V-Score Card

TICKER:          ZM (Zoom Communications, Inc.)
V-SCORE:         2.30
VERDICT:         AT_RISK
κ (conviction):  (2.30 − 3.0)⁺ = 0.00
BASKET:          FILTER (κ = 0 → excluded)

GATE 1 (E>1):       PASS (E=2)
GATE 2 (A>1∨Σ≥12):  PASS (A=2>1; C+E+U=8<12)
FAST SCREEN:         0/3 — no proprietary data, no regulatory mandate,
                     not transaction-embedded

DIMENSIONS
  C (w=0.25):  3  14yr platform, deep real-time media infra (30 DCs,
                  post-quantum E2EE, proprietary codec) + PSTN (45
                  countries, FCC-regulated). Shallow in newer products.
                  Agent re-derives meetings in months; Phone/CC
                  compliance takes 1-2yr.
  E (w=0.22):  2  NDE 98% = net contraction, multiple quarters. Online
                  churn 2.7%/mo (≈28% annual). RPO $4.2B growing 10%+.
                  FCC regulation on Phone (≈15-18% of rev) is real but
                  minority. EU Data Act reducing switching costs. No
                  proprietary data moat (explicit no-train policy).
  U (w=0.18):  3  13+ products across IT, CX, HR, Sales, Marketing.
                  Cross-sell materializing (6/10 top CC deals include
                  Phone). AI Companion as unifying layer. But many
                  products new/shallow, concentrated in communications.
  A (w=0.12):  2  3,100+ marketplace apps. Custom AI Companion wins
                  (Oracle, Salesforce). But AI revenue "de minimis."
                  Federated AI = consumes others' LLMs, not the
                  intelligence layer. Agents route through Teams/Meet.
  M (w=0.15):  3  $4.87B revenue, $7.8B cash, brand is a verb. 4,363
                  $100K+ customers +9% YoY. Phone 10M seats. But NDE
                  <100%, deliberately weak network effects, migration
                  measured in months not years.
  F (w=−0.06): 2  Meetings = lowest-friction collab tool. Self-serve
                  Online. Phone/CC add telephony complexity but that's
                  inherent to category.

ARITHMETIC
  Raw = 0.25(3) + 0.22(2) + 0.18(3) + 0.12(2) + 0.15(3) − 0.06(2)
      = 0.75 + 0.44 + 0.54 + 0.24 + 0.45 − 0.12
      = 2.30
  Gates: PASS · PASS = 1
  V = 2.30

DURABLE REV:   ≈30-35%
EXPOSED REV:   ≈65-70%

Dimension Analysis

C = 3 — Compound Cognition

Zoom has 14 years of accumulated logic, but it's bimodal: deep where it matters least (by revenue), shallow where it matters most.

Deep (1-2yr re-derivation): The real-time media processing stack — proprietary codec optimized for variable bandwidth, distributed architecture across 30 co-located data centers with private inter-DC links, post-quantum end-to-end encryption (first UCaaS provider). Phone's 45-country PSTN integration adds carrier relationships and regulatory complexity (E-911, USF, CPNI, number porting). Contact Center reached Gartner Magic Quadrant in 3 years from launch, with omnichannel routing and AI-assisted agent workflows.

Shallow (weeks-months): Everything else. AI Companion uses third-party LLMs (OpenAI, Anthropic, Meta) — the intelligence is rented, not crystallized. Docs, Tasks, Clips, Whiteboard, Revenue Accelerator are all well-understood application patterns. Workvivo was acquired, not organically built.

Revenue-weighted, shallow dominates. Meetings generates ≈60% of usage and its application-layer cognition is commodity — the edge cases are in the infrastructure layer underneath, which is real engineering but also held by Microsoft (Azure) and Google (GCP) with superior resources.

Frontier model risk: Application-layer cognition (65%+ of business) erodes toward C=2 by 2028 as AI generates equivalent UX. Infrastructure-layer cognition (codec, DC architecture) survives — physical testing requirements can't be shortcut. But this doesn't help competitively when hyperscalers have thousands of data centers to Zoom's 30. Net trajectory: C declines from 3.0 toward 2.5-2.7 over two years.

E = 2 — Irreducible Infrastructure

The binding constraint. NDE at 98% is the market telling you the gravitational well is shallow.

Switching cost by segment:

Revenue-weighted switching cost is months, not years. The FCC regulation on Phone is genuine — E-911 compliance, USF contributions, CPNI breach reporting, number porting processes — but covers less than 20% of revenue. You can't build a fortress on 20% of the building.

Self-refuting evidence: Zoom itself wins Phone/CC displacements. Fortune 10 Cisco replacement (140K seats), major bank Teams/Cisco displacement (50K to 150K seats), "top 10 CC deals all displacements of leading competitors." If switching costs were prohibitive, these wins wouldn't happen. Zoom's competitive success proves UCaaS switching costs are surmountable.

