← back to screen
EPAM Systems — source materials
Verification of documents pulled for the 12-agent deep-dive extraction. Every link opens in a new tab.
SEC filings — 10-Q / 10-K
8 most recent 10-Qs (Q2 2023 → Q3 2025) plus the most recent 10-K (FY2025, covers Q4 2025). Source: SEC EDGAR.
Earnings call transcripts
Full Q&A transcripts for all 8 quarters (Q1 2024 → Q4 2025). Sources: Motley Fool (6), Insider Monkey (Q2 2025), Investing.com via Playwright (Q1 2025 — curl got 403, headless Chromium passed the anti-bot). Note: the Investing.com transcript uses collapsed speaker labels ("EPAM Systems:") rather than individual names and has minor OCR-like errors; for deflection/tone agents, cross-reference with the Q1 2025 press release.
8-K press releases (supplementary)
Official earnings press releases filed as 8-K exhibits. Contain full GAAP/non-GAAP financials and forward guidance. Q1 2025 press release compensates for the missing transcript.
| Type | Period | Filed | Source | Local HTML | Local TXT (stripped) |
|---|
| 8-K Press Release | Q2 2025 | 2025-08-07 | SEC EDGAR ↗ | html (278KB) | txt (26KB) |
| 8-K Press Release | Q1 2025 | 2025-05-08 | SEC EDGAR ↗ | html (218KB) | txt (24KB) |
Coverage summary: 8 × 10-Q + 1 × 10-K + 8 × earnings transcript + 2 × 8-K press release = 19 documents locally saved.
Transcripts cover Q1 2024 → Q4 2025 — all 8 quarters present.
Deep-dive analysis
Short thesis (bottom line): EPAM's headline +15.4% FY2025 revenue growth is 76% inorganic: organic constant-currency growth is only +4.9% (Peterson, Q4 2025 call, line 204) — below the +11% LTM trading multiple and below peer INFY's raised FY26 guide of 3.0–3.5%. Management still frames AI as a pure tailwind while utilization has drifted 78.1% → 76.5% → 75.4% over three quarters, gross-margin % of COGS jumped 69% → 73%, and the Q4 2025 call introduced new hedging vocab — "elongated sales cycle" — for the first time. The Feb 2026 10-K itself concedes that "increased competition … from new and non-traditional market participants like AI-based task-specific tools, has negatively impacted the price of our stock" (Item 1A).
The edge is narrative-vs-fundamentals divergence: ACN/INFY deny AI pricing pressure AND are raising guidance; GLOB/DAVA have already repriced after admitting the compression. EPAM is running the mega-cap denial script with the mid-cap's actual headcount/utilization stress — roughly 2 quarters behind GLOB on the same curve. Q1 2026 guidance is already ~3% organic CC (vs 4.9% FY25 actual), with a 100bps NEORIS-largest-client drag (tariff exposure) that is an acquired problem not an organic one.
Signal strength: MEDIUM. Already-crowded short (18% of float, +48% MoM), -40% from 52w high, Fwd P/E 9.4x below peer median 11.6x — a lot is priced in. The trade is time, not direction: long-dated out-of-the-money puts where IV skew is negative (calls bid higher than puts at 6mo+), sized for a guide-down catalyst in H2 2026 when NEORIS lapping + organic deceleration converge.
Signal grid — 12 agents + 6-peer cross-read + priced-in check
Each row shows a specific angle of the thesis and the strength of evidence found. Red = strong short signal, orange = medium, yellow = weak, green = counter-signal (bull), gray = neutral/resolved.
