eClerx — a cash machine in the eye of the AI storm
eClerx Services Ltd
Snapshot
eClerx runs the back office of the world’s big companies. It does the un-glamorous, brain-heavy work — checking trades for big banks, running “know-your-customer” checks, cleaning product data and pricing for retailers, handling customer-service analytics for telecom and cable firms, and producing digital images for fashion brands. It sells this to roughly 50 Fortune-2000 clients, mostly in the US and Europe. Market cap ₹13,343 cr, price ₹1,419. P/E 18.9, P/B 5.2×, RoE 29%, RoCE 35%, EPS ₹75 (post-bonus), dividend yield basically zero (it returns cash through buybacks instead).
⚠️ A framing note: screener shows a 52-week range of ₹2,498 / ₹1,321, which looks like a 43% crash. But eClerx did a 1:1 bonus issue with its FY26 results (May 2026) — that mechanically halves the price. So much of the optical “drop” is the bonus, not a true de-rating. The real signal is bonus-neutral: the P/E has compressed to ~19×, well below the ~25–35× this business historically commanded, and it trades cheaper than its weaker peers. That de-rating is real and AI-fear-driven; the “43% crash” headline is half illusion.
What kind of animal is it? A Great, asset-light cash machine that the market has quietly marked down because it fears artificial intelligence (computers that can read, write and reason) will eat the very work eClerx sells. As of 2026-06-20, from screener snapshot.
The verdict in one box
| Lens | Result |
|---|---|
| QGLP score | 21.5 / 25 (Quality 10.5/12 · Growth 6/6 · Longevity 3.5/5 · Price 1.5/2) |
| Buffett rubric | 7.5 / 10 PASS |
| Business bucket | Great (asset-light, 35% RoCE, free-cash fountain) |
| Wealth-creator type | Enduring · Consistent — but with a live disruption asterisk |
| Economic Profit | ₹435 cr (RoE 29% − CoE 12% on ₹2,561 cr net worth) — strongly creating value |
| Margin-of-safety price band | ₹1,150–1,400 (PEG ≈ 1×; Emkay’s published target is ₹1,800). CMP ₹1,419 is fair; below ~₹1,200 you’re paid to take the AI risk |
A Great, beautifully-run business that is a genuine wealth creator, priced fairly-to-cheap versus its quality — with the entire bear case resting on one question: does AI feed it or eat it?
In plain English
Imagine a company that takes ₹100 of its owners’ money and turns it into ₹29 of profit every year, year after year, while barely needing to buy any machines or factories to do it. It collects its profit as actual cash — not paper — and hands most of that cash back to shareholders by buying back its own shares. That’s eClerx. By the plain numbers, it’s one of the highest-quality businesses on the Indian market: a 29% return on equity, a 35% return on capital, almost no debt, and cash profits that match or beat its reported profits nearly every single year. Warren Buffett would call this a wonderful “boat” — the kind of business that makes money without swallowing capital.
So why has the stock quietly de-rated? One word: AI. (Mind the optics first: the screen shows a 43% “fall,” but eClerx split each share into two via a 1:1 bonus in May 2026, which halves the price on paper — so the true de-rating is more like a third, not a half. Still real, just not a crash.) eClerx’s business is, at its root, renting out smart people in India to do work that costs more to do in New York or London — checking trades, cleaning data, answering analytics questions, tagging product photos. The fear, stated plainly, is that a machine can now do a lot of that for pennies, so clients won’t need eClerx’s people anymore. The market has reacted by marking a once-30×-earnings compounder down to a P/E of 19 — cheaper than its slower, lower-quality rivals.
Here’s the tension. The fear is real and serious — this is a labour-arbitrage business and AI is coming for routine knowledge work. But so far it is nowhere in the numbers. In the year just ended (FY26), revenue grew 18% in dollars, profit grew 30%, margins expanded, and the company won its first big “agentic AI” deal (where the AI does the task and eClerx gets paid for the outcome, not the headcount). Management’s argument is that AI is a tool they wield, not a flood that drowns them — because what clients actually pay for is domain knowledge (knowing how a derivatives trade settles, what a KYC rule requires), and AI without that knowledge is a clever intern with no judgement.
