V-Score Card

V = 0.25(C) + 0.22(E) + 0.18(U) + 0.12(A) + 0.15(M) − 0.06(F)
  = 0.25(3) + 0.22(2) + 0.18(4) + 0.12(3) + 0.15(4) − 0.06(3)
  = 0.75   + 0.44   + 0.72   + 0.36   + 0.60   − 0.18
  = 2.69

Gates: G₁ = 𝟙[E>1] = 𝟙[2>1] = 1  PASS
       G₂ = 𝟙[A>1 ∨ (C+E+U)≥12] = 𝟙[3>1] = 1  PASS

V = 2.69 × 1 × 1 = 2.69
Verdict: AT_RISK
κ = (2.69 − 3.0)⁺ = 0

Fast screen (Bustamante): 0/3. No proprietary data (Stock licensed, not owned; AEP data belongs to customer). No regulatory lock-in (PDF is open ISO 32000). Not transaction-embedded (adjacent to transactions, not the rail). Strong prior: E ≤ 3.

Dimension Analysis

C = 3 — Compound Cognition (Fragile)

Adobe's 43-year tool chain — Photoshop (1990), Illustrator (1987), Premiere (1991), After Effects (1993) — represents extraordinary crystallized cognition: layer compositing, color science, brush dynamics, non-destructive editing, broadcast delivery compliance, VFX compositing pipelines. $4.3B annual R&D, $12.86B goodwill from a dozen major acquisitions (Macromedia, Omniture, Marketo, Magento, Workfront). 30+ named products with deep inter-module dependencies. Segment merger into one reporting unit signals management views these as unified infrastructure, not separate businesses (10-K L852-857).

But the cognition is being bypassed, not re-derived. The compound knowledge is about HOW HUMANS MANIPULATE MEDIA. Foundation models skip this entirely — natural language to finished output. Adobe Stock's accelerating decline ($450M book, management admits "this shift is playing out more quickly than we had planned for," Q1 FY2026 transcript L135) is the leading indicator: the first segment where AI generation made the tool chain optional.

Revenue-weighted analysis: professional creative tools where cognition remains load-bearing (≈35% of revenue, VFX compositing, broadcast color, print production) score C=3-4. Enterprise DXP where cognition is genuinely complex (≈25%, customer journey orchestration, real-time CDP) scores C=4. But prosumer CC (≈25%, being bypassed by Canva/CapCut/Midjourney) scores C=1-2, and Acrobat (≈15%, PDF is ISO 32000, document AI commoditized) scores C=2. Blended: C = 2.9, rounds to 3. Direction is down.

Frontier models will not re-derive VFX compositing or customer journey orchestration within 2 years. They don't need to — they bypass the prosumer tool chain entirely, and that's where revenue erosion accelerates.

E = 2 — Irreducible Infrastructure (Confirmed, c_ℓ ≠ ∞)

Three claimed infrastructure moats were stress-tested:

AEP infrastructure (real, limited scope). 35 trillion segment evaluations and 70 billion profile activations per day genuinely cannot run locally (Q1 FY2026 transcript L164). Enterprise DXP contracts require 18-36 months and $5-20M to unwind. For this 25% of revenue, E = 3-4.

Firefly training data (competitive advantage, not infrastructure). "Commercially safe" licensed content is a brand trust narrative, not a wall. Shutterstock licenses its library to Google for training. Getty has comparable deals. Any company with $200-500M builds a similar dataset in 12-18 months. This is an M-dimension signal (gravity/brand), not E-dimension (infrastructure).

$22.22B RPO (contractual, not structural). RPO measures committed payments, not infrastructure dependency. When contracts expire, switching cost drops toward zero. Adobe's RPO-to-revenue spread is the narrowest in its peer cohort: ADBE +1pt vs CRM +4pt, NOW +6pt. If RPO were infrastructure lock-in, the spread would widen, not trail peers.

The FTC tell. DOJ alleged Adobe "failed to provide a simple cancellation mechanism." Settlement reached March 2026 (10-Q L1252-1272). If lock-in were genuine infrastructure, you wouldn't need dark patterns. The complaint reveals product-based lock-in was insufficient, so management manufactured friction. EU Data Act mandating portability (10-K L1961-1966) confirms regulators view the switching cost as artificial enough to require intervention.

Switching timelines by segment:

Compare to genuine c_ℓ ≈ ∞: SAP ERP = $50-200M, 3-5 years. ADP payroll = regulatory + employee records + tax filing. Adobe's highest switching cost ($20M/3yr for enterprise DXP) is an order of magnitude below true infrastructure lock-in.

