Moat and Peers

Moat and Peers

EPAM sells itself as a premium engineering firm, and exactly one number backs that claim: revenue per employee near Accenture's and well above the offshore majors. Everywhere else the premium disappears. Its GAAP operating margin (9.5%) and returns on equity (10.3%) sit in the lower half of a six-name peer set, below Accenture, Cognizant and Infosys. The durable evidence of a moat is narrower than the reputation: clients stayed through the loss of EPAM's entire Eastern-European delivery base.

How defensible the franchise actually is decides whether the de-rating is a permanent impairment or a cyclical markdown.

The field EPAM names for itself

EPAM does not have to be told who it competes with; it lists them. Its 10-K names Accenture, Atos, Capgemini, Cognizant, Deloitte Digital, DXC, Endava, Genpact, GlobalLogic, Globant, Grid Dynamics, HCL, Infosys, Tata Consultancy Services and Wipro [1], and it defines the competitive factors as technical expertise, industry knowledge, on-time delivery, recruiting and retention, scale, financial stability and price [2]. It stakes its differentiation on "complex and innovative software product development solutions," a technical employee base, and the ability to "deliver end-to-end AI-native solutions" [3].

That framing — engineering-led, not staff-augmentation — is testable. A firm doing harder work than its rivals should earn either a higher price per head, a higher margin, or higher returns. EPAM earns the first and not the other two.

Where the premium is real: revenue per head

EPAM ran 62,850 employees at the end of 2025, of which about 56,600 were delivery professionals [4]. Against $5.46 billion of revenue, that is roughly $87,000 of revenue per employee — near Accenture's ~$89,000 (779,000 people against $69.7 billion) [5] and well above Cognizant's ~$60,000 (351,600 employees against $21.1 billion) [6]. The offshore-heavy India majors sit lower still.

Source: derived from reported FY2025 revenue and disclosed headcount — EPAM 10-K [7]; Accenture 10-K [8]; Cognizant 10-K [9].

The comparison is imperfect: the India majors run a wider pyramid of lower-cost junior staff, which depresses their revenue per head without depressing their profitability. But the signal survives the caveat. On price-per-person, EPAM clusters with Accenture at the top of the field, not with the volume outsourcers. Its clients are paying for harder work.

Where it isn't: margins and returns

That pricing does not reach the bottom line. On GAAP operating margin, EPAM's 9.5% in FY2025 is the second-lowest of the group — barely ahead of Globant (9.4%), and far behind Accenture (14.7%) and Cognizant (16.1%).

Source: derived from FY2025 filings — EPAM 10-K MD&A [10]; Accenture 10-K income statement [11]; Cognizant 10-K income statement [12]. Infosys reports in rupees and is shown in the table below.

The same ordering holds on returns. Two adjustments are owed to EPAM here, and they matter. First, its GAAP margin is depressed by stock-based compensation (~$177 million, roughly 3.2% of revenue) and acquired-intangible amortization; on the non-GAAP basis the company reports, operating margin runs near 15% (Margin Bridge), closer to Accenture and Cognizant — though those peers also carry stock comp. Second, EPAM's return on equity is dragged down by a balance sheet still holding ~$1.3 billion of net cash, capital that Accenture and Infosys return rather than park. Even allowing for both, the picture is a firm whose economics are ordinary against the leaders and whose returns rank near the bottom.

Sources: derived from most-recent-fiscal-year filings (fiscal years end Aug 2025 for Accenture, Mar 2025 for Infosys, Dec 2025 for the rest; Infosys revenue converted from ₹1,629.9bn at ~₹83.5/$) — EPAM 10-K [13]; Accenture 10-K [14]; Cognizant 10-K [15].

The reason the premium price does not become premium profit is where EPAM once had its edge. Its differentiation was built on high-calibre engineers in Central and Eastern Europe, priced below Western consultancies but above commodity outsourcers. Two forces have compressed that arbitrage. The war pushed delivery into India and Latin America — the same geographies the majors already own — and EPAM's own filing warns that emerging-market wages "have increased and will make us less competitive if we are not able to increase the efficiency and productivity of our people" [16]. The talent-cost gap that funded the margin is narrowing toward the field.

Growth: fast, but not the fastest

EPAM's filings and calls lean on a claim of superior organic growth. The peer data does not clearly support it. Its FY2025 reported revenue grew 15.4%, but only 4.9% of that was organic constant-currency (Financials and Estimates); the rest was the NEORIS and First Derivative acquisitions. On a like-for-like basis that ~5% is roughly where Accenture (7.4% reported) and Cognizant (7.0%) landed, and behind Infosys and the fast-reaccelerating small-cap Grid Dynamics (17.5%). Over three years the ordering is starker: EPAM's revenue compounded at 4.2%, matching Accenture and beating Cognizant, but trailing Globant (11.3%), Infosys (10.3%) and Grid Dynamics (9.9%). EPAM is a mid-pack organic grower whose ~$912 million of acquisitions lifted reported growth above its organic rate.

The AI-scale gap

The most direct test of franchise durability now is AI, and here EPAM is a credible specialist rather than a leader. It guides its defined "AI-native" revenue above $600 million for all of 2026 (AI Demand and Deflation). Accenture booked $2.2 billion of advanced-AI work in a single quarter, "nearly doubling" year-over-year [17]. A firm competing on being the best engineer for the AI transition is being out-booked roughly fifteen-to-one by the largest generalist. Scale in this market compounds — more reference deployments, more data, more pricing leverage in vendor consolidations — and EPAM is on the wrong side of it.

The moat test EPAM has already passed

The counter-evidence is real, and it is the most important fact in this chapter. In 2022 EPAM lost access to its entire Russia, Belarus and Ukraine delivery base — for years the majority of its engineers — effectively overnight, and had to "shift portions of our delivery capabilities to other countries" under active sanction and client pressure [18]. A commodity staffing arrangement does not survive that; the client relationships were rebuilt on new continents, and the business grew through it. That is switching-cost evidence a peer table cannot show — the work is embedded enough, and the teams trusted enough, that clients relocated with EPAM rather than replacing it. It sits alongside the low client concentration and long tenure established at the outset of this report (Fallen Star): a top-five that is 13.7% of revenue, and 64% of revenue from clients of five years or more.

So the moat exists, and it is specific: deep, sticky engineering relationships that hold through disruption. What the peer numbers say is that it is narrow, not wide — real enough to retain clients through catastrophe, not wide enough to command a margin or return premium, and being pressed on price by wage convergence and on scale by Accenture's AI scale.

What would change the read

On today's evidence the franchise is durable but not dominant — a narrow, contested moat rather than a permanently impaired or a widening one. Two developments would move that judgment. If EPAM's GAAP operating margin converges back toward the mid-teens while organic growth holds at mid-single digits — showing the premium price finally reaching profit as the delivery base stabilises — the cyclical-markdown reading gains ground. If instead AI-native revenue stalls below its $600 million target, or margins stay stuck near 9-10% through the FY2026 guide period, the case that the engineering edge has been competed away, on both cost and scale, gets harder to answer.