The moat is a services moat — sticky client relationships, deep domain knowledge, and a few software products woven into clients’ workflows. It’s real (the returns prove it held for a decade) but it is narrower than a brand or a network, and AI is testing its edges right now. Where it’s widening: eClerx is repositioning fast, training thousands of staff on agentic AI, and pricing some work by outcome. Where it’s narrowing: the same AI that helps eClerx also lets a client do the work in-house.
The whole investment, then, comes down to one bet: is eClerx the blacksmith when the car arrives, or the toll-booth operator who learns to sell electricity? That is the crux, and I interrogate it below. The price you pay reflects which way you lean — at ₹1,419 the market is pricing meaningful fear, which is exactly when a clear-eyed owner starts paying attention.
Sitting down with the management
Two men built this: PD (Priyadarshan) Mundhra and Anjan Malik, both Wharton MBAs who walked away from finance careers — Mundhra from banking, Malik from Accenture Europe and Lehman Brothers — to start eClerx in 2000 and float it in 2007. The choice of word matters: they set out to build a knowledge-process outsourcer (judgement-heavy work), not a call-centre. Twenty-six years on they still own the place — promoters hold 54.5%, split almost evenly between the two founders (~26.85% each), and that stake has risen (it was 53.6% two years ago). Mundhra is an Executive Director, Malik a Non-Executive Director.
The succession shape is genuinely healthy. In May 2023 the founders handed the day-to-day to a professional CEO — Kapil Jain, an ~20-year Infosys BPM veteran (last role: global head of sales for Infosys BPM), on a five-year term. So a domain-deep outsider now runs the shop while the founder-owners set strategy and keep their capital in. That meaningfully lowers the key-man risk eClerx carried a decade ago, even though the company is still founder-controlled.
The capital-allocation story is the most admirable thing here, and it’s unusual. eClerx pays almost no dividend (payout ~1%) and instead returns cash through tender buybacks — repurchasing shares, which quietly raises every remaining owner’s slice. The cadence is steady: buybacks in 2022, 2024 at ₹2,800/share, and 2025 (₹300 cr at ₹4,500/share, record date Dec 2025) — with promoters abstaining, so the cash flowed entirely to public holders. Over the decade the share count trended down; no dilution, no serial equity raises. The two acquisitions — CLX Europe (2015, ~€25m, fashion/luxury digital) and Personiv (2020, ~$34m at ~5.5–6× EBITDA, US finance-and-accounting outsourcing) — are small, on-strategy bolt-ons bought at sane prices, not empire-building “toads.” Every retained rupee has compounded book value at ~29% — the one-dollar test passes cleanly.
One even-handed ding, though: that 2025 buyback at ₹4,500 (pre-bonus) now looks like cash returned near a cyclical high — the stock is far below that today. Buying back your own shares is only smart below intrinsic value, and on this one the timing was poor. It’s a venial sin (the cash still left the building to owners), but it’s not the textbook discipline of a Buffett buyback.
Do they talk straight? On the evidence of several concalls, yes. They guide conservatively — “top-quartile growth versus peers, EBITDA margin 24–28%, sequential profit growth” — and then deliver it. They flag the lumpiness of a small revenue base (“$120–130m a quarter, there will be aberration”) rather than pretending every quarter rises. On the hard AI question the CEO doesn’t dodge: the 15–20% of revenue that rolls off yearly is normal project churn, AI is “another productivity tool” that “takes time to implement,” and it’s “too early” to size agentic revenue. That’s a management resisting hype in a year when every services CEO claims to be an AI company.
