 | $92.80 |
| 26% (≈CRM) | $145.60 | $124.24 |
4.4 Multiples, and where TEAM sits in its own band
| Metric | FY26A | FY27E |
|---|---|---|
| EV / Revenue | 5.87× | 5.20× |
| EV / ARR | 5.84× | 4.95× |
| EV / FCF | 29.3× | 21.8× (base $1,770M) |
| P / non-GAAP EPS | 26.2× | 28.1× (on $5.44) |
| P / SBC-adj FCF | negative | ~107× |
| SBC / revenue | 24.4% | 19.0% |
⚠️ Forward P/E (28.1×) is above trailing P/E (26.2×) — per [[pitfall-forward-pe-above-trailing-pe-flags-an-inflated-base]] this is the tripwire: the multiple rises going forward because earnings fall. FY26's $5.85 was inflated by the DC pull-forward. The honest trailing base is ~$5.00–5.10.
TEAM's own band (raw closes, not adjusted — per [[pitfall-adjusted-close-breaks-multiple-bands]]):
| Date | Price | EV/Rev | P/FCF |
|---|---|---|---|
| FY24 end | $176.88 | 10.16× | 32.5× |
| Peak Jan-25 | $306.78 | 16.4× | 56.8× |
| FY25 end | $203.09 | 9.87× | 37.7× |
| 52w low Mar-26 | $56.01 | 2.20× | 10.3× |
| Apr-26 analysis | $71.55 | 2.87× | 13.3× |
| Pre-print 8/06 | $110.17 | 4.22× | 21.2× |
| Today | $153.04 | 5.87× | 29.4× |
Growth-adjusted — the honest measure:
| Date | Fwd EV/Rev ÷ fwd growth | % of own normal |
|---|---|---|
| FY24/FY25 steady state | 0.42 / 0.45 | 100% |
| Mar-26 low | 0.11 | 26% |
| Pre-print 8/06 | 0.28 | 67% |
| Today | 0.40 (0.46 on 11.25% organic) | ~92% |
The single-day move took TEAM from a ~33% growth-adjusted discount to its own history to a ~8% discount. Roughly 90% of the gap closed in one session, on a print that halved the forward growth rate.
4.5 Peer set
⚠️ The dual-class check found three more traps before any per-share math — WDAY understated 22.9%, ASAN 41.4%, and MNDY by ~26% in a variant where both Yahoo fields agree but diluted weighted-average is far higher. Without the check WDAY would have looked 23% cheaper than it is.
| TEAM | NOW | MNDY | ADBE | CRM | WDAY | HUBS | ASAN | |
|---|---|---|---|---|---|---|---|---|
| EV ($B) | 38.59 | 125.11 | 4.18 | 104.90 | 183.68 | 41.50 | 8.99 | 1.82 |
| Rev growth | 26.0% (13.0% gd) | 20.9% | 26.7% | 10.5% | 9.6% | 13.1% | 19.2% | 9.2% |
| EV / Revenue | 5.87× | 9.42× | 3.39× | 4.41× | 4.42× | 4.34× | 2.87× | 2.30× |
| EV / FCF | 29.3× | 27.6× | 13.5× | 10.6× | 12.8× | 14.9× | 15.6× | 23.6× |
| SBC / Revenue | 24.4% 🔴 | 14.7% | 14.4% | 8.2% ✅ | 8.5% | 17.0% | 16.9% | 27.2% 🔴 |
| EV / SBC-adj FCF | negative | 48.5× | 31.4× | 13.3× | 16.9× | 36.1× | 185× | negative |
| Fwd P/E | 28.1× | 24.8× | 16.7× | 9.6× | 12.3× | 14.1× | 12.8× | 19.7× |
TEAM ranks 5th of 8 on the metric that flatters it and 7th of 8 on the metric the knowledge base says actually screens this field. It carries the highest SBC/revenue in the set except ASAN, and is one of only two names with negative SBC-adjusted owner earnings. The other is a $2.2B decelerating micro-cap at 2.30× revenue. On growth-adjusted EV/Revenue at 11.25% organic growth, TEAM's 0.46 is joint-most-expensive in the set alongside CRM and above NOW.
