Cognizant Technology Solutions Deep Dive

Cognizant Technology Solutions Corporation (NASDAQ: CTSH) - Deep Dive

Technology sector | Research date: 2026-10-31

Reporting period anchor. Cognizant's financial year ends on December 31 and it reports every quarter. Q3 results usually come out in the last week of October, so the Q3 2026 release was expected around October 29, 2026. The company did publish its Q3 2026 results and held its call on October 29, 2026. The six results events used in this report are:

• Q2 2025 (July 30, 2025)
• Q3 2025 (October 29, 2025)
• Q4/FY2025 (February 4, 2026)
• Q1 2026 (April 29, 2026)
• Q2 2026 (July 29, 2026)
• Q3 2026 (October 29, 2026)

1. What the Company Does

Cognizant is paid by large companies to build, run and fix the software their businesses depend on. Picture a US health insurer: it has to decide claims, pay doctors, enrol members and stay compliant with Medicare rules. A bank needs payment rails, trading systems and loan-origination software. A pharma company has to run clinical trials and file drug submissions. Most of these systems are decades old, deeply customised and impossible to switch off. Cognizant supplies the engineers who write new code for them, the teams who keep them running around the clock, and increasingly the operations staff who process the transactions that flow through them. Roughly three-quarters of its people sit in India. Most of its clients are in North America. The gap between what an engineer costs in Chennai or Pune and what the same work is worth to a New Jersey insurer is the original economic engine of the business.

Where it came from. The company began on January 26, 1994 in Chennai as Dun & Bradstreet Satyam Software, a joint venture between Dun & Bradstreet (76%) and Satyam Computers (24%). Its job was to do software work for Dun & Bradstreet's own businesses. Wikipedia credits Kumar Mahadeva and Srini Raju as founders, and Mahadeva ran the company until 2003. In 1996 Dun & Bradstreet spun the unit into a new entity, Cognizant Corporation. In 1997 it bought out Satyam's stake for US$3.4 million and renamed the software arm Cognizant Technology Solutions. The company listed in 1998 through an IPO run by its then-parent IMS Health. It became fully independent in 2003, when IMS Health distributed its remaining 56% stake.

Three later decisions explain most of today's business:

1. Building a US-facing, India-delivered services firm, but with a US headquarters and client-facing staff close to the client. That positioning helped it win share from Indian-headquartered peers through the 2000s and into the S&P 500 by November 2006. It is also why the company is now weighing a listing in India (Section 7).
2. Buying TriZetto in September 2014 for US$2.7 billion. TriZetto makes the claims-administration software that a large part of the US health-insurance industry runs on. The deal turned Cognizant from a pure labour provider in healthcare into a software owner with a captive implementation and maintenance stream, and it is the reason Health Sciences has been its largest segment for years.
3. The 2019-2023 stall and the 2023 leadership change. Under Brian Humphries (CEO April 2019 to January 2023) Cognizant grew slower than its peers and lost share. Ravi Kumar S took over in January 2023. Per the 2026 proxy statement he spent 2002 to 2022 at Infosys, the last six years as President. He came with an explicit goal of getting Cognizant back to "Winner's Circle" growth, meaning top-tier among peers, by 2027. Since 2024 he has used acquisitions to add specialised capabilities: Belcan (aerospace and defence engineering, 2024), 3Cloud (Microsoft Azure, January 2026) and Astreya (AI-infrastructure managed services, June 2026). He has also rebranded the strategy around Cognizant as an "AI builder."

The value proposition. Cognizant solves one problem for large enterprises. They own vast, fragile, business-critical software estates and cannot afford to staff the people who understand them at onshore wages. It sells three things:

• Engineering capacity at offshore cost, which is the original pitch.
• Domain knowledge in a few regulated industries, mainly US healthcare, banking and insurance, where knowing the rulebook matters as much as knowing the code.
• Accountability: an increasing share of work is sold as a fixed-price outcome rather than hours billed (Section 4).

What makes it hard to replicate. The product looks like labour, but it is really a system for recruiting, training, deploying and managing more than 350,000 people. That includes the client-specific knowledge those people accumulate about systems nobody else understands. The details are in Section 3.

A concrete example. In Q3 2026 Cognizant announced a five-year extension with Gilead Sciences covering DevOps and agentic AI support. In practice this means a Cognizant team, mostly offshore with a smaller group near the client, runs the pipelines that build, test and release Gilead's internal software. It monitors and patches the applications, and it now deploys AI agents that handle parts of that work, such as triaging incidents or generating test code. Under a fixed-price structure, Cognizant agrees a price for keeping a defined set of applications running to agreed service levels. Any productivity it gets from automation shows up first in its own cost line, and some of it is passed back to the client over the life of the contract. That arrangement is the core of the business today and the core of the AI debate around it.

2. Business Segments

Cognizant reports four segments. They are defined by client industry, not by type of service: the same engineering, cloud and operations practices serve every segment. The segments are therefore sales and domain-knowledge units sitting on one shared delivery machine. Mix below is for FY2025 (10-K) unless noted.

The mix is moving. Financial Services overtook Health Sciences as the largest segment during 2026, after three straight quarters of double-digit growth (Q3 2026 release).

Health Sciences (30.1%)

What it does. It serves US health insurers (payers), hospital and provider groups, and life-sciences companies (pharma, biotech, medical devices). For payers it implements and runs claims administration, member enrolment, provider networks and care-management systems, and it increasingly operates those processes itself as outsourced business processes. For life-sciences clients it supports clinical-trial systems, regulatory submissions, pharmacovigilance and commercial operations.

Core capability. It owns TriZetto's payer platforms, the core administration systems a large share of US payers run on (products in Section 3). Owning the platform gives Cognizant implementation work, upgrades, hosting, maintenance and process outsourcing on top of the software. Management told the Q2 2026 call that TriZetto grows faster than the company, at substantially higher margins, and is "the blueprint" for the platform-led model it wants in other industries.

Why it is separate. US healthcare runs on its own regulation (Medicare and Medicaid rules, HIPAA, CMS mandates) and its own buyer, the payer CIO and operations chief. Its economics also differ because it includes software licences and third-party product resale.

Role in the group. The domain moat and the most defensible franchise, but also the segment most exposed to US government healthcare funding. Health Sciences shrank slightly in Q1 and Q3 2026. Management attributed Q1 to lower third-party product sales and Q3 to "regulatory uncertainty around Medicaid and Medicare funding" (Q3 2026 call).