| Angle | Key finding | Signal |
|---|
| 1 | Segment P&L | BIM vertical 4 consecutive qtrs flat/declining; COGS jumped 69%→73% in Q1 2025 — pricing-power erosion on top-line growth | medium-strong |
| 2 | Cost & capital structure | D&A stepped $20M→$31M/qtr from acquisition amortization; SBC hit new high $48.5M Q1 2025 despite buybacks | medium |
| 3 | Balance sheet | Goodwill 13%→25% of assets post-NEORIS+FD; cash drained $2.0B→$1.3B; DSO crept 71→77 days at peak | medium |
| 4 | Risk-disclosure drift | FY25 10-K adds explicit AI disruption risk: "AI-based task-specific tools … has negatively impacted the price of our stock" (Item 1A) | medium |
| 5 | Guidance walk / hedging vocab | "Elongated sales cycle" appears for FIRST time in Q4 2025 call; FY26 midpoint 4.5% < FY25 actual 4.9% | medium |
| 6 | Segment narrative shift | BIM disappeared as a discussed vertical; Travel renamed; pivot to "AI-native revenue" as headline KPI | medium |
| 7 | Competition / disruption posture | Management deflects on AI disruption; 10-K risk factor is pointed — tone mismatch with transcripts | medium |
| 8 | Capital allocation | $661M buybacks in 2025 into a falling stock; no insider sales; no dividend — aligned, but capital-intensive | weak |
| 9 | Customer & demand | Utilization drifting 78.1%→76.5%→75.4% Q2-Q4 2025; pricing JUST inflected positive; top-20 slowing vs non-top-20 | medium |
| 10 | AI practice trajectory | AI-native revenue $105M in Q4 2025 (~7.5%); AIRON/AI/RUN rebranded EVERY quarter since Q3 2024 — classic marketing cadence | medium |
| 11 | Analyst pushback substance | Same 3 unresolved issues 8 calls running (guide credibility, pricing vs wage, India mix) but tone softened 2H25 | medium |
| 12 | Mgmt deflection / non-answer | Targeted deflection on exactly the load-bearing bull-case fields: AI monetization, pricing power, 2026 decel | medium |
| 13 | Acquisition carve-out (new) | Reported 15.4% → organic 4.9%; 9.2pp inorganic + 1.3pp FX. NEORIS client tariff-exposed = -1pp FY26 drag (acquired problem) | strong |
| 14 | Peer: ACN (Accenture) | Same denial script BUT record $22.1B bookings back it up; Julie Sweet "AI is expansionary, not deflationary" — industry cover for EPAM | neutral |
| 15 | Peer: GLOB (Globant) | CFO: "AI Pod model by definition requires less people"; CTO admits per-profile bill-rate pressure. 2025 rev +1.3%, 2026 guide +0.2-2.2%. EPAM is ~2Q behind | strong |
| 16 | Peer: DAVA (Endava) | CEO: "higher productivity… eroding revenue" — explicitly acknowledged AI T&M erosion. Already repriced -77% — EPAM's future if denial breaks | strong |
| 17 | Peer: CTSH (Cognizant) | Same denial script but fixed-price rose 41%→50%+ of revenue 3yrs — quiet T&M migration under AI pressure, reframed as strategy | medium |
| 18 | Peer: WIT (Wipro) | CFO admits renewal productivity pass-through; utilization is RISING (opposite of EPAM); attrition 2% near-record low | medium
(isolates EPAM) |
| 19 | Peer: INFY (Infosys) | Raised guidance TWICE FY26; Project Maximus +30-50bps pricing lift; "pricing accretive"; explicit counter-signal | counter |
| 20 | Priced-in check (Q44-49) | Stock -40% from 52w high, -34% YTD; short% of float 18.1% (+48% MoM); Fwd P/E 9.4x vs peer med 11.6x. Analyst target mean $187 (+42.5% upside) — sell-side NOT capitulated; FY26 EPS revisions net -3 (30d) | medium priced-in, edge in timing |
Top 5 red flags (most concerning with citations)
1. Organic growth is 4.9%, not 15.4% — and is decelerating in 2026 guide.
FY2025 reported revenue growth of
+15.4% decomposes to +9.2pp inorganic (NEORIS $626M + First Derivative $301M + 3 smaller 2024 deals) + 1.3pp FX +
only 4.9pp organic constant-currency. FY26 midpoint guide is
4.5% organic CC — a decel, and the midpoint includes a -1pp drag from NEORIS' largest client (Mexico-headquartered industrial materials group) impaired by US tariffs.
"The midpoint of the range would produce 4.5%. … we did get incremental information on the NEORIS largest client … the decline on a year-over-year basis will have a negative 100 basis point impact on growth." — Jason Peterson, Q4 2025 call, line 224
Source: epam_agents/agent_13_acquisition_carveout.md
2. Utilization has fallen 3 quarters straight while management claims pricing is improving.
Q2 2025 78.1% → Q3 2025 76.5% → Q4 2025 75.4%. Utilization falling while pricing "just inflected positive" (Q3 2025) is a classic late-cycle tell: the company may be preserving rates by staffing light, sacrificing top-line. The FY26 guide implies this continues.
"We are observing modest extension of sales cycle across larger clients where procurement is more formally involved and where AI-based decision-making processes are being built into client evaluations." — Fejes paraphrase (Q4 2025 commentary on sales cycle)
Sources: epam_agents/agent_05_guidance_walk.md, agent_09_customer_demand.md
3. The 10-K itself ties AI-based competition to stock price decline.
In a new Item 1A risk factor added to the FY2025 10-K (Feb 2026), EPAM writes that "Increased competition, or the perception of increased competition, from new and non-traditional market participants
like AI-based task-specific tools, has negatively impacted the price of our stock." This is management-authored, SEC-filed ammunition — a far cleaner acknowledgment than anything in their transcripts.