(MNDY is the striking name here — 26.7% growth, positive SBC-adjusted FCF, 3.39× EV/Revenue. Flagged, not this report's subject.)
4.6 Bogle — the multiple dominates the answer
Buying at 28.1× FY27E non-GAAP EPS of $5.44. Dividend yield 0.00%.
| EPS path | 5yr CAGR | Exit 15× | Exit 18× | Exit 22× | Exit 28× |
|---|---|---|---|---|---|
| Base +22%/yr | 16.1% | +3.9%/yr | +7.7%/yr | +12.2%/yr | +17.3%/yr |
| Mid +16%/yr | 11.5% | −0.2%/yr | +3.5%/yr | +7.7%/yr | +12.6%/yr |
| Bear +12%/yr | 8.4% | −3.0%/yr | +0.6%/yr | +4.7%/yr | +9.5%/yr |
Holding EPS growth at base and moving the exit multiple across 15×–28× swings the answer 13.4ppt. Holding the multiple and moving EPS growth across the full range swings it 7.5ppt. The multiple assumption is 1.8× more powerful than the entire fundamental forecast. At 28× forward you are not primarily underwriting Atlassian's business — you are underwriting a multiple. The peer median forward P/E is 14.1×.
4.7 What is $153.04 actually paying for?
Reported-FCF basis, base $1,770M, 10.5% WACC, 3% terminal:
| Assumption set | $/share | % of price |
|---|---|---|
| (a) FY27 guide permanent — 13% fading to 5%, FCF margin flat, no improvement of any kind | $140.16 | 91.6% |
| (b) + management's FY28 re-acceleration, margin still flat | $161.86 | +14.2% |
| (c) + FCF margin recovery | not required | −5.8% |
91.6% of today's price is covered by extrapolating the FY27 guide with zero improvement. The price sits between (a) and (b): it requires roughly 59% of the promised FY28 re-acceleration and no margin recovery at all. That is a much less demanding ask than the pre-debate model implied (77% covered, 23% resting on an unguided promise).
On the owner-earnings basis the answer inverts. SBC-adjusted FCF is −$287.5M actual and ~+$357M in FY27E, of which 71% is working capital and 38% is a vanished restructuring charge. At any margin TEAM has ever achieved the stock is worth $71–87; at NOW's current 19.4% it is worth $93–109. Essentially 100% of today's price requires an outcome the company has never produced, at any scale, in any year.
The gap between those two frames is now the entire argument. Which is why conviction on the valuation call is 6.5 and not higher: better data widened the disagreement between the models rather than closing it, because the near-year base is load-bearing in one and nearly irrelevant in the other.
4.8 Fair value and zones
| Method | Range | Weight |
|---|---|---|
| Reported-FCF DCF, base $1,770M, 5–8% growth | $112 – $139 | High |
| SBC-adj DCF @ 15–19% terminal margin | $75 – $109 | High |
| SBC-adj DCF @ 22–26% (bull) | $106 – $146 | Medium |
| EV/FY27E Revenue 4.0–5.0× | $118 – $147 | High |
| Forward P/E 18–22× on $5.44 | $98 – $120 | High |
| Bogle PV, conservative | $110 – $150 | Medium |
| Graham's IV | VOID | — |
Fair value: $112 – $148 · central ~$128
Bull (re-acceleration delivered and SBC-adj margin reaching CRM's 26% by FY36): $145 – $162 Bear (FY27 guide is the permanent state, working capital does not reverse): $75 – $100
| Zone | Level | Basis |
|---|---|---|
| Strong buy | ≤ $85 | ≤15.6× fwd — the peer-median multiple. Here you are paid for the SBC problem rather than paying for a fix that has not happened. |
| Entry | $100 – $120 | 18–22× fwd. Below the SBC-adjusted DCF at NOW-level terminal margin. |
| Hold / no action | $120 – $163 | Current price sits here. Do not add. |
| Trim | Trim 30x fwd |
30× FY27 (June-2027) non-GAAP EPS — renders ≈$163 today and rises on its own as earnings grow. |
On the trim, per [[pitfall-stale-entry-zone-suppresses-a-name]]: written as a multiple, not a dollar — a $163+ trim would go stale the moment FY28 earnings arrive and would force selling a compounder for the wrong reason. "fwd" means FY27, ending 2026-06-30→2027-06-30, non-GAAP EPS. Re-set at the next /analyze or /value, and specifically after the Q2 FY27 print when the fiscal roll makes "fwd" mean FY28. If FY28 revenue growth prints above 15% with SBC below 15% of revenue, raise to 33×. Below 15% growth, cut to 24×.