Main rivals: Accenture, Optum (UnitedHealth's services arm), Infosys, HCLTech, NTT Data and payer-platform vendors such as HealthEdge.

Financial Services (29.2%)

What it does. Banking (retail, commercial, payments), capital markets and insurance. Work covers core-banking and policy-administration modernisation, regulatory reporting, cyber, cloud migration, and outsourced processes such as claims handling, onboarding and reconciliation. Named recent work includes Axis Bank (application management) and the European Bank for Reconstruction and Development (a five-year technology transformation) (Q3 2026 release).

Core capability. Long-standing embedded teams at large North American banks and insurers. Those teams carry undocumented knowledge of the client's systems, which is what makes renewals sticky.

Why it is separate. Distinct regulation, a distinct buyer (bank CIO and COO organisations), and the highest and earliest AI spending of any vertical.

Role in the group. The growth engine right now. Management calls it the leading indicator: "Financial services is well ahead with AI initiatives and advanced AI adoption" (Q2 2026 call). The company expects other industries to follow (Section 7).

Main rivals: TCS, Infosys, Accenture, Capgemini, HCLTech, Genpact (on processes).

Products & Resources (25.0%)

What it does. Manufacturing, retail and consumer goods, travel and hospitality, energy and utilities, logistics, and through Belcan, aerospace, defence and government engineering. Work spans ERP and supply-chain systems, e-commerce platforms, plant-floor and connected-product engineering, and engineering services for aircraft and defence programmes.

Core capability. A mix of enterprise-application implementation (SAP, Oracle, Salesforce) and physical-product engineering. Belcan adds engineering of hardware and systems, not just software, and access to US government programmes. Management cited a position on the Missile Defense Agency's SHIELD programme, an indefinite-delivery contract with a US$150 billion ceiling across many awardees (Q4 2025 call).

Why it is separate. Clients' own revenue is cyclical and tied to consumer demand and trade policy, so this segment's discretionary spend moves with the economy.

Role in the group. The cyclical swing factor. It shrank slightly in Q3 2026, with management citing retail and hospitality macro pressure (Q3 2026 call) and earlier trade-policy shifts (Q1 2026 call). Belcan gives a counter-cyclical defence leg.

Main rivals: Accenture, Capgemini, HCLTech, TCS, LTIMindtree, and on engineering, Cyient and Quest Global.

Communications, Media & Technology (15.7%)

What it does. Serves telecom operators, media and entertainment companies, education providers and technology companies, including hyperscale and software firms. Since June 2026 it includes Astreya, which manages IT and data-centre operations for large technology companies.

Core capability. Network and OSS/BSS systems for telecoms, digital-content platforms, and now AI-infrastructure operations through Astreya.

Why it is separate. Technology clients are both customers and competitors. They have strong in-house engineering and buy differently: often staff augmentation, trust and safety, or data-labelling work rather than outsourced application estates.

Role in the group. Small but growing fast in 2026, mostly from acquisitions (Q3 2026 call). Management flagged earlier that "a strategic shift at a major Communications and Media sector customer" was weighing on the segment (Q1 2026 call).

Main rivals: Accenture, TCS, Infosys, Tech Mahindra (telecom), TaskUs and Genpact (tech-client operations).

Geographic mix

Cognizant is overwhelmingly a North American business. FY2025 revenue mix (10-K) was:

• North America: 74.8%
• Continental Europe: 9.9%
• United Kingdom: 9.1%
• Rest of World: 6.2%

North America has been the core since the 1990s and is where TriZetto, the large US banks and the payer franchise sit. It was also the fastest-growing region in Q3 2026. The UK has been a market for decades, mostly in financial services and public sector: the multi-year HM Revenue & Customs contract cited on the Q2 2026 call is worth more than US$250 million over its life. Continental Europe has been weaker, shrinking in Q3 2026 (Q3 2026 release). Rest of World is mainly Asia-Pacific and the Middle East, plus a growing Indian domestic book (Axis Bank). The pending DyFlex acquisition (Section 7) is aimed at Australia and New Zealand.

The heavy US weighting makes Cognizant more exposed to US health policy and US immigration rules than its European-diversified peers (Section 8). The supply side runs the other way: at end-2025, 256,900 of 351,600 employees (73%) were in India, with 41,600 in North America (10-K).

3. Products and Operations

What it sells

Cognizant organises delivery into seven practices (10-K): Core Technologies and Insights, Enterprise Platform Services, Industry Solutions, Intuitive Operations and Automation, Software and Platform Engineering, Cognizant Moment (digital experience), and Security (set up as a stand-alone practice in 2026). In plain terms, the product lines that matter to the investment case are below.

1. Application development and maintenance (the core annuity). Building, enhancing, testing and running clients' custom and packaged applications. This is the oldest and largest line and the one most exposed to AI-driven productivity. Management says more than 50% of its software development is now "AI-assisted," up from about 30% in Q3 2025 and over 40% in Q2 2026 (Q3 2025, Q2 2026 and Q3 2026 calls).

2. Enterprise platform services. Implementing and running SAP, Oracle, Salesforce, ServiceNow, Workday and Microsoft platforms. The 3Cloud acquisition (closed January 1, 2026, US$733 million per the Q3 2026 10-Q) added more than 1,200 Azure specialists (Q4 2025 call). The pending DyFlex deal adds an SAP Platinum partner in Australia and New Zealand.

3. Cloud, infrastructure and data-centre operations. Managing clients' hybrid cloud, networks, end-user computing and, since Astreya (closed June 22, 2026, US$634 million per the Q2 2026 call), the physical and IT operations of AI data centres for large technology companies.

4. Business process outsourcing (BPO / BPaaS). Running whole processes: insurance claims, healthcare enrolment and billing, finance and accounting, trust and safety. Management says BPO was heading towards about US$3 billion of annualised revenue and was the fastest-growing offering in healthcare (Q3 2025 call). "BPaaS" means the process delivered on Cognizant-owned software, which is how TriZetto-based payer operations work.

5. Engineering and research services. Product and embedded-software engineering for manufacturers and medical-device companies, plus Belcan's aerospace and defence engineering.