The forensic red-flags? Almost none fire. Profit becomes cash (operating cash flow ran 90–104% of operating profit the last five years). Receivables improved sharply (debtor days 88 → 59). Promoters declared zero pledge/encumbrance for FY26. No SEBI run-in, no auditor drama, no accounting controversy on the public record. The balance sheet isn’t bloating; “borrowings” of ₹385 cr are overwhelmingly capitalised office leases (net of cash, effectively debt-free). The honest soft spots are cultural, not financial: a middling Glassdoor of 3.3/5 with recurring “low pay” complaints, and ~20% attrition — normal for Indian BPM, but a quality/margin headwind, and a reminder that the “asset” walks out of the building every evening.
Would Buffett and Agrawal shake hands on this management? Yes — comfortably. Owner-operators with deep skin in the game, a professionalised CEO, clean accounts and pledge, a shareholder-first buyback culture, disciplined M&A, and guidance they actually hit. The two things that would change their mind: if the AI pivot curdles into expensive, value-destroying M&A or a vanity “AI platform” land-grab — and if buybacks keep getting timed at the highs rather than the lows. Watch where the cash goes from here.
What’s on the horizon (live-issues tracker)
Three live threads will decide the next one-to-three years. The first is the whole ballgame.
1. THE CRUX — Does generative/agentic AI feed eClerx or eat it? 🟡 (too early, but tracking the right way)
Name it in one sentence: eClerx works as an investment if, and only if, AI ends up augmenting its domain-heavy services faster than it lets clients automate that work away.
The mechanism, in plain words. eClerx’s profit comes from a price gap: a trade-checking or data-cleaning task that costs $40/hour in London costs eClerx perhaps $12/hour in Pune, and eClerx keeps the spread. AI threatens this in two different ways, and they’re worth separating because they do different damage:
- Damage to volume/headcount — if an AI agent does the task, you need fewer people, so the hours eClerx bills shrink.
- Damage to price/take-rate — even if eClerx runs the AI, clients may demand the savings be passed to them, compressing the fee.
The bull’s reply is a real one: the moat isn’t the labour, it’s the domain. An LLM can draft an email, but it doesn’t know that a particular derivative needs a specific collateral treatment, or which KYC exception a regulator will accept. eClerx wraps the AI in domain knowledge and “human-in-the-loop” judgement, and — crucially — it’s beginning to sell the outcome rather than the hour. Its KYC “Compliance Manager” already cut a client’s refresh cost ~50% using agentic AI, and eClerx kept the contract. If you’re paid for the outcome, an AI that does the work cheaper makes you more profitable, not less.
Test the analogy. Is this like the blacksmith when the automobile arrived (doomed), or like a logistics firm when barcodes arrived (made stronger)? The honest answer: it depends which task. The most commoditised, rules-based work (basic data entry, tier-1 customer support) is the blacksmith — AI will compress it. The judgement-heavy, regulated, deeply-embedded work (financial-markets operations, financial-crime compliance, pricing analytics) is the logistics firm — AI is a power tool that a domain expert wields. eClerx’s mix leans toward the second, which is why the numbers haven’t cracked. But the line between the two moves every quarter, in AI’s favour.
Named-threat map (fuller version pending web research):
| Who can exploit AI against eClerx | Backed by | How they pressure | Proof point so far |
|---|---|---|---|
| WNS | Being acquired by Capgemini for $3.3bn (July 2025) explicitly to build “agentic-AI-powered intelligent operations” | Scaled, AI-blended ops at a price pure-plays can’t match | The clearest signal yet that the industry expects AI to favour scale over pure arbitrage |
| Genpact, Accenture (diversified majors) | Huge AI practices, CXO access | Bundle “AI transformation,” push pure BPM out | Winning AI-transformation budgets; mixed on pure ops |
| Firstsource, Sagility, IKS Health, Indegene | Public-market scale | Same agentic pitch, more scale | None has shown AI eroding revenue yet — all still growing |
| AI-native startups | VC funding | Cheap outcome-priced agents for single tasks | Strong in narrow tasks; weak on regulated, end-to-end domain work |
| The client itself (insourcing) | Off-the-shelf LLMs | ”Why pay eClerx, we’ll run it ourselves” | The real long-run risk — but no eClerx client has cut spend citing AI yet (the one soft patch, Q4 BFSI, was a consulting wind-down, not AI) |
Precedent — two films, one reassuring, one cautionary. First, the RPA (robotic process automation) wave of ~2016–2020 was supposed to gut BPO. It didn’t — it shifted work up the value chain and the survivors grew. That’s the reassuring precedent: these firms have absorbed an automation wave before. Second, and more sobering, is what’s already happened to the freshly-listed pure-plays: Sagility and IKS Health — both 2024 IPOs — fell ~21% and ~12% respectively on AI-disruption fears even though their revenues kept growing. That tells you the market is willing to de-rate the multiple well ahead of any actual revenue erosion — which is precisely what has happened to eClerx. So the precedent cuts both ways: the business has survived automation before, but the stock can stay de-rated for a long time on fear alone. GenAI is broader than RPA, so whether it’s a wave (absorbable) or a tide (structural) is the genuinely open question.