⚠️ The prior watchlist zone of Add <$68 was 123% away from spot — an automatic-invalidation condition under the pitfall note. The zones above are re-derived from scratch on post-print models, not adjusted off the April numbers.
Phase 1b — Sentiment & Intelligence
5.1 Analyst reaction — and how much of the book is stale
Seven dated PT raises on 2026-08-07, all reiterations, zero rating changes: Oppenheimer $110→$200, KeyBanc $115→$185, Wells Fargo $140→$180, BTIG $130→$180, Macquarie $130→$170, Guggenheim $115→$165, Truist $100→$160. Web-reported and not yet in the dated feed: BofA upgrade Neutral→Buy, $105→$175 ("AI beneficiary rather than an AI victim"), Morgan Stanley $180, Baird $200, Jefferies $200, Bernstein $309.
⚠️ Do not use the consensus mean. Yahoo still reports targetMeanPrice $153.77 and targetMedianPrice $130.00. The median sitting below the print price is the tell — most of the book has not moved. Unrevised since before 8/06: Barclays $112 (5/07), Citigroup $110 (5/01), Cantor $107 (5/01), Mizuho $145 (4/14), BMO $95 (6/25). The targetHighPrice $480 is a legacy pre-crash artifact. Expect the mean to drift toward ~$175 on catch-up revisions alone.
5.2 Insider activity — verified row by row
Pulled insider_transactions, not the insider_purchases summary, per [[pitfall-yahoo-insider-purchases-counts-rsu-grants]]. The pitfall held — Duffy 121,512, Chuong 297,030, Liu 3,257 and nine directors at 1,885 each are all $0.00 grants that the summary field would report as purchases.
A portfolio-specific passage was removed from the public build.
🔍 The founders' lockstep sell program — and the finding nobody has reported. Cannon-Brookes and Farquhar each sold 7,665 shares every trading day from at least 2025-11-17 through 2026-02-06 — ~467,565 shares each, ≈$65M apiece at a ~$140 average. This reconciles to Cannon-Brookes' Class B position falling 47,534,373 → 47,066,808 across the quarter.
They stopped on 2026-02-06 and have not sold a share in six months — right as the stock began its slide from ~$100 to the $56.01 April low. That is a cleaner signal than the buy-plan announcement, and it is verifiable from Form 4 cadence rather than from a press release.
Sizing the $250M buy plan: Cannon-Brookes' stake is ~47.07M Class B ≈ $7.2B (22.7% as-converted, ~42.7% of votes). $250M is ~3.5% of his stake and ~4× what he sold in the FY26 program. No prior open-market purchase appears anywhere in his Form 4 record — his 2025 and 2022 plans were both sell plans.
⚠️ But the timing deserves a hard flag. The 8-K says he intends to enter the plan "no earlier than the close of trading on August 7, 2026", subject to a cooling-off period. He had not bought a single share when the stock rose 39%. He announced at $110 and will transact at $150+. The announcement was worth far more than the purchase will be.
5.3 🔇 Noise — a material part of the move is mechanical
sharesShort 17,645,455 as of 2026-07-13 = 11.08% of float, short ratio 4.19 days, and rising into the print. Volume on 8/07 was 15.4M against a 4.96M three-month average — 3.1×. A +39% gap-and-go with 11% of float short, 3× volume, a founder-buy headline, and a 24%-in-30-days run into the event has a substantial covering component. Do not read the full 39% as a fundamental re-rating.