6. Owned software and AI platforms

• TriZetto: the payer-administration suite, Facets and QNXT, which are core claims and benefit-administration systems; it is the only large owned software franchise.
• Neuro: an orchestration layer for running AI and automation across a client's estate.
• Flowsource: an AI-assisted software-engineering platform. Management said it was live at more than 70 clients with 120 in the pipeline (Q3 2025 call).
• Agent Foundry: industry-specific libraries of AI agents. The company says it has built more than 1,500 agents.
• AI Nexus: launched in Q3 2026, described as a layer to "govern AI quality, manage spending, and connect consumption to business outcomes."

None of these, apart from TriZetto, is a stand-alone software business. They are delivery accelerators that make Cognizant's labour more productive and its pitch more distinctive.

7. AI data and training services. Cognizant runs a pool of more than 10,000 specialists who label and evaluate data and train models for AI developers and enterprises (Q3 2025 call).

How it is delivered

The delivery model is a global network, but the gravity sits in India. The company runs large campuses in Chennai (Siruseri, where it is building a 14-acre immersive learning centre), Pune, Hyderabad, Bengaluru, Kolkata, Kochi and Coimbatore. Since 2024 it has opened newer centres in Bhubaneswar, Indore, GIFT City in Gujarat (a fintech centre) and Visakhapatnam, which was announced June 2025. It also has nearshore and onshore centres in the US, UK, Continental Europe, the Philippines and Latin America.

A typical engagement runs like this:

1. Pursue. A client-facing team of account managers, domain consultants and solution architects, mostly onshore, shapes the deal.
2. Transition. For outsourcing deals, a three-to-nine-month phase in which Cognizant staff shadow the incumbent or in-house team, absorb system knowledge, and move work offshore. Cognizant bears cost here before revenue arrives. This is why management talks of large deals moving from "cost-incurring to revenue-accruing phases" (Q1 2026 call).
3. Steady state. Mostly offshore teams run the work against service levels, with a small onshore layer. Productivity tools such as Flowsource, Neuro and agents reduce effort over the contract life.
4. Expand. Embedded teams spot new projects and upsell.

Labour is the inventory. Utilisation held at about 85% through 2025 (Q3 2025 call). Voluntary attrition in technology services fell from 15.2% in Q2 2025 to 12.6% trailing twelve months at September 30, 2026 (Q2 2025 call, Q3 2026 release). Headcount was 359,900 at September 30, 2026, up 10,100 in a year (Q3 2026 release). The company hired about 20,000 college graduates in India in 2025 and planned about 20,000 in India plus 2,000 in the US for 2026 (Q4 2025 call). By Q3 2026 it was citing 1,500 US graduates.

The pyramid is being redesigned. The classic model hires many juniors to do routine coding and testing under a few seniors, and that is what AI tools now do best. Management says it is abandoning the traditional pyramid for interdisciplinary teams that pair domain experts with AI-fluent engineers (Q1 2026 call). It is creating two new certified roles:

• "Frontier Certified Engineers," who build AI systems.
• "Frontier Business Operators," who manage mixed teams of people and agents.

It targets 15,000 people across the two roles (Q3 2026 release). More than 45,000 associates had earned 95,000+ "frontier AI" certifications by Q3 2026.

Partner capability groups. Cognizant increasingly organises large, partner-specific benches:

• An 18,000-person Gemini Frontier Capability Group with Google Cloud, focused on legacy modernisation and cyber defence (Q3 2026 call).
• A delivery group of more than 25,000 professionals dedicated to OpenAI technologies, announced October 30, 2026.
• A collaboration with Anthropic to deploy Claude models and agentic tooling internally and with clients, first disclosed on the Q3 2025 call. A Travelport engagement built on it was expanded in Q3 2026.
• An OpenAI Codex hackathon with more than 2,600 participants (Q3 2026 release), and a July 2025 "vibe coding" hackathon that drew more than 53,000 employees across 40 countries (Q3 2025 call).

Recent operational events:

• Project Leap. Announced with Q1 2026 results on April 29, 2026. It is a restructuring that cuts layers, closes facilities and retires software, with savings redirected into AI capabilities, partnerships and reskilling. The estimated cost was lowered on the Q3 2026 call because more staff were redeployed rather than let go. It now runs into Q1 2027.
• Furloughs. Clients' year-end shutdowns, standard in Q4 for this industry, are expected to be heavier than usual in Q4 2026 (Q3 2026 call).

A simple value-chain view for the chart-generator:

Client business need → onshore consulting and solution design → transition (knowledge capture from the incumbent) → offshore delivery centres (India: about 73% of staff) running build / run / operate → AI accelerators (Flowsource, Neuro, Agent Foundry, AI Nexus) compressing effort → steady-state service-level delivery → expansion into adjacent processes (BPO, platforms).

4. Customers

Who buys. Cognizant's clients are large enterprises, mostly Global 2000 companies. Named public accounts from the last six quarters include:

• Health and life sciences: Gilead Sciences; Benchling and Kyowa Kirin (R&D platforms).
• Financial and public sector: Axis Bank; the European Bank for Reconstruction and Development; HM Revenue & Customs.
• Travel, payments and logistics: Travelport; SITA (the airline IT cooperative); Ingenico (payments terminals); Odyssey Logistics.
• Defence: US defence programmes through Belcan.

The 10-K does not disclose the share of revenue from top clients. It says only that services "are distributed among a number of clients in each of our reportable business segments" and that losing "a significant client or a few significant clients in a particular segment could materially reduce revenues for that segment." No client is singled out as material at group level. Concentration risk shows up in segments instead: one large media client's strategic change visibly hurt CMT in early 2026 (Q1 2026 call).

Who decides, and how. There are three distinct buying motions.

• Large outsourcing and managed-services deals (application management, infrastructure, BPO; often US$100 million+ total contract value, multi-year). The CIO and CFO, often run through a procurement team and frequently advised by a sourcing consultant (ISG, Everest, EY). The main criteria are guaranteed savings over the incumbent, transition risk, domain credentials and references. Sales cycles run 6 to 18 months, plus a transition period. Cognizant signed 28 such deals in 2025, including five "mega" deals above US$500 million, one worth about US$1 billion (Q4 2025 call), then seven in Q2 2026 and three in Q3 2026.
• Discretionary transformation projects (a cloud migration, a new digital channel, an AI pilot turned into production). Business-unit heads plus the CIO. Criteria are speed, specialist skills and proof of value. Cycles run weeks to a few months. This is the spending that froze in 2023-2026, and management keeps saying it is waiting for it to recover (Section 9).
• Platform-anchored healthcare work (TriZetto implementations, upgrades, payer process outsourcing). The payer's operations and IT leadership. Since the client already runs Cognizant's software, choosing Cognizant to implement, upgrade or operate it is often the default.