Follow-on questions, answered:
- Is the damage to share or to fee? So far, neither is visible — FY26 grew 18% USD with margins up. The first pressure, if it comes, will likely be on fee (clients demanding AI savings) before share.
- Which segment is hit first? The most exposed is routine customer-operations/support; the most protected is financial-markets data ops and financial-crime compliance (regulated, judgement-heavy).
- What is the incumbent doing, and is it credible? eClerx trained 3,000+ staff on agentic AI, built internal platforms (Roboworx/Cogniflows), launched an agentic data-sourcing platform for banks, won its first large agentic deal in Q4 FY26, and is shifting to outcome-based pricing. This is credible action, not just talk — but the agentic revenue is still “too early” to size.
- Has anyone actually moved yet? No. As of the May 2026 call, no client had cut spend citing AI; the order book and ACV (new deal value) were healthy.
Honest verdict: genuinely two-sided, leaning constructive but watch-listed. This is not a “too hard” — eClerx’s domain-heavy mix, clean execution, and early agentic wins tilt the odds toward “AI is a tool it wields.” But it is not settled either; the market’s fear is rational, not silly. The tell will be in the organic revenue growth and margin over the next 4–6 quarters: if AI is eating the business, you’ll see USD growth slow toward zero and margins slip below 24% despite a strong pipeline. So far the opposite is happening.
2. US “offshore call restrictions” (the FCC NPRM) 🟡 monitoring
On 26 March 2026 the US FCC adopted a Notice of Proposed Rulemaking that proposes to cap offshore-handled customer-service calls (a 30% cap is up for comment), set English-proficiency standards, and bar “foreign-adversary” delivery locations. It’s still only a proposal at the comment stage (mid-2026), but it lands squarely on eClerx’s customer-operations / CMT vertical serving US telecom and cable clients — a meaningful slice of the non-BFSI ~57% of revenue (eClerx hasn’t quantified the exact exposure). The hedge: eClerx is diversifying delivery into Manila, Cairo and Fayetteville (US) — its new Cairo centre reached “top-quartile” performance within four months. But note the limit of that hedge: Cairo is still offshore, so it doesn’t answer a rule that mandates onshore work. A real, unquantified tail risk — worth watching, not yet sized.
3. Margin discipline through the AI investment 🟢 on track
The risk in any AI pivot is that you spend heavily and margins crater. So far eClerx has expanded margins (FY26 EBITDA +132 bps) while investing in AI, sales and tech, and reaffirms its 24–28% band. The discipline is holding.
The watch-list (check these next quarter):
- Organic USD revenue growth — stays double-digit = thesis intact; slides toward low-single-digit despite a “strong pipeline” = AI is biting.
- Operating EBITDA margin — holds inside 24–28% = discipline intact; breaks below 24% = AI deflation or pricing pressure showing up.
- Agentic-AI revenue — management starts sizing it (stops saying “too early”) = the pivot is monetising.
- Any client reducing spend and citing AI — the first real one would be a red flag; none yet.
- New deal wins / ACV — Q4 FY26 was $46m; a sustained fall would signal the funnel drying up.
- Where the cash goes — buybacks (good) vs a large “AI platform” acquisition (caution).
