Sterlite Technologies — a bruised fibre cyclical riding an AI dream
Sterlite Technologies Limited
Snapshot
Sterlite Technologies (STL) makes the glass thread that the internet runs on — optical fibre, the cables that bundle it, and the connectors that plug it together. It is India’s largest and lowest-cost fibre maker and holds about 8% of the optical-fibre-cable market outside China. Market cap ₹31,943 cr, share price ₹654, 52-week range ₹84.6–₹684, P/E 671, P/B 14.1×, RoE 2.24%, RoCE 7.75%. In one phrase: a capital-hungry, cyclical manufacturer — recently loss-making, now turning the corner — whose share price has run roughly eight-fold in a year on a genuine AI-datacentre demand story. As of 2026-06-20, from screener snapshot.
The verdict in two boxes — the business first, the price second
Keep them apart on purpose. Box 1 asks “what kind of business is this?” — it would read the same if the share price doubled or halved tomorrow. Box 2 asks “what is Mr. Market charging for it today?” — a separate, perishable reading.
Box 1 — The business (durable):
| Lens | Result |
|---|---|
| Business-quality score | 12 / 23 (Quality 5/12 · Growth 3.5/6 · Longevity 3.5/5) |
| Buffett rubric | 3 / 10 PASS |
| Business bucket | Good tilting Gruesome — a capital-heavy cyclical, not a cash fountain |
| Wealth-creator type | Transitory · Volatile (profit has swung from ₹578 cr to two years of losses) |
| Economic Profit | −₹221 cr (RoE 2.24% − CoE 12% on ₹2,268 cr net worth) — destroying value on FY26 numbers |
A Good-bordering-Gruesome business that is not currently a wealth creator — it earns less than the cost of its owners’ money — independent of what it costs today. The open question is whether the AI-fibre upcycle can finally lift returns above that hurdle and keep them there.
Box 2 — The price today (a current phenomenon):
| Reading | Result |
|---|---|
| CMP | ₹654 (as of 2026-06-20) |
| Price pillar | 0 / 2 (PEG not meaningful · 5-yr payback ≈ 7–14×) |
| Margin-of-safety band | Roughly ₹130–₹230 to satisfy QGLP’s payback ≤ 1× test on a generous normalised-earnings view; today’s price is ~3–5× above it |
| Mr. Market’s mood now | Euphoric — a $1 bn AI-datacentre order and a “fibre is the backbone of AI” narrative |
| CMP vs the band | Very demanding — priced for a turnaround that has barely begun to show in the accounts |
Today the market is pricing it for a near-perfect AI-fibre boom — a mood driven by one large hyperscaler order and the AI-capex frenzy, which can deflate quickly while the business above (a cyclical fibre maker only just back to a thin profit) does not change.
In plain English
Imagine a company that makes the thinnest, purest glass thread in the world, then bundles thousands of those threads into a cable and ships it to telecom companies and, increasingly, to the giant warehouses full of computers (“data centres”) that run artificial intelligence. That is STL. It does the whole chain itself — from melting sand-grade silicon into glass, to drawing the fibre, to making the cable, to the little connectors at the end. Doing all of it under one roof (“backward integration”) is its main edge: it can make fibre cheaper than most rivals.
Here is the trouble, and you should hold it firmly in mind before any AI excitement. This is a cyclical commodity business, and a capital-hungry one. Optical fibre is, at bottom, a thing sold by the kilometre, and China can flood the world with it. When demand is hot, STL earns lovely margins — in FY17–FY20 it earned a return on owners’ money (“RoE”) of 22–34%, which is genuinely excellent. Then India’s telecom companies stopped buying, China’s gluts arrived, STL had borrowed heavily to expand, and the whole thing fell over: revenue collapsed from ₹6,925 cr in FY23 to ₹4,083 cr, and the company posted two straight years of losses (FY24 and FY25). It is a “Good” business at best — one that needs constant feeding with capital to grow — and in bad years it behaves like a “Gruesome” one, the kind Buffett tells us to avoid: it eats cash just to stand still.
What’s happening now is real and worth understanding. Artificial intelligence is being built inside data centres, and AI data centres are extraordinarily hungry for fibre — by the company’s own slides, an AI rack needs up to 36× more fibre than an old-style one. So after two flat years, world fibre demand has turned up, and STL has won a string of orders, including a headline over-$1 bn supply deal with an unnamed US hyperscaler for AI-ready data centres, spread across FY27–FY29. Orders for the first nine months of FY26 were up 40%. The business is back to a small profit. This is a genuine tailwind, not a fairy tale.
But the share price has run eight times — from ₹84.6 to ₹684 — in a single year, while the company earned all of ₹56 cr. That is a price-to-earnings ratio of about 670 and fourteen times its book value. The market is no longer buying a fibre maker; it is buying a beautiful story about AI, priced as if the boom is permanent and STL captures most of it. The one-line tension is simple: the business has just clawed its way back to thin profitability after nearly drowning, and the price already assumes it will become a high-return compounder. The boat is at sea again; the ticket to ride is being sold at a champagne price.
Sitting down with the management
Dear reader — let me tell you who you’d be partnering with, because the name on the door matters here.
STL sits inside the Vedanta group. The non-executive chairman is Anil Agarwal, the mining billionaire who built Vedanta from a scrap-metal trader into a global metals-and-mining empire — and who is famous, in equal measure, for boldness and for debt. Vedanta’s defining deal, the ~$9 bn Cairn India acquisition, was half-funded by borrowing and seeded years of refinancing and credit-rating worry at the parent. That DNA shows up at STL. The day-to-day boss is his nephew, Ankit Agarwal (Managing Director), London Business School MBA, a 14-year STL veteran who cut his teeth doing M&A at Vedanta. Vice-chairman Pravin Agarwal has run the Sterlite telecom and power businesses for decades. So this is a promoter-run family business under a famously leveraged group — that is the first fact, and it cuts both ways.