Why they choose Cognizant

1. Incumbency and system knowledge. Teams embedded for years know undocumented logic in client systems that no competitor can price confidently.
2. Healthcare domain depth through owning TriZetto.
3. A US-centric client-facing presence with Indian delivery costs.
4. Willingness to take fixed-price risk. In management's telling, also willingness to share AI productivity gains up front in exchange for scope (Q1 2026 call).
5. Partner standing with hyperscalers and model vendors (Section 3).

Switching costs. The 10-K says it directly: "The services we provide to our larger clients are often critical to their operations and termination of our services would typically require an extended transition period with gradually declining revenues." Leaving Cognizant means:

• Running a competitive bid.
• Paying for a reverse transition while two vendors overlap.
• Accepting a period of higher operational risk on systems that process claims, payments or orders.

For TriZetto payers, switching the platform itself is a multi-year core-system replacement that few attempt. The switching cost is real but not absolute. Clients do re-tender at renewal and ask for price cuts, and AI is giving them a new reason to (Section 8).

Contract structure. Revenue mix by contract type for the first nine months of 2026 (Q3 2026 10-Q):

Fixed-price was about 43% in 2024 and about 47% in 2025 (Q3 2025 call; 10-K), so the shift is steady.

• Fixed-price contracts are mostly multi-year managed services with annual productivity commitments. The longer the contract, the better the revenue visibility. The catch is that Cognizant carries execution risk and has to deliver promised productivity.
• Time-and-materials work is more discretionary and can be cut at short notice.
• Transaction-based pricing is mostly BPO and platform-based work.

Management is now introducing "tokenized rate cards," pricing that runs from human-led work to fully autonomous "digital labour" (Q1 2026 call). It amounts to an attempt to charge for outcomes and machine output rather than headcount.

What this means for predictability. Book-to-bill of about 1.3x on trailing-twelve-month bookings (Q3 2026 release) and a growing fixed-price share make the base fairly predictable. Quarter-to-quarter variation comes from the discretionary time-and-materials layer and from client furloughs.

5. Competitive Landscape and Moat

The structure of the industry. Enterprise IT services is fragmented at the bottom and oligopolistic at the top. For large, multi-year outsourcing deals at Global 2000 clients, the real shortlist is usually five to eight firms:

• Accenture, which wins on consulting and strategy.
• Five Indian-heritage firms: TCS, Infosys, HCLTech, Wipro and Cognizant, which win on offshore scale and price.
• Capgemini, the European anchor.
• IBM Consulting, Deloitte and NTT Data in specific niches.

Below them sits a long tail of specialists (EPAM, Globant, Endava in engineering; Genpact, EXL and WNS in processes) and thousands of boutiques. The 10-K names Accenture, IBM Consulting, Infosys, TCS, Wipro, Capgemini and Deloitte Digital among its competitors.

Market caps are peer-size references from CompaniesMarketCap, October 2026.

How Cognizant wins and loses against each:

• Against Accenture, it wins on cost and on healthcare depth, mainly in payer operations, where owning TriZetto beats Accenture's partner-led approach. It loses where the client wants a strategy-led transformation or a CEO-level relationship. Accenture's spending on acquisitions and AI dwarfs Cognizant's.
• Against TCS and Infosys, the contest is closest, since the delivery models are nearly the same. Cognizant's edges are a US-based leadership and sales culture, which helps in US mid-market and healthcare accounts, and TriZetto. Its weaknesses are smaller scale in Europe and, historically, slower adoption of large-deal pricing. Under Ravi Kumar it has been taking share in large deals: management cites total contract value of large deals up nearly 50% in 2025 (Q4 2025 call). Industry trackers also listed Cognizant among the firms gaining contracts in 2026 while Infosys, TCS and Wipro faced client changes (InfotechLead, 2026). These rivals have also had leadership changes and slower growth.
• Against HCLTech, it wins in applications and healthcare and loses in infrastructure-heavy deals. The Astreya and 3Cloud acquisitions aim partly at narrowing that gap.
• Against Capgemini, it wins in North America and loses in Continental Europe, where Capgemini is entrenched and Cognizant's growth is weakest.
• Against Wipro, it has been a consistent share taker.
• Against EPAM and Globant, it competes on price for engineering work. They tend to win high-end greenfield product engineering, Cognizant wins scale programmes.
• Against in-house AI and the model labs themselves. This is the new competitor (Sections 6 and 8). Cognizant's response is to partner, through the Google, OpenAI and Anthropic groups, rather than to compete.

Barriers to entry. Entering the US$100 million-deal tier is hard. A new firm needs:

• Tens of thousands of trained engineers.
• A delivery footprint that passes client security and regulatory audits (HIPAA, SOC 2, banking-regulator reviews).
• A balance sheet that can carry fixed-price and transition risk.
• Years of reference clients.

No new firm has broken into the top tier in two decades. The tier's membership has stayed the same while share has moved within it. Entry into small, discretionary projects is easy, which is why that layer is where pricing pressure and AI-native boutiques appear first.

Market share. No single firm holds a large share of the roughly US$1.87 trillion IT-services market (Gartner, July 2026, a figure that includes IaaS). Share within the top tier moves slowly, through renewals and large-deal wins. Cognizant lost share between 2019 and 2023 and has been regaining it since 2024: it grew 6.4% in constant currency in 2025, beating the top end of its own guidance (Q4 2025 release).

Structural shifts:

1. Generative and agentic AI shrinks the hours needed for coding, testing and support. This is deflationary for the time-and-materials layer and a pricing fight at every fixed-price renewal (Section 8).
2. Vendor consolidation. Large clients cut their supplier lists to two or three strategic partners, which favours scale players. Morningstar has cited this as a tailwind for Cognizant.
3. Model labs and hyperscalers are building partner channels and need integrators to reach enterprises, as Cognizant's partner groups show.
4. The scale players are buying specialists: Accenture's Faculty deal, Capgemini's WNS deal, Cognizant's Belcan, 3Cloud, Astreya and DyFlex deals.