5.4 The bear case, stated fairly
- Cash conversion deteriorated while the income statement improved, and everything that improved improved partly on a recognition-timing change. The one metric timing cannot flatter went the wrong way. Best bear point, and under-covered.
- The pull-forward is admitted, not alleged — "pulled greater up front term license revenue into FY26 from FY27." Part of the beat was borrowed from the year they just guided down.
- Reported revenue declines sequentially for two quarters: Q3 $1,787M → Q4 $1,766M → Q1 guide $1,705–1,715M.
- ARR deceleration is not an accounting artifact — 23% → 18% on the metric management themselves call the clean one. ~1pt is DX; the rest was not quantified.
- GAAP economics are still thin — FY26 net loss $53.8M, accumulated deficit $6.11B, FY27 GAAP op margin guided 4.5%.
- The balance-sheet cushion is largely spent, having repurchased $1.8B across a year in which the stock fell 38%.
- AI revenue remains undisclosed. Every Rovo metric is a usage metric. Bulls read "attach"; bears read "free."
- Seat-based pricing under agentic AI — the structural version of the March thesis. Counter-evidence in this print is real (Cloud accelerated four straight quarters, NRR >120%, cross-sell not migration driving the Q4 beat); the fair bear reply is that seat contracts run 1–3 years so compression cannot have shown up yet.
5.5 What management did not guide
🔴 Ten metrics were guided to one decimal place. FCF got an adjective. Revenue growth, Subscription ARR, Cloud, Data Center, Marketplace, GAAP and non-GAAP gross margin, GAAP and non-GAAP operating margin, and share count — all numeric. On cash: "We expect to generate healthy free cash flow, allowing us to opportunistically offset dilution and return capital to stockholders."
They can model FY27 GAAP operating margin to 4.5% but will not put a number on cash — the one line that turns on a ~$545M working-capital swing. In a year following a 6.8% FCF decline, that omission is information.
🚩 And the share-count language is weaker than it reads. Exact text: "We expect our net diluted share count to remain, at a minimum, relatively flat in FY27 versus FY26." FY26 weighted-average was 260,163k; the Q4 exit was 253,399k. Holding flat against the FY26 average permits a 2.67% rise from the Q4 exit. The −3.61% Q4 shrink was bought with 136%-of-FCF buybacks and is not repeatable — and management has quietly said so.
5.6 Regime placement — and the second data point in an open retrospective
Per Knowledge/Market/regime.md, TEAM is an Act I casualty of the Feb–Apr 2026 SaaSpocalypse ($184.00 → $56.01, −70%), analyzed by this agency at $71.55 on 2026-04-26.
Knowledge/Themes/software-saas.md records ADBE as half-falsifying the March thesis: the sell-off was indiscriminate AI fear, but Adobe then guided ARR down on deferred price increases — a pricing-power failure the screens had no column for. TEAM resolves the same test in the opposite direction:
| March 2026 thesis | ADBE (scored 8/04) | TEAM (scored today) |
|---|---|---|
| Sell-off is indiscriminate AI fear, not a business problem | ✅ half right | ✅ confirmed — Cloud accelerated 26→26→29→31 through the fear |
| A moated compounder is mispriced | 🔻 falsified — ARR guided down on deferred price rises | ✅ held — DC prices raised 15% in Feb 2026, strong DC retention still guided |
| Price re-rates as the fear passes | ⚠️ de-rated further, partial recovery | ✅ +173% off the low in ~3.5 months |
The discriminating variable is the same in both cases and it is not the multiple — it is whether the company can still take price. That diagnostic is now two-for-two, and the outcome one-for-two. This is the most consequential Act I retrospective data point yet and it belongs in the open retrospective.
