Cognizant: A Good Price for a Shrinking Moat
Opening
Every so often the arithmetic and the business disagree, and the harder job is deciding which one to believe. Cognizant (CTSH) presents exactly that problem. The numbers are genuinely appealing: a P/E of 9.33, an owner earnings yield of 12.2%, essentially no debt, and a stock down 36% in 2026. The business, however, sits squarely where AI-driven disruption is aimed. We walked it through the four gates in order. Three were fine. The one that matters failed.
Gate 1: Circle of Competence
Cognizant makes money the old-fashioned IT services way: it supplies large volumes of skilled, lower-cost labor — largely India-based — to develop, run, and maintain the IT systems of global enterprises, mostly large American corporations, and captures the spread on labor costs under long-term contracts. It operates in four industry segments: Health Sciences, Financial Services, Products & Resources, and CMT. FY2025 revenue was roughly $21 billion. Aside from some IP assets — the healthcare SaaS platform TriZetto chief among them — the essence of the business is billing by the hour for skilled people. The company is now pushing an “AI Builder” strategy, consulting on and building AI adoption, with TTM bookings of $29.6 billion, up 11% year over year.
Is it knowable? The first layer, information, is comfortably yes: filings, earnings calls, and a well-mapped competitive set (Accenture, TCS, Infosys, Wipro). The second layer — predictability — is where we hedge. The question we ask is whether an explanation of this company ten years from now would be analysis or a probability bet. Here it is a bet. The unresolved proposition is whether AI code generation will net-shrink demand for labor-based IT services, or whether the demand for AI adoption itself more than compensates. Bullish evidence exists: Q1 2026 bookings up 21%, record large deals. So does bearish: revenue growth slowing to 5.8%, guidance below consensus, the stock down 36% this year. Neither camp has won, which places this at the edge of our circle — a business whose model’s survival is itself the unsettled variable.
Gate 2: Moat — the decisive failure
The moat’s source is identifiable: switching costs. Multi-year operations contracts embedded in client workflows, accumulated domain knowledge, and platform assets like TriZetto, which handles US health insurance claims processing. Morningstar rates it a narrow moat. That much is fine.
Pricing power is not. IT services is a competitive-bid industry. The 10.6% net margin matches or trails Accenture (~11–12%) and falls far short of the Indian peers (TCS ~19%). Can the company raise prices without a prayer meeting? No. If anything, industry-wide pressure runs the other way: to pass AI productivity gains on to customers as price cuts.
The direction is the real problem. The moat’s root is the mass supply of low-cost skilled cognitive labor — precisely the point AI code generation targets. Low-value areas like QA and maintenance already face automation pressure, and the time-and-materials billing model conflicts structurally with rising AI productivity. Switching costs shield existing contracts, but they cannot prevent deal-size deflation at each renewal. On our AI-era framework, this is a first-order erosion case: a business whose moat was scale of people, in a domain where AI collapses marginal cost. The AI-consulting tailwind is real but open to the whole industry — not a defensible advantage unique to Cognizant. What survives a leveling of AI models — TriZetto and regulated-industry domain knowledge — is only a partial defense of a $21 billion revenue base.
A specific, real moat, narrowing: Gate 2 fails.
Gate 3: Management (record kept, judgment moot)
With Gate 2 lost, the management gate is not needed, but the record deserves a note. Capital allocation has been reasonable: FY2024 dividends of $600 million plus $605 million in buybacks; $2 billion returned in FY2025. During the 2026 selloff, management raised the buyback target from $1 billion to $2 billion (with $3.45 billion in remaining authorization) — the right instinct, buying weakness. The Belcan acquisition (2024, $1.3 billion, aerospace engineering, mostly cash) was restrained, and there is no empire-building history. The retained-earnings test technically fails over five years, but with market price as the dominant factor we do not treat it as a standalone fault. CEO Ravi Kumar (since 2023) has acknowledged the growth slowdown while explaining the AI transition consistently. No red flags.
Gate 4: Price
Owner earnings are about $2,492 million — plausible for a light-capex services business where accounting and cash diverge little. At a $20.4 billion market cap, that is a 12.2% owner earnings yield, well above the ~4.3% ten-year Treasury. Our intrinsic value range: no-growth at 10x is $24.9 billion; low-growth at 12x is $29.9 billion — call it $25–30 billion, applying the lower end given AI erosion risk. Against the current price, the discount is roughly 18% on the no-growth case (short of our 30% requirement) and about 32% on the low-growth case.
On price alone, there is a borderline margin of safety. But the numerator is the issue: Gate 2 called into question the durability of those owner earnings over ten years. If AI deflation shrinks them, today’s price is not cheap. Bad, deteriorating business plus good price equals bad result — the Berkshire textile lesson, verbatim.
Verdict
Rejected — Gate 2 fails: the moat’s source is real but narrowing, and pricing power is absent. The quantitative profile (five-year average ROE of 16.6% with remarkably low volatility, near-zero leverage with $1.9 billion in cash, D/E around 0.04, P/E 9.3) is genuinely good, and the 36% drawdown is acknowledged. But we read the selloff not as a market overreaction to a temporary headwind, rather as the market asking a structural question — and we cannot answer “no” to whether the disruption could destroy the moat. The cheap price does not rescue the thesis; it merely prices the question.
What would change it: four consecutive quarters of 7%-plus year-over-year revenue growth together with adjusted operating margins sustained above 16% (guidance: 15.9–16.1%) — evidence that bookings convert to revenue and AI productivity converts to margin rather than price cuts. If those two appear, we will re-hear the moat.
This analysis is AI-generated, educational, and not investment advice. Figures may contain errors or be delayed. Disclaimer