Moat: Narrow - switching costs, intangible assets (TriZetto domain franchise)

The evidence is the extended-transition language in the 10-K and the multi-year managed-services base described in Section 4. TriZetto adds a platform position in US payer administration that no peer has. But the moat did not stop the 2019-2023 share loss to peers with the same model, which is why it is at most Narrow. The thing most likely to erode it is AI compressing the effort behind maintenance and support work: once clients believe a contract renegotiation can capture that productivity, incumbency protects the relationship but not the price.

6. Industry

What drives demand. Enterprise IT-services demand has three layers, each with a different driver:

1. Run spending (maintaining and operating systems) follows the size and complexity of the installed base. It is fairly stable and gets outsourced more over time as clients look for savings.
2. Change spending (projects, modernisation, new channels) follows business confidence, interest rates and the technology cycle. It is the volatile layer.
3. Operations spending (BPO) follows clients' transaction volumes and their willingness to hand over whole processes.

Technology cycles change the content of the work: mainframe to client-server, then internet, ERP, cloud, and now AI. Each cycle has eventually expanded the market by creating a new modernisation wave.

Size and growth. Gartner's July 2026 forecast puts worldwide IT spending at US$6.37 trillion in 2026, up 14.2%. IT services is the largest segment at more than US$1.87 trillion, but grows only about 5.3%. Data-centre systems grow about 55.8% as AI infrastructure absorbs the incremental budget (Gartner, July 2026). The gap matters: enterprise technology budgets are rising, but the extra money is going to chips, servers and software, not people. Gartner's longer-range work projects that server spending will exceed all IT-services spending by 2030 (Software Strategies blog, October 2026). Cognizant's management frames a larger prize: a "$1 trillion system integration market" opening into a "$5-6 trillion enterprise operations market" as AI agents do work humans now do (Q2 2026 call). The company's own research says AI could unlock US$4.5 trillion of US labour value (Q4 2025 call). Those are vendor estimates, not independent ones.

Where Cognizant sits in the supply chain. Between the technology vendors and the enterprise:

• Upstream: hyperscalers (AWS, Azure, Google Cloud), model labs (OpenAI, Anthropic, Google), enterprise-software vendors (SAP, Salesforce, ServiceNow) and chip and infrastructure makers.
• Downstream: the enterprise that has to make all of it work inside its own processes and data.

The integrator's value is the "last mile." Cognizant's Chief AI Officer put it this way in February 2026: "That mapping is our job, it does not come just automatically out of the box" (Babak Hodjat, Yahoo Finance/Reuters, February 26, 2026).

Regulation. Four regulatory forces shape this industry:

1. US immigration rules, especially H-1B visa costs and caps, set the cost and feasibility of putting Indian engineers onsite. The US added a new H-1B fee in 2025, which management flagged as a cost item for 2025-2026 planning.
2. Data-protection and sector rules (HIPAA, GDPR, banking outsourcing guidelines) dictate where data can be processed and how offshore centres are audited.
3. US healthcare funding policy (Medicaid and Medicare) changes payer IT budgets directly.
4. Indian tax and SEZ policy affects delivery costs. The 10-Q tracks Indian tax disputes before the Income Tax Appellate Tribunal and the Supreme Court of India.

Cyclicality. IT services is mildly cyclical. Run spending held up in 2008-09 and 2020. Change spending froze, recovered sharply in 2021-22, and has been subdued again since 2023 under high interest rates, tariffs and AI uncertainty. Large outsourcing deals are counter-cyclical to a degree, because clients outsource more when they are cutting costs.

Tailwinds:

• Enterprise AI adoption needs integration, data cleanup, governance and change management. That is labour-intensive work that has not yet been automated.
• Legacy modernisation (mainframe, old ERP) becomes cheaper and faster with AI, which may release deferred projects.
• Vendor consolidation favours scale players.
• Defence and public-sector technology spending is rising.

Headwinds:

• AI-driven deflation of hours in coding, testing and support.
• Clients pausing AI deployments while they work out token costs and productivity: "One in four have paused AI deployments," Ravi Kumar said of the Global 2000 (Q2 2026 call).
• Budgets moving from services to infrastructure.
• Investor and client doubt about labour-based models. Anthropic's product launches in February 2026 triggered a sell-off in Indian IT stocks, and by July 2026 the combined market value of India's five largest IT firms had fallen more than 46% from its peak (Business Today, July 4, 2026), before a partial rebound in October.

7. Growth Triggers

Ranked by how much each could change the business.

• Converting AI from a deflationary threat into new scope through the "AI builder" model. Management's framework, repeated in some form on every call since Q2 2025, has three lanes:

  • "traditional work done more productively";
  • "old things in new ways" (AI-accelerated mainframe and SAP modernisation);
  • "entirely new things" (context engineering, physical AI, agent operations) (Q2 2026 call).

  Tokenized rate cards will price human and digital labour together (Q1 2026 call). AI Nexus is the new layer meant to make AI spending governable for clients (Q3 2026 call). Management says the "AI tailwind will spread from financial services to other industries" as discretionary spending recovers in 2027 (Q3 2026 call).

• Partner-specific delivery benches at frontier-model scale. An 18,000-person Gemini Frontier Capability Group with Google Cloud focused on legacy modernisation and cyber defence (Q3 2026 call, October 29, 2026). A delivery group of more than 25,000 professionals dedicated to OpenAI technologies (announcement, October 30, 2026). A collaboration with Anthropic for Claude models and agentic tooling, first disclosed on the Q3 2025 call (October 29, 2025) and expanded through Travelport (Q3 2026 release). Repeated across three calls.

• Large-deal and new-business momentum carrying into 2027. New and expansion work rose to 60% of bookings, more than 10 points higher than a year earlier, and grew 16% year-to-date (Q3 2026 call). Management positions 2027 as a potential inflection, contingent on geopolitical stability and clarity on healthcare funding (Q3 2026 call). Large deals signed in Q4 2025 and Q1 2026 were expected to reach full revenue run-rate from mid-2026 (Q1 2026 call). The large-deal story has been repeated on all six calls.

• Replicating the TriZetto platform model in other industries. Management presents TriZetto as the template for platform-led growth beyond healthcare (Q2 2026 call). Separately, BPO and BPaaS are described as a "multiyear tailwind" as clients rebuild operations around AI-driven automation (Q4 2025 call, February 4, 2026; also Q3 2025 call).