QGLP scorecard (the Motilal Oswal lens) — the receipts
| # | Question | Score | Evidence |
|---|---|---|---|
| Quality of Business | 5.0/6 | ||
| 1 | Large opportunity? | 1 | Global BPM + data/AI services TAM; ~50 Fortune-2000 clients, low penetration |
| 2 | Favourable industry structure? | 0.5 | BPM is fragmented & competitive; OPM compressed mid-30s→mid-20s over the decade, though stable recently |
| 3 | Defensible moat? | 1 | RoCE > cost of capital 10 of 10 years; domain depth + embedded products (Compliance Manager, Market360) |
| 4 | Return ratios > 15%? | 1 | RoE 29%, RoCE 35%; both >15% every year for a decade |
| 5 | Asset-light? | 1 | FCF ₹756 cr > PAT ₹706 cr; modest capex, RoCE 35% — the See’s profile |
| 6 | Favourable terms of trade? | 0.5 | Positive working capital (not negative-WC), but cash conversion ~100% and debtor days improved 88→59 |
| Quality of Management | 5.5/6 | ||
| 7 | Unquestionable integrity? | 1 | Profit backed by cash (CFO/OP ~100%), no pledging, clean accounts, buybacks at fair value |
| 8 | Proven execution? | 1 | Hits its own guidance (top-quartile growth, 24–28% margin) quarter after quarter; revived growth post FY18–20 stall |
| 9 | Growth mindset? | 1 | Investing in analytics ($90m book), AI/agentic, new geos & verticals |
| 10 | Superior capital allocation? | 1 | Buybacks not dilution; no value-destroying M&A; reinvests at ~29% RoE — one-dollar test passes |
| 11 | Succession plan? | 0.5 | Professional CEO (Kapil Jain) under founder-chairman; still founder-anchored / key-man on promoters |
| 12 | Minority interests protected? | 1 | Buybacks benefit all holders; reasonable promoter pay; rising DII + promoter holding |
| Growth | 6.0/6 | ||
| 13 | Structural tailwind? | 1 | Digital/data/AI services growing well above nominal GDP |
| 14 | Volume-led growth? | 1 | FY26 +18% USD, volume/headcount + value-led (analytics, automation) |
| 15 | Operating leverage? | 1 | FY26 EBITDA margin +132 bps while revenue grew; recent trend clearly positive |
| 16 | Manageable leverage? | 1 | Effectively debt-free; “borrowings” are mostly leases; D/E ~0.15 |
| 17 | Market-share gains? | 1 | Winning ACV “from competition,” top-quartile growth, top-10 concentration falling to 59% |
| 18 | Earnings growth > 15%? | 1 | 5-yr PAT CAGR 20%, FY26 PAT +30%; (3-yr 13% the only soft spot) |
| Longevity | 3.5/5 | ||
| 19 | Relevant for 10–15 yrs (low disruption)? | 0.5 | The crux — labour-arbitrage services are genuinely AI-exposed; management adapting but risk is real |
| 20 | Extend Competitive Advantage Period? | 0.5 | Domain + products help; AI pressures the edge of the moat |
| 21 | Sustain Growth Advantage Period? | 0.5 | Large TAM, but AI could shrink the billable-hours pool |
| 22 | Diversification headroom? | 1 | New geos (Manila, Cairo, US), new verticals (BFSI analytics, F&A), Adobe partnership |
| 23 | Adaptive culture? | 1 | Survived FY18–20 earnings recession and reinvented; pivoting hard to agentic AI now |
| Price | 1.5/2 | ||
| 24 | Reasonable valuation vs growth? | 1 | P/E 18.9, PEG ≈ 1×, below its own history and below peers despite higher quality; post 43% drawdown |
| 25 | Margin of safety (PEG<1 / payback<1)? | 0.5 | PEG ~1 (borderline); 5-yr payback 2.4× (fails the textbook <1×, as most quality names do) |
| Total | 21.5/25 |
The pattern: Quality and Growth are the strength (16.5 of a possible 18). The entire deduction sits in Longevity — the AI-disruption question — with Price a notch off textbook-cheap. In plain terms: this is a textbook wealth creator on every axis except the one the market is obsessed with, and the price has already moved to reflect that obsession.