The EU Data Act (disclosed as risk factor in 10-Q) imposes "interoperability and switching obligations to enable users to switch between cloud service providers without undue delay or cost." The regulatory vector points toward lower switching costs, not higher.

RPO of $4.2B growing 10%+ with lengthening contract durations (long-term RPO +15% YoY) is the strongest E argument — contractual lock-in is real. But contracts expire. RPO is a time delay on switching, not a structural barrier.

U = 3 — Ecosystem Breadth

13+ products spanning 5-6 organizational departments: IT (Phone, Rooms), CX (Contact Center, ZVA), HR (Workvivo, BrightHire), Sales (Revenue Accelerator), Marketing (Events, Webinars), and universal tools (Meetings, Chat, AI Companion, Docs).

Cross-sell is materializing: 6 of 10 largest Contact Center deals also pulled Phone. AI Companion provides a unifying intelligence layer across all products. The "system of action" narrative has real evidence behind it.

But product count is not workflow depth. Three products have genuine depth (Meetings, Phone, Contact Center). The rest are 1-3 years old with shallow adoption data. And all products live within the communications domain — Zoom doesn't touch finance, HR administration, supply chain, or engineering workflows. Compare to SAP (U=5, 20+ workflows across every department) or ServiceNow (U=4, 15+ workflows spanning IT, HR, CS, procurement). Zoom's breadth is real but narrow.

A = 2 — Distribution & Discoverability

The critical structural weakness. In the AI agent era, discoverability means: do agents route through you?

Zoom has 3,100+ marketplace apps (19.3% AI-powered) and a Custom AI Companion paid add-on with Fortune 200 wins (Oracle, Salesforce adopted it). ZVA operates on a consumption-based pricing model — genuinely agent-friendly monetization.

But the federated AI architecture (using OpenAI, Anthropic, Meta LLMs) means Zoom consumes AI intelligence rather than providing it. Agents route through Microsoft Copilot into Teams. They route through Google Gemini into Meet. Zoom is the destination, not the routing layer. AI revenue remains "de minimis" in Q4 FY2026.

Peter Levine asked the right question on the Q4 call: "AI providers controlling the intelligence layer could theoretically build AI-native collaboration suites... what are the structural barriers?" Management deflected to enterprise relationships and brand equity. No structural answer was given. Because there isn't one.

M = 3 — Ecosystem Gravity

$4.87B revenue, $23B market cap, $7.8B net cash, brand literally a verb. 4,363 customers contributing over $100K annually, growing 9% YoY. Phone at 10M seats creates PSTN stickiness. Analyst consensus 61% bullish, mean target $97 (+25%).

But gravity requires escape velocity — it must be too costly to leave. NDE below 100% means customers are net-contracting, which means they're finding it plenty easy to reduce exposure. Zoom deliberately chose weak network effects: open platform, browser-based guest access, PSTN connectivity. This drove adoption but forfeited lock-in. Brand recognition is not switching cost — everyone knows what Zoom is, but that doesn't stop them from clicking a Teams link instead.

The Anthropic stake ($2-4B estimated at IPO pricing, on ≈$51M original investment) adds hidden NAV but doesn't contribute to ecosystem gravity. It's a financial asset, not a structural moat.

F = 2 — Ecosystem Friction (penalty: −0.12)

Low friction, which is simultaneously Zoom's greatest strength and its structural vulnerability. Meetings is the lowest-friction synchronous collaboration tool ever built — create account, share link, join. This UX simplicity was the original competitive advantage and persists.

Phone and Contact Center add inherent telephony complexity (provisioning, E-911, carrier integration), but that's the category, not the product. Channel partnerships growing (9/10 large CC deals through partners, 50%+ of large Phone deals through partners) add deployment complexity but also demonstrate maturation.

Low friction means easy to adopt and easy to leave. The F penalty is small (−0.12) because friction taxes survival but doesn't determine it.

Regime Context

T = 15 weeks (2025-12-16 to 2026-03-27, 70 trading days)

FACTOR DECOMPOSITION (ZM ~ SPY + WCLD)
  α̂ = +26.1% ann   (t=0.33, p=0.745 — NOT significant)
  β_SPY  = 0.503
  β_WCLD = 0.459
  R²     = 0.197
  σ_idio = 40.6%    σ_total = 45.7%
  %Idio  = 80.3%    → PASS (>75%)
  IR     = 0.643    (unreliable — regime noise)

INTRA-SECTOR CORRELATION
  ρ_intra = 0.522   → INDISCRIMINATE SELLOFF regime
  Range: [0.187, 0.793]
  Trend: 0.659 → 0.438 (falling — differentiation emerging)

15-WEEK RETURNS
  TEAM  -59.1%    NOW   -35.0%    DOCU  -33.0%
  FIVN  -30.0%    CRM   -29.5%    WCLD  -24.4%
  TWLO  -10.4%    ZM    -10.2%    SPY    -6.3%

  ZM vs WCLD: +14.2%
  ZM vs SPY:   -3.9%

The sector is in a correlated selloff (ρ_intra = 0.522). When correlations are this high, IR measures the regime, not the stock. The α of +26.1% annualized is statistically zero (p = 0.745, μ̂ − 1.5×SE < 0). The IR of 0.643 is noise.