• Project Leap savings fully in the run-rate in 2027. About US$200 million of savings in 2026, with the full-year benefit expected in 2027 (Q2 2026 and Q3 2026 calls). Two-thirds is earmarked for reinvestment in growth, AI and partnerships, one-third for reskilling (Q1 2026 call). Execution continues to Q1 2027.

• Acquired capabilities ramping: AI infrastructure, Azure, SAP and defence.

  • Astreya brings AI data-centre and IT managed services, with "momentum through [the] existing Google partnership" (Q2 2026 call).
  • 3Cloud is a Microsoft Azure partner (Q4 2025 call).
  • DyFlex Solutions is an SAP Platinum partner in Australia and New Zealand under a definitive agreement (Q3 2026 call).
  • Belcan holds a position on the Missile Defense Agency's SHIELD contract vehicle (Q4 2025 call).

• New market: mid-sized companies. Cognizant Activate is a new business unit selling enterprise-grade AI transformation to mid-market and "emerging" enterprises, a client tier Cognizant has historically under-served (Q3 2026 call).

• A possible listing in India. The company is evaluating a primary offering and secondary listing in India, potentially through Indian Depository Receipts. It described the plan as "still in its early phase" (Q3 2025 call). Later: "We will make a decision once we have visibility of the revised regulatory framework" (Jatin Dalal, Q2 2026 call). On the Q3 2026 call, management said draft regulations were expected "in a few weeks." Repeated on three calls.

8. Key Risks

8.1 AI deflation outruns AI-created demand. Mechanism: Application maintenance, testing and tier-one support are where AI tools already handle much of the work. Cognizant reports more than half of its own development is AI-assisted (Section 3). Clients know this, and at every fixed-price renewal they ask for the productivity to be handed back as a lower price or smaller scope. If the new AI work (agent building, data engineering, modernisation) arrives more slowly than legacy hours disappear, revenue per client shrinks even when Cognizant keeps the account. The 2026 pattern fits this risk: bookings held up, yet growth guidance was cut twice. Clients are signing, but signing for less work per dollar. Reports of planned cuts of 12,000 to 15,000 roles under Project Leap (Startup Fortune and others, 2026) show the industry is absorbing its own disruption. Management has not confirmed that number, and headcount actually rose in Q3 2026. Calibration: high probability, moderate-to-severe drag. This is the central risk of the investment case.

8.2 Discretionary spending stays frozen, and the 2027 recovery slips again. Mechanism: Management has assumed some discretionary recovery in each of the last several guidance cycles (Section 9), and it has not come. The Q3 2026 outlook adds one more layer of caution:

"[Guidance] assumes higher impact from furloughs in the fourth quarter than previously expected." - Jatin Dalal, CFO (Q3 2026 call, October 29, 2026)

If tariffs, geopolitics or AI uncertainty keep clients in "pause and evaluate" mode through 2027, the large-deal base grows but the high-margin project layer does not. Without acquisitions, organic growth would sit in the low single digits. Calibration: medium probability, moderate drag on growth. Margins are cushioned by Project Leap.

8.3 US healthcare funding cuts hit the largest franchise. Mechanism: Health Sciences (Section 2) depends on US payers whose economics are tied to Medicaid and Medicare funding. Spending cuts, eligibility changes or rate pressure lead payers to freeze discretionary IT, re-bid outsourcing contracts on price, or consolidate. TriZetto clients rarely leave, but they can delay upgrades and squeeze maintenance fees. Health Sciences already shrank in Q1 and Q3 2026, and management named Medicaid and Medicare uncertainty explicitly (Q3 2026 call). Calibration: medium probability, moderate impact, concentrated in the segment that carries the platform moat.

8.4 US immigration and protectionism. Mechanism: About 73% of staff are in India and about 75% of revenue is in North America (Section 2), so the model depends on moving some Indian engineers onsite and on US clients being allowed to send work offshore. Higher H-1B fees (in place from 2025), tighter visa caps, proposed taxes on outsourcing payments, or rules on offshore processing of healthcare and financial data would push Cognizant towards more expensive onshore hiring or local subcontracting. Its high US weighting makes it more exposed than peers with European revenue. Calibration: medium probability of further tightening, low-to-moderate impact if gradual, severe if a direct outsourcing tax passed.

8.5 Fixed-price execution risk rises as the mix shifts. Mechanism: Fixed-price work is now nearly half of revenue (Section 4), and management is pricing AI productivity into bids up front, sometimes before it has been delivered. Overrunning a transition, missing a productivity commitment, or mispricing "digital labour" on tokenized rate cards falls straight onto Cognizant's costs. Mega-deals concentrate the risk: one troubled US$1 billion programme can absorb a year of margin improvement. Calibration: low-to-medium probability of a material write-down. Cognizant has no recent history of one, but the exposure grows each year.

8.6 The acquisition push and partner dependence dilute focus. Mechanism: In 2026 alone the company has spent about US$1.3-1.4 billion on 3Cloud and Astreya and has DyFlex pending (Section 3). Acquired specialist firms often lose key people after earn-outs end. The capability groups also tie large benches to the roadmaps of Google, OpenAI and Anthropic, which are building their own enterprise services teams and may decide to keep more of the integration work for themselves. Calibration: medium probability of under-delivery, moderate impact. This is a drag, not a threat to the business.

8.7 Legal exposure around TriZetto intellectual property. Mechanism: TriZetto (Cognizant) is suing Infosys in Texas federal court, alleging Infosys took TriZetto software and code to build competing payer tools. The case passed a motion to dismiss in August 2025 and was in depositions in June 2026 (Bloomberg Law, Law360). Infosys has filed antitrust counterclaims, which a court told it to rework (Bloomberg Law). A separate TriZetto trade-secret verdict against Syntel was upheld in March 2026 but with a reduced award (Law360 coverage). If the counterclaims succeed or discovery exposes weak protection of the platform, the moat underpinning Health Sciences weakens. Calibration: low probability, moderate impact. Noted because TriZetto is the one asset that differentiates Cognizant.

9. Walk the Talk

Concalls used:

• Q2 2025 (July 30, 2025)
• Q3 2025 (October 29, 2025)
• Q4/FY2025 (February 4, 2026)
• Q1 2026 (April 29, 2026)
• Q2 2026 (July 29, 2026)
• Q3 2026 (October 29, 2026)

The most recent is two days before this report.

