Engineering Premium

Engineering Premium

EPAM's edge was never a patent or a network. It was a reputation for the hardest engineering, staffed from a scarce, low-attrition talent pool — and for a decade it showed up as revenue growth two to three times the scale incumbents' at comparable margins. That growth premium has now closed: organic growth runs slightly below Accenture's and Cognizant's, and reported operating margin sits beneath both. What remains is genuine client stickiness — being tested on price and mix, not yet on retention.

In a labor-based services business there is no moat you can point to on a balance sheet. The tell is indirect: clients choosing a more expensive provider, and paying for that choice in the vendor's growth rate and pricing. On that test EPAM was, for years, the clearest case in its cohort. Read at the competitive level, this is the report's cyclical-versus-structural read again: the evidence below weighs whether the premium is dormant or spent.

The premium showed up as growth

For six years through 2022, EPAM compounded revenue at 25% to 41% a year — 28.4% in 2022 alone, to $4.825 billion — while the scale IT-services incumbents grew in the low-to-mid single digits [1]. That gap is the visible fingerprint of an edge: enterprises routed their most demanding build work to EPAM and let it grow into their budgets.

Source: derived from reported financials, FY2017–FY2025 10-Ks; FY2022 growth per FY2022 10-K MD&A [2].

The 2025 figure needs a caveat. Reported growth of 15.4% was lifted by the late-2024 NEORIS and First Derivative acquisitions; on an organic constant-currency basis, revenue rose 4.9% [3]. Over the same year Accenture's revenue grew 7.4% and Cognizant's 7.0%. On the cleanest available measure, EPAM is now growing marginally slower than the two incumbents it historically outpaced. Management's 2026 guidance of 3% to 6% organic constant-currency growth keeps it in that same band [4].

What built the edge: scarce talent, embedded clients

The premium rested on two things a competitor cannot quickly copy: the quality of the delivery pool and the depth of the client relationships.

The talent signal is retention. EPAM's voluntary attrition was 8.5% in 2025, 8.9% in 2024 and 8.6% in 2023 [5]. Cognizant, whose model leans on a much larger India base, ran 15.9% in 2024 and 13.8% in 2023 [6]. A workforce that turns over at half the peer rate keeps institutional knowledge inside client accounts and lowers the cost of staffing complex work — the operational core of the reputation.

That reputation converts into rate. EPAM earned roughly $87,000 of revenue per employee in 2025 (5,457 million dollars across about 62,850 people) [7] against about $59,000 at Cognizant, whose roughly 336,800 staff generated $19.7 billion [8]. Geographic mix explains part of the gap, but the direction is the point: EPAM sells scarcer, higher-value hours.

Sources: attrition and headcount, FY2025 10-K [9] [10]; client concentration, FY2025 10-K MD&A [11].

The second pillar is how diversified and embedded the client base is. EPAM's top five clients were 13.7% of revenue in 2025, the top twenty 31.9%, and everything below the top twenty 68.1% [12]. For a firm this size that is unusually low concentration, and it has kept falling as the base broadens.

Source: FY2025 10-K MD&A, Revenues by Client Concentration (2023–2025) [13].

The mechanism behind the falling concentration is land-and-expand still working. In the fourth quarter of 2025, revenue from clients outside the top twenty grew 15.5% year over year while the top-twenty accounts grew 7.3% [14]. New and smaller relationships are compounding faster than the anchor accounts — the opposite of a franchise losing its pull with customers.

Where the premium has narrowed

Three lines of evidence show the edge being priced, not abandoned.

First, pricing power. In 2025 EPAM's cost of revenues rose to 71.2% of sales from 69.3%, and management's own explanation was blunt: the increase was "primarily due to compensation increases which we were not able to fully offset through pricing increases" [15]. A firm with unassailable rate power passes wage inflation through; EPAM could not fully do so, and gross margin absorbed the difference.

Second, the margin gap to the scale peers. On reported figures EPAM's 2025 operating margin was 9.5%, against roughly 15.6% at Accenture and 16.1% at Cognizant. Much of that gap is the intangible-amortization drag from the 2024 acquisitions, examined in Earnings Quality: on the adjusted basis management guides to, EPAM's operating margin was 15.2% [16]. The honest read is that EPAM's margin premium over the incumbents has closed to rough parity, not that it has inverted.

Sources: EPAM GAAP and adjusted operating margin, Q4 FY2025 earnings call [17]; Globant, Accenture and Cognizant operating margins derived from reported financials, latest fiscal year.

Third, the delivery base is drifting toward the same geography as the low-cost incumbents. India is now EPAM's single largest delivery location, at roughly 12,200 professionals, ahead of Ukraine (8,750), Poland (5,050), Belarus (3,400) and Mexico (2,950) [18]. The Central and Eastern European engineering heritage that once set EPAM apart is being blended, of necessity, with the same India talent pool that Infosys, Cognizant and TCS draw on. EPAM itself frames the field this way, listing competitors from Accenture and Cognizant to Endava, Globant, Grid Dynamics, Infosys and TCS, and naming "price" among the principal competitive factors [19].

What holds, and what would decide it

The distinction that matters is between a moat under pricing pressure and a moat under retention pressure. EPAM's evidence points to the first. Clients are not leaving — attrition of the workforce is low, the client base is broadening, and the long tail is growing faster than the anchors. What has weakened is the ability to convert that loyalty into premium growth and premium price, in a market where the scarce-talent story is diluting and AI reframes what an engineer-hour is worth (the pricing channel traced in AI and Billable Hours).

The strongest fact against a durable-moat read is that the whole pure-play engineering cohort re-rated together — EPAM, Globant and Grid Dynamics — which suggests the market is repricing the engineering-hours premium itself rather than any one firm's execution. Globant, the closest listed comparable, decelerated from 15.3% revenue growth in 2024 to 1.6% in 2025 on the same forces.

What would decide durability is observable and near-term: whether gross margin stabilizes or keeps eroding as wage costs outrun pricing; whether organic growth re-accelerates back above the incumbents as deferred programs convert; and whether revenue per delivery professional holds as the India mix rises. Retention metrics turning down — attrition climbing, the client base concentrating, the long tail stalling — would move the read from "premium under pressure" to "moat impaired." None of those has turned yet.