Buffett lens (the Berkshire-letters read)
| # | Test | Verdict | Evidence / the Buffett line |
|---|---|---|---|
| 1 | Good boat? (business > management) | PASS | Great: 35% RoCE, asset-light, free-cash fountain. “A good managerial record is far more a function of what business boat you get into.” |
| 2 | Moat + franchise + pricing power | PARTIAL | RoE beat CoC 10/10, but it’s a services firm — limited raw pricing power (decade margin compression); outcome-pricing only now emerging |
| 3 | See’s test — high returns, little capital | PASS | FCF ₹756 cr > PAT ₹706 cr; tiny capex vs cash flow |
| 4 | One-dollar test (capital allocation) | PASS | Book & market value compounded at high RoE; buybacks below value; no dilution, no toad-kissing M&A |
| 5 | Owner-oriented, candid management | PASS | 54.5% promoter skin, conservative guidance hit repeatedly, plain talk on AI. “We eat our own cooking.” |
| 6 | Integrity / no “credit P&L, debit balance sheet” | PASS | OCF ≈ PAT (CFO/OP ~100%); receivables improved; balance sheet not bloating |
| 7 | Circle of competence / predictability | PARTIAL | The weak test — can you say what eClerx looks like in 10 years? AI clouds it. “If there’s lots of technology, we won’t understand it.” |
| 8 | Mr. Market — gift or trap now? | PASS | P/E de-rated from ~25–35× historically to 19× on AI fear (the headline “43% drop” is half a 1:1 bonus); cheapest-yet-highest-quality in the peer set — the “Bruised Blue Chip” |
| 9 | Patience / compounding runway | PARTIAL | Long runway at high RoE if AI doesn’t shrink the pool — runway real but clouded |
| 10 | The honest red flag (not scored) | — | See below — AI disruption of the labour-arbitrage core |
Score: 6 PASS + 3 PARTIAL = 7.5 / 10 — “a real business with real gaps,” and every gap clusters on the same fault line: durability/predictability under AI.
The See’s test, in numbers. See’s Candies needed only $32m of reinvestment over 35 years yet threw off $1.35bn. eClerx is cut from similar cloth: in FY26 it earned ₹706 cr of net profit and converted it into ₹756 cr of free cash — more cash than reported profit — on a business needing little capital to run. That’s the mark of a Great boat: it doesn’t drink its own profits.
The one-dollar test, in numbers. Has each retained rupee created at least a rupee of value? Over the decade eClerx retained almost everything (≈1% payout), compounded equity at a 29% return, and shrank the share count via buybacks rather than diluting. Book value per share, market value, and per-share earnings all compounded together — the textbook pass. The only smudge is timing: an owner who bought at the 2025 high is underwater today. But that’s a price mistake, not a capital-allocation one — the business kept its side of the bargain.
The framework metrics
- Economic Profit = Net Worth ₹2,561 cr × (RoE 29% − CoE 12%) = +₹435 cr. Strongly creating value — it earns ₹435 cr a year above the cost of its owners’ money. Top-quintile economics.
- Terms of Trade = Debtors (~₹665 cr) ÷ creditors ≈ near/above 100% — not a negative-working-capital business (normal for services), but cash conversion ~100% offsets it.
- 5-yr Payback = Mcap ₹13,343 cr ÷ projected 5-yr cumulative PAT ≈ 2.2–2.4× (assuming 15–20% PAT growth). Above the <1× multibagger threshold — as nearly all quality names are.
- PEG = P/E 18.9 ÷ growth = ~0.9–1.05 at 18–20% growth (1.45 if you use the soft 3-yr 13%). Around fair.
- RoE − CoE spread = +17%, and RoE/RoCE > 15% in 10 of the last 10 years — a wide, durable “uncommon profit.”