The commitments that tell you something:

The 2026 revenue guide is the story of the year. In February management set a wide range that assumed things would get better in the second half. It held the range in April, cut the top in July, and in October cut the top again, ending close to the original floor. The bookings data did not deteriorate at the same pace: book-to-bill stayed about 1.3x and new business grew. The gap between signing and revenue is the same pattern Section 8.1 describes.

"Management expressed confidence that AI tailwind will spread from financial services to other industries as discretionary spending recovers in 2027." - Ravi Kumar, CEO (Q3 2026 call, October 29, 2026, as summarised in the transcript)

The recovery that was meant to arrive in late 2026 is now being described for 2027. That is the second time the recovery has been pushed back, so the 2027 claim needs to be checked against the Q4 2026 and Q1 2027 calls.

Margin and capital return have been reliable and conservative. Every margin guide since the 2025 Investor Day was met or raised. Capital return beat its plan by more than a third, mainly because the company sped up buybacks when the stock fell in the second quarter (Section 10). On the levers it directly controls (costs, pricing discipline, capital allocation), management has under-promised.

Project Leap is a quiet win that came with some sleight of hand. The lower cost estimate came with savings at the bottom of the range and a one-quarter extension. Management explained the lower severance by pointing to staff being reskilled and redeployed rather than dismissed. That is consistent with headcount rising 3,200 in Q3 2026. It also means less of the program is the cost removal investors first modelled.

The AI operating metrics have arrived on time. More than 50% AI-assisted code, 8,500+ AI engagements (up from about 2,500 in Q2 2025), and the frontier-certification numbers are internal metrics management defined itself. It has hit each one quickly. They are not yet visible in revenue growth, which is the investment question in Section 8.1.

Assessment. On the variables it controls, this management team does what it says. Margins, cost programmes, capital return and the AI workforce build-out have all been delivered at or above guidance for six quarters. On revenue it has been consistently too optimistic about the timing of a demand recovery. It missed the 2026 guide by repeatedly assuming a second-half pickup that did not come, while keeping a confident tone on bookings.

10. Ownership, Governance and Shareholder Friendliness

Part 1 - Ownership and control. Cognizant is a widely held company with one class of shares and no controlling holder. There is no founder, family, parent company or state stake. The largest holders are index and institutional managers:

• Vanguard Capital Management, which reported 35,975,456 shares (7.52% of the class) as of March 31, 2026. This followed a Vanguard internal realignment, after which The Vanguard Group itself no longer reports beneficial ownership (Schedule 13G filings, 2026).
• BlackRock, listed in the 2026 proxy as a holder of more than 5%. Its exact stake was not captured in the extract reviewed.

There were 447,030,425 Class A shares outstanding at October 23, 2026 (Q3 2026 10-Q cover). For a minority shareholder this means one share, one vote, and a full free float. Board control is decided by institutional votes, and there is no related-party channel through which a controlling holder could extract value.

Part 2 - Governance. Governance is mostly clean. The 2026 proxy shows 13 director nominees, 12 of them independent. The chair, Stephen J. Rohleder, has been the independent chair since 2022, separate from the CEO. The proxy reports no material related-party transactions, and PricewaterhouseCoopers was ratified as auditor for 2026.

CEO pay is 93% variable and 79% equity. Performance-share units are tied to revenue growth, adjusted operating margin and TSR relative to peers. The 2023-2025 PSU cycle paid out at only 57% of target, which shows the metrics do bite. The 2025 annual cash incentive paid out at 169% of target for corporate leaders (2026 proxy).

Two historical items are now closed. First, the India FCPA matter: the company settled its own case with the SEC in 2019. The DOJ dropped criminal charges against former President Gordon Coburn and former Chief Legal Officer Steven Schwartz in April 2025, and the SEC dismissed its civil case against them with prejudice in July 2025. Second, the 2026 proxy carried a shareholder proposal for the right to act by written consent, which the board opposed.

Part 3 - Capital returns

Dividends. Dividends declared per share were:

So the dividend has grown every year, by mid-single digits, with a 6.5% increase for 2026 (FY2025 and FY2024 annual reports; Q4 2025 release). The 2026 dividends paid totalled US$465 million by September 30.

Buybacks and dilution. Buybacks are now the main channel, and they were accelerated in 2026.

• Authorisation. The board added US$2 billion in March 2025, taking the remaining authorisation to about US$3.1 billion at March 24, 2025 (8-K, March 2025). About US$1.9 billion remained at end-2025 and US$2.1 billion at September 30, 2026 (Q4 2025 and Q3 2026 releases).
• 2025. Repurchased 17.4 million shares for US$1.3 billion (FY2025 release). The plan for 2025 had been raised to US$1.1 billion at the Investor Day, so this beat it.
• 2026, January to September. Repurchased 34.8 million shares for US$1.877 billion at an average of about US$53.93 (Q3 2026 10-Q). This includes US$1.1 billion in Q2 alone (22.5 million shares at about US$51), a May 2026 accelerated share repurchase of 9.7 million shares at about US$51.54, and 4.9 million shares (US$258 million) in Q3. The original plan for 2026 was about US$1 billion.
• Last ~90 days. MoatMap's buyback table shows no buyback rows since August 2, 2026. The company's own Q3 release shows 4.9 million shares bought in the July-September quarter, part of which falls in that window. The company figure is the authoritative one; MoatMap's US buyback feed typically lags until it parses the 10-Q.
• Net share count. Shares outstanding fell from about 495 million at end-2024 to 479 million at end-2025, then to 447 million by October 2026. Weighted diluted shares fell from 505 million in 2023 to 489 million in 2025. That is roughly a 10% net reduction since end-2024, well beyond any option dilution.

Verdict: Returns Capital. Cognizant has returned more than it planned in each of the last two years and has cut its share count by about a tenth in under two years, while still paying a growing dividend.

11. Insider Activities

The MoatMap disclosure database (US market, current as of October 30, 2026) is the base for this section. It was cross-checked against SEC Form 4 filings through the secform4.com mirror of EDGAR. In the window it records 158 Form 4 lines from 16 insiders: 1 purchase, 12 sales and 145 "other" entries. The "other" lines are almost entirely vesting of restricted and performance stock units on September 1 and September 15, 2026, plus the shares withheld to cover tax on them. Recipients included CEO Ravi Kumar S, CFO Jatin Dalal, Chief Legal Officer John Kim, Chief People Officer Kathryn Diaz, President of the Americas Surya Gummadi, and Rajesh Varrier (President, Operations and India CMD). These are compensation events, not decisions to trade. No newer open-market trades appeared in the cross-check for October 17-31, 2026.