ZM outperformed WCLD by 14.2 percentage points over 15 weeks — massive relative outperformance while the cloud sector bled. This is the valuation floor: P/E 12.6 with $7.8B cash (≈34% of market cap) makes ZM a value name hiding in a growth index. When growth sells off indiscriminately, value provides a floor.

But defensive value in a selloff is not structural durability. V-Score is orthogonal to price: V(s) is scored against structural properties, not returns. ZM held up because it's cheap, not because it's durable.

The ρ_intra trend is falling (0.659 in weeks 1-7 to 0.438 in weeks 8-15). As correlations break down and the market returns to stock-picking, structural differentiation will reassert. AT_RISK names that held up on valuation alone will face the reckoning that the selloff temporarily masked.

Thermodynamic Summary

Intelligence flows to minimum energy. For synchronous communication (video/voice), cloud infrastructure remains the minimum-energy path — you cannot do multi-party real-time communication locally. This gives Zoom a floor.

But the floor is commodity. Teams runs on Azure. Meet runs on GCP. Both provide equivalent synchronous communication at lower marginal cost (bundled with Microsoft 365 / Google Workspace at near-zero incremental pricing). Zoom's 30 data centers compete against Microsoft's 60+ regions and Google's 40+ regions.

For asynchronous communication and AI-powered knowledge work, intelligence flows local. Meeting summaries, task extraction, document generation, email drafting — these are the expanding use cases, and they run on local LLMs. Every hour that shifts from synchronous meeting to asynchronous AI interaction is an hour that doesn't need Zoom's infrastructure.

Zoom's exposed surface is the application layer atop commodity infrastructure. Phone and Contact Center add genuine regulatory friction (FCC, E-911) but represent less than 20% of revenue. The explicit no-train policy on customer data (a privacy differentiator) also means no proprietary data flywheel — the very policy that builds trust prevents the accumulation that would build a moat.

The Anthropic stake is the darkest irony: Zoom's most valuable non-operational asset belongs to the entity best positioned to render Zoom's software layer unnecessary.

Stress Test: E and C

Both dimensions were challenged adversarially. Neither shifts the verdict.

E = 2 holds. Switching costs are finite and measurable (hours for Online, days for enterprise meetings, 3-6 months for Phone, 2-4 months for CC). Zoom's own competitive displacement wins prove UCaaS switching costs are surmountable. NDE at 98% is the definitive market signal — if infrastructure were irreducible, customers wouldn't be contracting.

C = 3 holds, under pressure. Real-time media engineering (codec, DC architecture) genuinely takes 1-2 years to re-derive, but is competitively irrelevant against hyperscalers with superior infrastructure. PSTN integration is regulatory/commercial, not cognitive (properly belongs in E). Application-layer cognition (65%+ of business) will erode toward C=2 as frontier models generate equivalent UX. Net trajectory: 3.0 to 2.5-2.7 over two years.

Sensitivity:

No reasonable perturbation reaches EMBEDDED (V ≥ 3.0). To get there requires upgrading four dimensions simultaneously — which means Phone/CC exceeding 50% of revenue, deep cross-domain cognition, genuine platform gravity, and agent routing. At current growth rates: 5-7 years. The score is robust.

Evidence Table

Cumulative LR: 8.78 (13 worldview evidence items). Geo mean: 1.182. Net direction: bullish on the business, bearish on AI survival. These are orthogonal assessments.

Conviction & Basket Verdict

V(ZM)  = 2.30
κ      = (2.30 − 3.0)⁺ = 0.00
w_i    ∝ κ_i = 0

IR     = 0.643  (regime noise, not signal)
ρ_intra = 0.522 (indiscriminate selloff)
δ      = N/A   (κ = 0 → no structural premium to restore)

Verdict: FILTER. Excluded from the SaaS survival basket. κ = 0 means zero weight regardless of regime.

ZM may be a perfectly fine value trade at P/E 12.6 with $7.8B cash and a potential Anthropic IPO catalyst. That's a different thesis with a different framework. For AI survival — the question of whether this software business structurally endures as intelligence gets cheaper — the answer is: the dominant revenue stream (Meetings, ≈65%) sits on commodity infrastructure with near-zero switching costs, and Zoom consumes AI intelligence rather than producing it. No single dimension, nor any reasonable combination of dimension improvements, moves the score to EMBEDDED within the measurement horizon.

The path to EMBEDDED exists in theory: Phone and Contact Center grow past 50% of revenue, AI monetization inflects, agent routing shifts from Teams/Meet to Zoom's infrastructure. At current growth rates, that's 5-7 years. The V-Score measures what IS, not what might be. Today, V = 2.30. FILTER.