Margin Bridge

Where the Margin Went

EPAM's GAAP operating margin fell from 11.5% in 2024 to 9.5% in 2025 [1]. That drop is a cost-of-delivery event, not overhead bloat: cost of revenue rose from 69.3% to 71.2% of sales while SG&A actually fell as a share of revenue [2]. Acquisition dilution, a lapsed Poland tax credit, higher variable pay and currency did the damage. The company's own guidance puts margin back to 10–11% in 2026 [3] — a claim this chapter tests against the exit run-rate.

The two-point drop, decomposed

The move from an 11.5% to a 9.5% operating margin is worth about two points of revenue, or roughly $110 million. Almost all of it sits above SG&A. Gross margin — revenue less cost of delivery — narrowed 1.9 points, from 30.7% to 28.8%. Amortization and depreciation, swollen by intangibles from the late-2024 acquisitions, took another 0.4 points. Working the other way, SG&A leveraged down by about 0.3 points as revenue outgrew the sales-and-admin base.

Source: derived from FY2025 Annual Report (Form 10-K), MD&A results-of-operations table [4].

The bridge below isolates the three moving parts. Gross margin is the story; SG&A is a modest offset; amortization is the acquisition tail.

Source: derived from FY2025 Annual Report (Form 10-K), MD&A results-of-operations table; components sum to the reported 2.0-point decline [5].

What compressed the gross line

Management attributes the 1.9-point gross-margin fall to four named causes, and the mix matters because they age differently. First, the NEORIS and First Derivative acquisitions completed in late 2024 carry lower delivery profitability and were consolidated for a full year in 2025 [6]. Second, a $13.6 million cumulative Poland R&D tax credit that had inflated 2024's cost of revenue did not recur, so the year-over-year comparison is flattered on the base and penalised in 2025 [7]. Third, variable compensation rose with a stronger second half — a cost that tracks performance, not a leak [8]. Fourth, the appreciation of delivery-location currencies added about 0.8 points of unfavourable cost, partly offset by hedging [9].

The through-line running under all four: "compensation increases which we were not able to fully offset through pricing" [10]. That sentence has recurred across the record since 2024, when a promotion campaign lifted pay faster than bill rates [11]. It is the same wage-versus-price squeeze the moat chapter traced to the post-war shift of delivery into India and Latin America and to emerging-market wage inflation (Moat and Peers).

Two of the four causes are essentially non-recurring in nature — the Poland credit lapped clean, and a full year of acquisition consolidation is a one-time step in the base. The other two — wage inflation outrunning price, and currency — are the durable open questions, and they are where the AI thesis meets the income statement.

Utilization and pricing: no deflation in the numbers yet

If AI were already deflating EPAM's billable hours (AI Demand and Deflation), the tell would be falling bill rates and slumping utilization. Neither is visible. Utilization has held in a tight mid-70s band — 75.4% in Q4 2025, 77.0% in Q1 2026 [12] — with the modest recent dip explained by holiday timing and the deliberate intake of junior staff who ramp at lower initial utilization [13].

Pricing, if anything, turned the right way late in the period. The CFO described "slightly better pricing" through the second half of 2025 [14] and confirmed outright "rate increases in Q1" 2026 that "helped profitability" [15]. That is the opposite of the deflation bear case, at least so far: the pricing pressure that compressed 2024–25 gross margin was wage-led, not AI-led, and it was easing as 2026 opened. The counter is that AI-driven price deflation, when it comes, would show up here first — so the absence of it in 2025 data is reassuring about the present, not dispositive about the thesis.

The quarterly rebuild, and whether the guide is credible

The full-year 9.5% understates where the business exited 2025. Operating margin bottomed at 7.6% in the seasonally weak first quarter — payroll-tax caps reset in January and fewer billable days — then rebuilt each quarter to 10.6% by Q4. The first quarter of 2026 fell back to 8.3%, the same January seasonality repeating.