- Consistent/Volatile test = PASS (Consistent) — over 12 years, PAT fell >10% only twice (FY18, FY19), no fall exceeded 50%, terminal PAT (₹706 cr) far exceeds initial (₹230 cr). Value it on earnings, not book. Note the FY18–20 dip is a real reminder this isn’t bulletproof.
Peer comparison
| Company | Mcap (₹cr) | CMP (₹) | P/E | P/B | RoE | RoCE | OPM | FY26 Sales (₹cr) |
|---|---|---|---|---|---|---|---|---|
| eClerx | 13,343 | 1,419 | 18.9 | 5.2 | 29.0% | 34.8% | 26% | 4,117 |
| Firstsource (FSL) | 17,301 | 245 | 23.0 | 4.0 | 17.7% | 16.2% | 16% | 9,556 |
| Sagility | 18,487 | 39.5 | 19.5 | 1.9 | 10.5% | 13.4% | 24% | 7,193 |
| Datamatics | 4,737 | 802 | 19.8 | 3.1 | 16.4% | 20.8% | 19% | 1,987 |
The relative read flips nothing — it confirms. eClerx has the highest return ratios in its asset class by a wide margin (29% RoE vs 10–18% for peers) and yet trades at the lowest or near-lowest P/E of the group. You’re being asked to pay less for the best business in the room. The usual reason a quality leader trades cheap to weaker peers is a perceived terminal risk — and here that risk has a name: AI exposure of the labour-arbitrage model, which the market judges higher for eClerx’s data/ops-heavy mix than for, say, Sagility’s healthcare-ops focus. Whether that discount is wisdom or fear is the whole debate.
Latest quarter & what’s happening now
Q4 / FY26 (reported May 2026, HARD). FY26 operating revenue $469m (+18% USD, +22% INR to ₹4,217 cr); EBITDA +29% to ₹1,153 cr; PAT +30% to ₹706 cr; ACV (new deal value) +24% YoY. Alongside results the board declared a 1:1 bonus (cosmetic — doubles share count, halves the optical price; the reason the chart looks like a 43% fall). Q4 itself was sequentially soft (+0.6% QoQ revenue) but margins held at 25.7% and new deal wins were $46m. Utilisation 74%, attrition 21% (low for the sector), top-10 client concentration down to 59%. Sell-side stays constructive — Emkay rates it Buy with a ₹1,800 (post-bonus) target, ~27% above CMP.
Concall takeaways: (1) First large-scale agentic-AI win in Q4, deploying Q1 FY27 — outcome-focused, not headcount-priced (MEDIUM). (2) Analytics & automation now a $90m book (HARD). (3) FY27 guidance reaffirmed: top-quartile growth, 24–28% EBITDA margin, sequential profit growth (MEDIUM). (4) Watching the US NPRM offshore-call rule, diversifying delivery to Manila/Cairo/Fayetteville (SOFT). (5) On AI cannibalisation, management is firm that the 15–20% annual project roll-off is normal churn, not AI deflation (MEDIUM).
Where the two lenses agree — and disagree
They agree almost entirely, which is itself the message. QGLP scores it a 21.5/25 wealth creator; Buffett scores it 7.5/10 Great-boat-with-gaps. Both flag the exact same single weakness — Longevity (QGLP Q19–21) = Predictability (Buffett test 7 & 9) — and both arrive there from different doors. That convergence tells you the analysis is honest: this is a high-quality, well-run, fairly-priced business whose only serious question is whether AI extends or ends its runway. There is no hidden accounting flag, no capital-allocation skeleton, no governance trap that one lens caught and the other missed. The debate is entirely about the future of the industry, not the quality of this company.
Margin-of-safety price band
Not a recommendation — the arithmetic. On QGLP’s Price pillar, PEG ≈ 1× is satisfied around a P/E of 15–19 on ~18% sustainable growth, i.e. ₹1,125–1,425. For reference, Emkay’s published target is ₹1,800 (~18× FY28 EPS), and the business historically commanded 25–35× — so today’s 19× is a genuine de-rating, not a bubble. The Buffett Mr.-Market read on CMP ₹1,419: the crowd is fearful on this name (a real multiple compression on AI worry, cheapest-best-in-class in its peer set), which is when quality is worth examining, not avoiding.