❌ Vendor Traps — five found, all confirmed against primary source
| # | Trap | Reality |
|---|---|---|
| 1 | SharesOut 159.63M |
Class A only. True total 253,767,862 (Class A 159,634,245 + Class B 94,133,617, 10-Q cover 2026-04-24). 37.1% understatement. [[pitfall-yahoo-share-count-dual-class-fpi]] — the MktCap ÷ Price detection test works cleanly. The same check then caught WDAY (−22.9%), ASAN (−41.4%) and MNDY (~−26%) in the peer set. |
| 2 | EV $28.06B |
True EV $38.59B. Off by $10.5B (27%) — it implies $10.5B of net cash against an actual +$250M. This error propagated into the Moat analyst's first draft as "~$10B net cash… can fund this fight for a decade," which is false and was corrected in synthesis. |
| 3 | PE(fwd) 24.90 |
Stale pre-print consensus implying FY27 EPS $6.11–6.15 against a guidance-derived $5.44–5.57. True forward P/E ~28×. [[pitfall-vendor-forward-eps-is-stale-on-the-day-of-a-guidance-cut]] |
| 4 | P/B 43.98 · D/E 141.41 · CurrentRatio 0.70 · BVPS 3.44 |
None reconcile to either balance sheet. Correct: 36.7× · 93.5% · 0.78 · $4.17. They mix Q3 FY26 TTM, the FY25 balance sheet and the Class-A-only count. |
| 5 | Yahoo totalRevenue $6.19B, revenueGrowth 0.317 |
Not refreshed post-print (actual FY26 $6.572B; the growth figure is Q3's). Do not run valuation off vendor fields within 24h of a print. |
⚠️ New tooling limitation worth a note: roic.ai returned only 2 years of history on this account (Free plan), making it unusable for the 5–8yr series §1 requires. The entire long history here came from SEC XBRL companyfacts instead. The Tool Hierarchy in CLAUDE.md assumes roic.ai supplies 20–40yr statements; on this account it does not.
✅ [[pitfall-roicai-fcf-field-returns-ocf]] was checked and did not fire — cf_free_cash_flow reconciled exactly to the filing for FY25.
Phase 4 — Synthesis & Verdict
The weighted view
Weighting per §5 of the framework: for a high-growth software name, Moat and Sentiment normally outweigh static valuation. I am departing from that default here, and the reason is specific. Knowledge/Themes/software-saas.md records that in this exact field, across 12+ screens, SBC-versus-FCF was the single most common disqualifier and the headline multiple was routinely misleading. That is not a generic valuation objection; it is the field's own scored lesson, and it earns the weight.
What genuinely improved, and it is a lot: - GAAP operating income positive for the first time since FY22; Q4 GAAP net income +$139M — the first genuinely profitable quarter - Cloud accelerating four quarters running (26→26→29→31%) straight through a "SaaS is dead" tape - RPO +44% against cRPO +27% — customers signing longer and larger precisely as the AI question peaks. The hardest datapoint in the print to explain away - Enterprise inflection unambiguous: $5M+ ARR customers +70% to 69, $3M+ +50% to 164 - SBC/revenue improving and committed in writing to ~19%; buybacks exceeded SBC for the first time; share count actually shrank - Underlying operating margin expands, and underlying operating profit grows +22.2% — the loudest bear claim in Phase 1 was withdrawn - Founders stopped selling six months ago and the CEO announced a $250M buy plan
What did not improve, or got worse: - SBC-adjusted FCF is −$288M, the worst in the series, and reaches only ~+$357M (4.8% of revenue) in FY27E — of which 71% is a working-capital reversal and 38% a vanished restructuring charge - Underlying pre-working-capital cash generation grows 3.5% on 13% revenue growth. Negative operating leverage on cash - The FY27 SBC reduction is a reclassification. Total compensation cost rises $29M - Net cash fell 87%, $1,949M → $250M. A pillar of the April thesis is gone, and ~$40M of net interest goes with it - FY26's entire headline margin expansion is one-off (−1.83ppt underlying) - FCF has been flat-to-down for three years while revenue grew 50.8% - ~1.5–2pts of growth may be inorganic (DX), putting FY27 organic at ~11.25% - Two self-inflicted switching-cost resets run concurrently through FY27–28 - ~11% of float was short into a 3× volume session — part of the move is mechanical
And the price: $153.04 against fair value $112–148. The stock went from a ~33% growth-adjusted discount to its own history to a ~8% discount in one session. The April thesis was right and has been paid.