Source: quarterly income statements, FY2025 10-Qs and Q4 2025 release; FY2026 guide of 10–11% GAAP operating income [16].

This reframes the credibility question. The 10–11% FY2026 guide is not a promised recovery from a 9.5% base; it roughly annualises the 10.4–10.6% run-rate the company was already printing in the third and fourth quarters of 2025 [17]. Management framed 2026 as "committed to improving overall profitability, and specifically gross margin," with the CFO expecting year-over-year gross-margin improvement in the first half [18], [19]. That is a lower bar than the full-year 2025 optic implies.

The bar is not free, though. The guide leans on the same second-half-weighted seasonality repeating, and 2026 opened with softer demand: in May 2026 the company cut its full-year revenue-growth outlook to 4–6.5% from 4.5–7.5%, citing delayed client decisions and higher energy prices [20]. It held the 10–11% margin band while trimming the top line — meaning the margin now depends more on cost discipline than on volume leverage.

The "one-time" charges that keep arriving

That cost discipline has a recurring line item. EPAM has run a named "cost optimization program" — workforce reductions booked to SG&A — in each of the last three years, and the charges have grown, not faded: about $35 million in 2023, $31 million in 2024 and $48 million in 2025, with a further roughly $25 million flagged to complete the 2025 program by mid-2026 [21].

Source: FY2025 Annual Report (Form 10-K), Note 12 — Cost Optimization Programs; 2026 is the company's estimate of charges remaining to complete the 2025 program [22].

A charge that recurs every year for four years is, in substance, an operating cost of running the post-war delivery model — continuously re-shaping the workforce toward India and Latin America and out of higher-cost geographies. It is real cash. But note where it sits: in SG&A, the very line that leveraged down in 2025. So even carrying these charges, overhead is not the margin problem. The margin problem is gross, and it is largely acquisition mix and a lapsed tax credit — items that do not repeat at the same scale.

The GAAP-to-non-GAAP gap is structural, and rising

One reason EPAM's headline margin looks thin against peers is the size of the wedge between its GAAP and non-GAAP numbers. Non-GAAP operating margin was 15.2% in 2025 [23] — about 5.7 points above the 9.5% GAAP figure. The wedge is stock-based compensation (about 3.2% of revenue and growing, $147.7M in 2023 to $176.8M in 2025) [24], amortization of acquired intangibles ($71.4M), and the restructuring above.

The cleanest read on delivery economics strips both corporate stock comp and acquired-intangible amortization: EPAM's own segment operating profit — revenue less delivery cost, SG&A and depreciation, before those unallocated items — was $855.1 million in 2025, or 15.7% of revenue, down from $825.6 million and 17.5% in 2024 [25]. Two points there: the compression is real even before any GAAP adjustment noise — the delivery margin genuinely fell about 1.8 points — but at 15.7% that pre-corporate margin sits close to Accenture's 15.6% reported operating margin, which reframes the "bottom-of-peer" GAAP optic from the moat chapter as substantially a stock-comp-and-amortization difference rather than a delivery-quality one (Moat and Peers).

The read

The evidence points to a cyclical-and-mix margin trough rather than a structural reset. The 2025 compression is concentrated in the gross line, and its largest single drivers — a full year of lower-margin acquisitions and a lapsed Poland tax credit — are step-changes in the base, not a deteriorating run-rate. Utilization is stable, pricing turned positive into 2026, and the business exited 2025 at a 10.4–10.6% quarterly margin that already sits inside the FY2026 guide. On those facts the 10–11% guide is credible.

The strongest fact against that read is the cost-optimization line: a "one-time" restructuring charge in its fourth consecutive year, escalating, is the tell of a delivery model that needs continuous re-shaping to hold its margin — and the guide now leans on that discipline because the top-line assumption was cut in May 2026. What would change the read: gross margin failing to improve year over year in the first half of 2026, against management's explicit commitment, or the cost-optimization charges continuing past the 2025 program's mid-2026 completion. Either would move the story from mix-driven trough toward structural reset.