U = 4 — Ecosystem Breadth (Solid)

30+ named products spanning 18 distinct workflow categories across creative, document, marketing, and AI domains (10-K L487-850). Five customer audiences from consumer (billions of Acrobat Reader users) through enterprise (99 of Fortune 100). 10+ departments served: creative/design, video/film, marketing, IT/data, document/legal, e-commerce, project management, content operations, B2B sales, customer service.

Deep cross-module data flows: Firefly → Creative Cloud → GenStudio → AEP → activation → analytics creates a closed loop that makes wholesale departure impractical. Not quite SAP-level (doesn't touch HR, finance, operations), but genuine breadth that forces customers to evaluate leaving 5-10 tools simultaneously rather than switching one at a time.

A = 3 — Distribution & Discoverability (Upgraded from 2)

MCP endpoint strategy is architecturally correct — Adobe atomizing Photoshop, Express, and Acrobat capabilities as Model Context Protocol endpoints for agent-first discovery (Q4 FY2025 transcript L48, L214). ChatGPT integration live in Q1 FY2026 (Acrobat, Express, Photoshop conversational editing). LLM Optimizer trials grew 13x quarter-over-quarter (50 customers Q4 → 650+ trials Q1, Q1 FY2026 transcript L177). 70% of AEP customers on agentic capabilities (Q1 FY2026 transcript L178). Record API calls for content automation. Firefly as model marketplace (30+ partner models, credit monetization) positions Adobe as orchestration layer, not just model provider.

Why not A = 4: Firefly $250M = less than 1% of $26B ARR. Most LLM integrations are ChatGPT only; Copilot, Claude, Gemini are "expected" (future tense, Q1 FY2026 transcript L118). No specific API volume metrics — "record" is qualitative. Agents still default to foundation models (Midjourney, DALL-E, Runway) for creative tasks, not Adobe. Adobe is embedding IN other platforms, not being the platform agents default to.

M = 4 — Ecosystem Gravity (Holding at core, eroding at edges)

$26.06B ARR, 850M+ MAU, 96% subscription revenue. Three "industry-standard" products (Photoshop, Illustrator, After Effects). PDF and PostScript are Adobe-originated standards. Content Credentials emerging as content provenance standard. 150+ customers with >$10M ARR, growing >20% YoY. Massive SI partner ecosystem (Accenture, Deloitte, IBM, Publicis, WPP). Commerce has "thousands of third-party extensions" (10-K L784-785). Firefly Foundry custom models (2,500+ created) create deep enterprise lock-in.

Why not M = 5: No counterparty network effects (unlike SAP where suppliers/customers coordinate through the system). Figma displaced Adobe XD. Canva took prosumer graphic design. CapCut gaining video. Adobe Stock declining structurally. Gravity weakening at prosumer edges despite holding at enterprise core.

F = 3 — Ecosystem Friction (Bifurcated)

Consumer/prosumer: deliberately LOW onboarding friction — freemium entry (Reader, Express, limited Firefly credits), AI reducing need for training (10-K L396-398), self-serve purchasing via app stores and adobe.com. Enterprise DXP: HIGH friction — SI/consultant dependency (Solution Partner Program, professional services), long implementation cycles, "solution-specific skills assessments" for enterprise customers (10-K L993-997). FTC settlement reveals some of the consumer stickiness was coercive rather than product-driven.

Regime Context (15-Week Window)

Regression: r_ADBE = α + β_mkt·SPY + β_sector·IGV + ε
Window: 2025-12-09 to 2026-03-27 (75 trading days)

α̂ = −43.2% ann    (t = −0.80, p = 0.42 — NOT significant)
β_mkt = −0.52      β_sector = 0.76
σ_idio = 28.2%     R² = 0.345

Performance attribution: Sector contribution −22.4% (73% of total loss). Alpha contribution −12.9% (ADBE-specific: CEO exit, AI structural concerns, Q2 guidance). Market contribution +3.5% (negative β helped in downtape).

IR measures the regime, not the stock. At ρ_intra = 0.52, the market is selling "software" as a category. Idiosyncratic signal is buried in sector noise. The −1.53 IR looks like signal but isn't — 15 weeks is too short to distinguish ADBE-specific alpha from the sector gravitational field. IR does NOT gate the verdict.

δ — The Structural Discount

V is scored against structural properties (filings, transcripts, product architecture), not price. V ⊥ r_sector(t) by construction. The −30.8% drawdown does not change C, E, U, A, M, or F.