10-K FY2025, Item 1A risk factors — quoted in agent_04_risk_drift.md
4. Acquisition accounting is ~100% goodwill + intangibles on both deals.
NEORIS: $401.6M goodwill + $259M intangibles = 106% of net consideration. First Derivative: $171.7M goodwill + $124.8M intangibles = 98% of net consideration. These weren't purchases of hard assets or durable IP — they were client-list-and-headcount roll-ups. Combined with the NEORIS client already deteriorating, goodwill impairment tail risk is real: goodwill on the balance sheet has grown 13%→25% of assets.
5. AIRON / AI/RUN rebrand cadence + no disclosed AI-engineer headcount.
AI/RUN was introduced Q3 2024 and has been re-launched or re-scoped EVERY quarter since (AI/RUN → AI/RUN Transform → AI/RUN Transform playbook + Agentic QA + AIron developer agent). Management has never disclosed how many engineers are on AI projects, how many deployments have gone live, or the dollar size of any single named AI win — despite being asked repeatedly. The $105M Q4 2025 "AI-native revenue" figure appears for the first time on the Q4 2025 call, using a restrictive definition that excludes the larger "AI-foundational" data/cloud work.
Sell-side on the Q4 2025 call pressed repeatedly for quantification on AI revenue and AI-engineer capacity; management deferred to qualitative narrative ("hundreds of AI native projects," "60-70% expanded from POC").
Source: epam_agents/agent_10_ai_trajectory.md, agent_12_deflection_detector.md
Top 3 counter-arguments (what the longs would say)
1. Customer concentration is actively de-risking. Top-5 clients: 16.6% (2023) → 15.8% (2024) → 13.7% (2025). Top-10: 23.6% → 23.4% → 21.6%. No single >10% client. Non-top-20 growing 2-3x faster than top-20. If AI were cannibalizing enterprise consulting demand, you'd expect concentration to rise (small clients churn first, large clients renew under pressure) — EPAM shows the opposite.
2. Pricing has just inflected positive after ~8 quarters of compression. FY24 management conceded "some ongoing pricing pressure," "modest pressure on average bill rates," "the pricing environment did not grow." Q3 2025 saw modest price improvement; Q4 2025 guided low-single-digit 2026 rate increases. If this sticks, the bear thesis (AI eating T&M rates) is empirically refuted — the inflection matters more than the magnitude.
3. Large-cap peer set is fully aligned (ACN, INFY, CTSH) — this is not an EPAM-only story. Accenture's Julie Sweet: "We do not see AI as deflationary. We do see and are seeing it as expansionary." Infosys raised FY26 guide twice with Project Maximus delivering 30-50bps pricing lift. If the entire large-cap IT services cohort is wrong, the contrarian call is on the industry — and EPAM is a high-beta mid-cap expression of that same thesis, not a uniquely-flawed name. ACN has record $22.1B bookings backing its narrative; EPAM's bookings aren't broken either.
Priced-in check — what the market has already digested
18.1%short % of float
+47.6%MoM Δ shares short
4.59M → 6.77M
-40.68%from 52w high
($221.40 → $131.34)
-34.48%YTD return
5.44days to cover
9.4xFwd P/E
(peer med 11.6x = -18.9%)
8.4xEV/EBITDA
(peer med 7.9x = +6.2%)
+42.5%upside to mean target
($131 → $187 mean; $193 median)
17analyst opinions
rec key: "buy" (mean 1.84)
Net -3FY26 EPS revisions (30d)
5 down / 2 up
+2.7%26-DTE put-call ATM skew
puts bid higher near-term
-5.3%180-DTE put-call ATM skew
calls bid higher — long-dated puts CHEAP
Reading: Short interest is crowded (18% of float, +48% MoM) and the stock is already -40% from highs — a lot of the thesis is priced in. But sell-side has not capitulated: target mean $187 (+42.5%), recommendation mean 1.84 (buy), only 3-5 analysts cut EPS in the last 30 days. The long-dated IV skew is inverted vs near-dated: 6-month OTM puts are relatively cheap on the vol surface. This is the structural edge — if the denial cracks the way GLOB's did, the guide cut arrives in H2 2026 when NEORIS lapping reveals true organic ~3-4%, and sell-side will be forced to re-rate down from $187 target.
Trade structure (thesis-driven): Modest-sized position in Oct 2026 OTM puts (~180 DTE, $110 strike = ~16% OTM). The IV skew at that tenor is negative (puts cheaper than calls on the vol surface), the catalyst window (Q2 2026 guide refresh, Aug 2026) lands inside expiry, and position sizing should be small because 18% short % of float is already crowded. Avoid front-month expiries — that IV is elevated (62% ATM put) and the short-squeeze-on-any-beat risk is high with an entrenched short book.