- ₹1,150–1,400 — the “paid to take the AI risk” zone (P/E ~15–19, PEG ≤ 1). CMP ₹1,419 sits at the top edge: fair, not a steal.
- Below ~₹1,200 (P/E ~16) — a genuine margin of safety opens: you’re buying a 29%-RoE cash machine at a price that already assumes AI bites.
- Above ~₹1,900 (P/E ~25) — you’d be paying for AI being a clean tailwind, with little cushion if it isn’t.
Plainly: a wonderful business at a fair price today, a cheap one on any further fear — provided the AI longevity question breaks the right way.
Conviction texture
Bull case, strongest form: This is the best business in its peer group — 29% RoE, 35% RoCE, more free cash than profit, owner-operators buying back stock — trading at a discount to weaker rivals because of an AI fear that, four quarters in, is completely absent from the numbers (18% USD growth, expanding margins, first agentic wins already landing). If AI proves to be a tool eClerx wields (as RPA was), the multiple re-rates and earnings compound — a double engine from a depressed base. You’re buying a Bruised Blue Chip at PEG ≈ 1.
Bear case, strongest form (the honest red flag): eClerx sells labour arbitrage dressed as domain expertise, and generative AI is the first technology that attacks the labour and the expertise at once. The decade-long margin slide (mid-30s → mid-20s OPM) already shows the pricing power is bounded. The FY18–20 earnings recession proves the business can stall. If clients use AI to insource even a fifth of this work, eClerx’s growth and margins go backwards together — and a 19× P/E on falling earnings is not cheap, it’s a value trap with a delay. No client has cut spend for AI yet — but “yet” is doing a lot of work, and the market is paid to anticipate.
What the numbers actually support: A genuinely Great, Consistent, value-creating business (EP +₹435 cr, RoE>CoE 10/10 years, cash-backed profits) priced fairly versus its quality and cheaply versus its peers. The numbers do not yet show any AI erosion; they show acceleration. The bear case is a forecast, not a fact — but it’s a reasonable forecast on a real fault line.
The two or three things that tip it: (1) organic USD growth — double-digit = bull intact, low-single-digit “despite strong pipeline” = bear winning; (2) EBITDA margin holding ≥24% through the AI spend; (3) the first client that cuts spend and names AI as the reason. Watch those, not the share price.
No buy/sell/hold — the reader decides.
Sources
- Screener.in — eClerx Services (consolidated), snapshot fetched 2026-06-20.
- eClerx Q4/FY26 earnings call transcript, 19 May 2026 (primary — AI commentary, guidance, NPRM, agentic wins).
- eClerx Q3 FY26 earnings call transcript, Feb 2026 (guidance arc, analytics segment detail).
- eClerx Annual Reports FY24 & FY25 (governance, board, remuneration).
- Peer snapshots (Firstsource, Sagility, Datamatics) — screener.in, fetched 2026-06-20.
- eClerx Q4/FY26 results, 15 May 2026 — Businesswire; bonus issue — ICICIdirect.
- CEO appointment (Kapil Jain, May 2023) — eClerx newspaper intimation.
- Buybacks: 2024 at ₹2,800 — Business Today; 2025 ₹300 cr at ₹4,500 — Chittorgarh. Acquisitions: Personiv 2020, CLX 2015.
- No promoter pledge FY26 — ScanX. Glassdoor 3.3/5 — Glassdoor.
- AI-disruption / BPM context: Outlook Business — GenAI threatens Indian BPOs (Sagility/IKS de-rating, WNS-Capgemini $3.3bn deal). FCC offshore-call NPRM (26 Mar 2026) — Davis Wright Tremaine.
- Assumptions: Cost of Equity = 12% (mid of the studies’ 10–15% range); forward PAT growth 15–20% for payback/PEG (vs FY26 actual +30%, 5-yr CAGR 20%, 3-yr 13%).