Forward P/E of 8.91 on 12% revenue growth implies the market is pricing structural impairment beyond what V-Score measures. Comparable AT_RISK name Snowflake (V = 2.78) trades at ≈25x forward. ADBE at 8.91x implies V_market ≈ 1.5-2.0, somewhere between COLLAPSED and AT_RISK.

δ = V − V_market ≈ 0.7-1.2. The market is applying more structural discount than the dimensions justify. But δ is context, not a trade signal — κ = 0 means ADBE receives zero weight in the survival basket regardless of the discount. Over-discounting relative to structure doesn't make it a survival position. The V-Score says the structural defenses themselves are insufficient.

Thermodynamic Summary

Adobe sits at the fault line between compound cognition (C = 3, 43 years of crystallized tool-chain knowledge) and infrastructure irreducibility (E = 2, no mandate, no proprietary data moat, cloud is commodity). The core thermodynamic function — creative intent → finished media via tool chain — is being bypassed by foundation models that generate finished output directly from natural language prompts. Intelligence flows AROUND the tool chain, not through it.

What prevents collapse? Breadth (U = 4) and gravity (M = 4). Too many workflows to leave at once. Industry-standard tools with decades of skill investment. 850M MAU creating distribution momentum. These are real resistances — but they're dissipative, not conservative. They slow the flow but don't redirect it. Under sustained pressure from AI generation making individual creative apps optional, breadth and gravity decay.

Adobe Stock's accelerating decline ($450M, confirmed cross-ticker via SSTK −5% subscribers, GETY −14% purchasing customers) is the canary. The first segment where intelligence flowed around the tool chain. The question is the propagation speed: t*(prosumer) ≈ 3-5 years, t*(professional) ≈ 7-10 years, t*(enterprise DXP) ≈ never.

Management understands the physics. MCP endpoints (A = 3), Firefly as orchestration layer, ChatGPT integration — these are attempts to position Adobe as infrastructure that intelligence flows THROUGH rather than around. Execution is early: Firefly $250M vs $26B ARR = less than 1%. AI-first ARR tripled YoY to ≈$750M but must grow 10x to offset structural erosion. The math: $750M at 200%/yr vs $5-6B exposed prosumer CC eroding at 5-10%/yr — current trajectory favors survival but with margin compression and multiple contraction.

Revenue Decomposition

Total ARR: $26.06B

Durable ≈55% (≈$14.3B)

Exposed ≈45% (≈$11.7B)

Conviction & Basket Verdict

κ = (V − 3.0)⁺ = (2.69 − 3.0)⁺ = 0
w_ADBE = W_S · κ / Σ_j κ_j = 0

Basket verdict: WATCH. AT_RISK with κ = 0. Zero weight in survival basket. Monitor for structural upgrades.

Upgrade triggers (both required for EMBEDDED)

• E → 3 (+0.22): AEP/DXP becomes dominant revenue driver, or Firefly Foundry creates genuine infrastructure lock-in at scale (not 2,500 models — 25,000).
• A → 4 (+0.12): Agents default to Adobe APIs for creative tasks. MCP endpoints become the standard creative tooling surface. Firefly ARR crosses $2B+.
• Combined E→3 + A→4: V = 3.03, κ = 0.03. Minimum viable EMBEDDED.

Downgrade triggers

• C → 2 (−0.25): C&MP growth decelerates below 10% by Q2 FY2026 (currently 50/50). Prosumer CC erosion becomes visible in segment revenue. V drops to 2.44.
• E → 1 (−0.22): EU Data Act forces portability. NRR/GRR eventually disclosed and shows deterioration. V drops to 2.47.

Monitor predictions

• C&MP growth slowdown (50%): C&MP subscription revenue growth < 10% YoY by Q2 FY2026. Q1 came in at 12%, but Q2 comp is harder. If hit → C downgrade signal.
• SSTK organic decline (70%): SSTK organic content revenue declines YoY FY2026. Cross-ticker confirmation of stock image marketplace structural decline.
• CEO successor: Timeline "few months." Strong hire = potential regime break. Weak hire or extended vacancy = accelerating talent drain (anti-dilutive shares 2.4x YoY).

Sensitivity

Only path to EMBEDDED requires simultaneous E→3 AND A→4. Single-dimension upgrades are insufficient.

Sources: 10-K filed 2026-01-15 (FY2025), 10-Q filed 2026-03-25 (Q1 FY2026), Q1 FY2026 transcript 2026-03-12, Q4 FY2025 transcript 2025-12-10, yfinance market data 2026-03-28.