Kill triggers: (a) pricing inflection holds into Q1 2026 print with rate increases ≥3%, (b) FY26 guidance raised at mid-year, (c) NEORIS client stabilizes earlier than the 2026 stabilization implied on the Q4 call.
Remaining edge — what's NOT yet priced in
- NEORIS client impairment magnitude. Management disclosed "-1pp on 2026 organic CC" but the client-specific deterioration was only surfaced via analyst questions, not proactive disclosure. If the client goes from a 1pp drag to a 2-3pp drag (tariff scenarios escalate), FY26 organic could print ~2% — below the entire peer set including INFY.
- AI practice rebrand fatigue. AIRON/AI/RUN rebranded five times in six quarters. At some point a sell-side analyst will demand a stable definition, stable numerator, and a multi-quarter growth rate. When that happens, the $600M 2026 "AI-native" target becomes falsifiable — and it isn't currently in the consensus model.
- Utilization floor. 75.4% in Q4 2025 is already below EPAM's historical 77-80% range. If it breaks 73% in any quarter of 2026, the gross-margin math forces a guide cut regardless of pricing. Utilization trend is EPAM-specific — WIT's utilization is rising, ACN's is flat at 93%.
- Goodwill impairment test (Q3 2026). EPAM's annual goodwill impairment testing date is in Q3. With goodwill now 25% of assets and NEORIS' largest client in decline, an impairment charge in the $100-300M range is plausible by late 2026. This is not in any consensus model.
- Peer rating migration. If GLOB's two-quarter-ahead trajectory holds, EPAM sees a materially worse print around Q2-Q3 2026. Sell-side will cut in a cluster (ACN/INFY raised twice shows the opposite pattern when story is confirmed). The 17-analyst coverage means the rating migration can move the stock $15-25 in a single week if 4-5 analysts cut inside a month.
Agent reports (all 12 + 7 extensions)
Full evidence trail. Each agent's output is linked below for direct verification.
| # | Agent | Report |
|---|
| 1 | Segment P&L engine (Q1-4, 6) | agent_01_segment_pnl.md |
| 2 | Cost & capital structure (Q7-10, 20) | agent_02_cost_capital.md |
| 3 | Balance-sheet warning lights (Q11-16) | agent_03_balance_sheet.md |
| 4 | Risk-disclosure drift (Q5, 17-19, 21) | agent_04_risk_drift.md |
| 5 | Guidance walk + hedging vocab (Q22-24) | agent_05_guidance_walk.md |
| 6 | Segment narrative shift (Q25-27, 29, 32) | agent_06_segment_narrative.md |
| 7 | Competition & disruption posture (Q30, 31, 35) | agent_07_competition_disruption.md |
| 8 | Capital allocation tone (Q33, 34, 36) | agent_08_capital_allocation.md |
| 9 | Customer & demand commentary (Q37-40) | agent_09_customer_demand.md |
| 10 | AI practice trajectory (Q28) | agent_10_ai_trajectory.md |
| 11 | Analyst pushback substance (Q41, 43) | agent_11_analyst_pushback.md |
| 12 | Mgmt deflection / non-answer (Q42) | agent_12_deflection_detector.md |
| 13 | Acquisition carve-out (NEW — NEORIS + First Derivative → true organic) | agent_13_acquisition_carveout.md |
| P1 | Peer cross-read: ACN (Accenture) | peer_ACN_crossread.md |
| P2 | Peer cross-read: GLOB (Globant) — most relevant competitor | peer_GLOB_crossread.md |
| P3 | Peer cross-read: DAVA (Endava) — the repriced comp | peer_DAVA_crossread.md |
| P4 | Peer cross-read: CTSH (Cognizant) | peer_CTSH_crossread.md |
| P5 | Peer cross-read: WIT (Wipro) | peer_WIT_crossread.md |
| P6 | Peer cross-read: INFY (Infosys) — counter-signal | peer_INFY_crossread.md |
| — | Priced-in check data (raw) | epam_priced_in_report.md · epam_priced_in.json · epam_yf_analyst.json |
Generated via 12-agent parallel extraction (see .claude/skills/shorts-deep-dive/SKILL.md) on 19 local source documents (epam_sources/), plus 3 extensions: acquisition carve-out agent, 6 peer cross-read agents (ACN, GLOB, DAVA, CTSH, WIT, INFY — transcripts saved in peer_sources/), and priced-in market data (epam_priced_in_check.py + epam_yf_analyst.py; EV/EBITDA / short interest / IV skew / estimate revisions via yfinance; FMP rate-limited on daily tier). Every claim in this analysis is cited to a specific agent report, filing, or transcript — click any linked report to verify. Generated 2026-04-18.