The tension I am not resolving, because it should not be resolved yet
The two valuation frames now disagree more than they did before the debate round, not less:
| Frame | Fair value | Says |
|---|---|---|
| Reported FCF | $112–139, 91.6% of price guide-covered | Roughly fairly valued |
| SBC-adjusted owner earnings | $75–109 | 40–100% above fair value |
Better data widened the gap, because the near-year base is load-bearing in one model and nearly irrelevant in the other. Naming that honestly is more useful than splitting the difference and calling it conviction. One more year of data on SBC as a percentage of revenue will resolve more than another quarter of FCF ever will.
Verdict
HOLD — conviction [6.5]. Unchanged score, entirely different reasons.
Business quality moved up. Price attractiveness moved down, and by more. They net to the same number, which is the least interesting fact about this analysis and the one most likely to be misread — every input behind the 6.5 changed.
Do not add at $153. Entry is $100–120, strong buy ≤$85, and the stock is 28% above the top of entry. Do not trim either —
Trim 30x fwdrenders ≈$163, and the position (1.588sh ≈ $243) is too small for a trim to mean anything. The correct action is no action, which is a real answer, not a dodge.A portfolio-specific passage was removed from the public build.
Break triggers
Break: FY27 SBC above 21% of revenue (the comp-swap commitment failing) · Cloud growth below 25.5% in any two quarters · DSO above 75 days at Q2 FY27 (the working-capital drag becoming structural) · share count rising year-over-year · Data Center declining worse than −22% with no Cloud offset · FY28 revenue growth guided below 15% (the promised re-acceleration failing) · non-GAAP gross margin below 85.5% (AI COGS outrunning the AI gateway) · any material Marketplace app-count decline through the Forge cutover.
🟢 Upgrade to [7.0]: FY27 SBC printing at or below 17% of revenue — this is the single variable that reconciles the two valuation frames and it moves the terminal requirement more than any plausible working-capital outcome · Rovo overage billing switched on with disclosed consumption revenue · an explicit agent/non-human SKU · SBC-adjusted FCF durably above 10% of revenue · Cannon-Brookes' $250M actually executed rather than announced.
⏳ Next gate: Q1 FY27 prints ~2026-10-29. Watch, in order: DSO (does the working capital reverse?), SBC as a percentage of revenue (is the swap real?), Data Center against the −4.0% Q1 guide (is the H2 cliff coming early?), and any FCF guidance at all. Revenue will be the least informative number in the release.
Also pending: the FY26 10-K, which will carry the AR note, the DX purchase-price allocation, cash taxes paid, and the year-end share count — four of the largest open questions in this file.
Data gaps — stated explicitly
- FY26 10-K not filed. Missing: receivables note, acquisition PPA, cash taxes paid, year-end share count, SBC vesting schedule, RPO duration split.
- No FY27 FCF guidance — treated as signal, not omission.
- ~$165–185M of the Q4 AR build has no DC-EOL explanation. The innocent reading (enterprise payment terms) is probably right — and that portion does not reverse.
- DX revenue contribution never disclosed, so organic growth cannot be computed from public data. The ~1.5–2.0pt estimate is inferred from the guided Cloud growth shape.
- ARR +18% vs revenue +13% cannot be reconciled without the Cloud/DC ARR split, which Atlassian does not disclose.
- The verbatim earnings-call transcript could not be retrieved (SEC 403s automated access; Yahoo's URL 404s; agent-browser timed out twice). Quotes attributed to the call are as rendered by third-party transcript services — treat wording as approximate. Every quote attributed to the shareholder letter or press release is primary and exact; I read both PDFs directly.
- FY19–20 are IFRS, FY21+ US GAAP, with a −6.1% break on the overlap year. Any CAGR spanning them is flagged.
- Post-print institutional positioning unknown — Q3 13Fs are not due until mid-November.
- Acquisition attribution (Browser Company ~$488M / DX ~$720M) is from press and third-party summaries; cash out $1,228.9M and goodwill +$998.3M are primary.