MoatMap counts 12 sales in the window. The Form 4 extract itemised the eight from June to September, which make up the table. The rest were not itemised in the data available.

Buys - reading the signal. The chair bought 4,034 shares on the open market on June 10, 2026. Based on the shares it reported, this was his first open-market purchase recorded, which takes it beyond routine director stock grants. The timing is notable: it came during the steep spring-summer fall in IT-services stocks triggered by fears of AI disruption (Section 6), and alongside the company's own accelerated buyback. The amount, in the order of US$0.2 million, is modest compared with a board chair's typical holdings, and no other insider bought in the same window. It is a positive signal from the most senior independent director, not a cluster buy.

Sells - working out the why

• CFO Jatin Dalal's sale of 12,000 shares on August 3, 2026 came a few days after the Q2 2026 call, in the normal post-earnings trading window. The Form 4 extract showed no 10b5-1 plan marker. The reason was not disclosed.
• Chief Legal Officer John Kim's identical 2,500-share sales on August 3 and September 2 look like a pre-set schedule. The plan was not confirmed in the data reviewed.
• Chief Accounting Officer Alina Kerdman's sales are tiny and regular, consistent with selling vested shares.
• The CEO made no open-market sales in the window. His only Form 4 activity was vesting and tax withholding.

Net assessment. By number of trades, insiders were net sellers, but the selling is small, concentrated in three officers, and routine in pattern. The single meaningful discretionary decision on either side was the independent chair's open-market purchase near the lows. The CEO neither bought nor sold. Read: neutral, tilting mildly positive because of the chair's buy. Nothing in the window points to insiders reducing their exposure because of a view on the business.

12. Scenarios

Bull case. Financial services keeps showing that AI expands scope rather than shrinking it. Banks and insurers move from pilots to production agents and pay Cognizant to build, govern and run them through AI Nexus and the partner groups. By 2027 the same pattern reaches healthcare payers as Medicaid and Medicare rules settle, and modernisation work comes back on TriZetto. Tokenized rate cards let Cognizant keep part of the productivity it creates instead of passing it all back. Astreya, 3Cloud and DyFlex give it a credible position in AI infrastructure and enterprise platforms that it lacked. Project Leap's reinvestment shows up as differentiated offerings rather than lost cost savings. Cognizant meets its "Winner's Circle" growth goal in 2027 on organic growth, keeps taking share from slower Indian peers in North American large deals, and an Indian listing widens its shareholder base. The business that results is a smaller, more productive workforce earning more per person, with a growing share of fixed-price and platform revenue.

Base case. Management keeps delivering on what it controls, as Section 9 shows: steady margin expansion, buybacks and dividends at or above plan, a reskilled workforce and quick AI-adoption metrics. Revenue growth stays mid-single-digit. Large deals and new-business wins keep the base growing, but AI deflation in maintenance work offsets much of it, and discretionary recovery comes partially and late. Financial services keeps leading, Health Sciences is flat until US healthcare funding is clearer, and Products & Resources follows the consumer economy. The acquired businesses add a point or two of growth. In this path Cognizant stays competitive and keeps its share among the large offshore firms, but the "top-tier growth by 2027" goal is met only with acquisitions or against weak peers. That leaves the business stable, cash-generative and shrinking its share count, with no clear move to a new growth trajectory.

Bear case. 8.1 and 8.2 combine. Clients keep discretionary budgets frozen through 2027 while using AI as leverage at every renewal, so bookings stay healthy while revenue per contract falls. 8.3 hits at the same time: payer clients facing Medicaid funding cuts re-bid application and BPO work on price, and the platform franchise grows only in maintenance. Headcount is no longer the growth lever, and Project Leap becomes a recurring program rather than a one-off. 8.5 adds a fixed-price stumble on one of the 2025 mega-deals, priced on productivity assumptions that turn out optimistic. 8.4 is a further risk if the US makes offshore outsourcing more expensive. Under this combination Cognizant's margins hold only through repeated cost-cutting, acquisitions mask organic shrinkage, and the business becomes a slowly declining labour franchise defending its incumbency.

Sources:

• Cognizant Q3 2026 results release
• Q3 2026 earnings call transcript (Motley Fool)
• Q3 2026 Form 10-Q
• Q2 2026 earnings call transcript (Motley Fool)
• Q2 2026 guidance (Investing.com transcript)
• Q1 2026 results release
• Q1 2026 call summary (Yahoo Finance)
• Q4/FY2025 results release
• Q4 2025 earnings call transcript (Motley Fool)
• Q3 2025 earnings call transcript (Motley Fool)
• Q2 2025 call highlights (Yahoo Finance)
• Q2 2025 results release
• FY2025 Form 10-K
• 2026 Proxy Statement (DEF 14A)
• Investor Day 2025 release
• Investor Day 2025 8-K
• Cognizant (Wikipedia)
• Gartner IT spending forecast, July 2026
• Gartner 2030 forecast commentary (Software Strategies blog)
• Business Today on the Indian IT sell-off
• InfotechLead on 2026 IT services customer shifts
• Babak Hodjat on AI threat (Yahoo Finance)
• Startup Fortune on Cognizant job cuts
• Morningstar on vendor consolidation
• Cognizant considers India listing (Motilal Oswal)
• Infosys can't escape Cognizant's trade secrets suit (Bloomberg Law)
• Infosys told to rework Cognizant monopoly claims (Bloomberg Law)
• TriZetto and Infosys CEO deposition fight (Law360)
• DOJ dismisses FCPA charges against former Cognizant executives (Baker Donelson)
• SEC dismisses FCPA case against former Cognizant executives (Willkie)
• Insider transactions (secform4.com, EDGAR mirror)
• Stephen Rohleder insider purchase (GuruFocus)
• Vanguard Schedule 13G/A
• Cognizant forms OpenAI delivery group (ROI-NJ)
• Peer market caps, October 2026: Accenture, TCS, Infosys, HCLTech, Wipro, Capgemini, EPAM