Now the capital-allocation scorecard, which is the part Buffett and Agrawal (the QGLP author) weigh most heavily — and here STL stumbles. Over the last decade management borrowed aggressively to chase growth (borrowings climbed to ₹3,834 cr by FY23), expanded capacity into the teeth of a demand collapse, and watched returns on capital fall from ~30% to near zero, with two years of outright losses. Economic profit — what the company earns above the cost of its owners’ money — is deeply negative. That is the textbook one-dollar-test failure: rupees were retained and reinvested, but they did not reliably come back as a rupee of value. To their credit, they have since deleveraged hard — net debt is down to ₹1,331 cr, debt-to-equity 0.87× — and they spun off the lumpy, low-margin services/projects business (now “STL Networks / Invenia”) to focus the listed company on the higher-value optical and digital pieces. That is sensible housekeeping. But it followed a near-accident, and dividends have been zero for three years.
On candour, the concalls are a mixed read. Ankit and CFO Ajay Jhanjhari are fluent and answer the AI-demand questions well, but they decline to disclose volumes, capacity, utilisation, or the old-vs-new contract pricing gap — “competitive reasons” — which leaves an outside owner unable to verify the margin-recovery story for himself. That is a yellow flag, not a red one; many manufacturers do this, but it asks you to take the recovery partly on faith.
On the forensic checklist, a few things fire softly. The promoter sold its stake down from ~54% to ~44% over the last three years (screener flags −9.6%) — never a comforting tell, though there is no pledging (the MD has publicly confirmed no encumbrance in FY26). The MD took ₹2.96 cr in FY25 as “minimum remuneration” in a loss-making year, which required a special shareholder resolution — legal, disclosed, and modest in absolute terms, but worth noting in a year the company lost money. There’s a transfer-pricing income-tax notice and an ongoing US lawsuit (now at the Court of Appeals, with a bond posted, no payout yet) — management calls both immaterial; we can only watch. Accounts are audited by two of the Big Four and reported cash backs reported profit reasonably well (operating cash flow has tracked operating profit over the cycle), so I don’t see the “credit P&L, debit balance sheet” pattern.
Would Buffett and Agrawal shake hands on this management? Cautiously, and not warmly. They’re capable operators with real domain depth and an improving balance sheet — but the decade’s capital-allocation record destroyed value, the disclosure is guarded, and the Vedanta-group association carries a permanent “watch the debt and the related parties” caveat. The single thing that would change their mind: two or three years of returns on capital comfortably above 15% with the leverage kept low — proof that this time the cash actually sticks.
What’s on the horizon (live-issues tracker)
1 — The AI-datacentre fibre upcycle (the crux). 🟡 Real, early, and richly priced. This is the whole investment, so it gets the full interrogation below.
2 — The US tariff hit on margins. 🔴 Hurting now. STL is a net exporter — ~80% of revenue is from outside India, and North America jumped from 25% to 36% of sales in FY26. That is exactly where a 50% US tariff now bites. Management says the tariff knocked roughly 760 basis points off Q3 FY26 EBITDA margin, dragging reported margin down to 10.3% from an underlying ~18%. They’re mitigating by passing some cost to customers at contract renewal (a 2–3 quarter process) and ramping their own US plant — but the US factory cannot supply the full US demand (“not sufficient,” they admitted), so India shipments and the tariff stay in the mix. What’s next: whether an India-US trade deal lands, and how much tariff sticks to customers vs to STL over the next 2–3 quarters. This directly governs whether the AI orders are as profitable as the headline suggests.
3 — Germanium supply. 🟡 A genuine raw-material risk. Glass-grade fibre needs germanium, which China tightly controls. Management is sourcing from other geographies and using its own China facility, and sounds calm — but a supply squeeze here would hit the one thing that makes STL special (its from-scratch glass-making). Worth watching, not yet a problem.
4 — Deleveraging holding. 🟢 The good-news thread. Net debt ₹1,331 cr, D/E 0.87×, down sharply from the ₹3,800 cr peak. If margins recover and the AI orders convert to cash, the balance sheet is finally in a position to support growth without another near-death scare. What’s next: keep net-debt/EBITDA (now 2.58×) falling as revenue scales.
The crux, interrogated
The crux in one sentence: This investment works if and only if the AI-and-fibre demand boom lets STL earn returns on capital well above its cost of capital — and hold them — rather than being one more turn of a cycle that always reverts to a low-return commodity grind.
The mechanism — in plain English, with a tested analogy. What physically protects a fibre maker? Two things: being the lowest-cost producer (STL’s backward integration), and a technology lead in high-value cable (multi-core fibre, ultra-slim 864-fibre ribbon, hollow-core). The bull case is that AI data centres need so much specialised, dense, high-count fibre that the product stops being a plain commodity and becomes engineered kit with fatter margins — moving STL “up the value chain.”
Is that like, say, a memory-chip maker in an AI boom? Test it. Memory chips are a brutal commodity — every up-cycle (high prices, fat margins) pulls in capacity and a down-cycle follows; the cure for high prices is high prices. Optical fibre has behaved exactly the same way — STL’s own history (34% RoE in FY19, losses by FY24) is the cycle in one chart. The bull’s hope is that AI fibre is different — more like a specialised component than a bulk commodity. That is partly true at the leading edge (hollow-core, multi-core), but management itself concedes hollow-core is years from scale, has no industry standard yet, and is very costly to make; and on the ordinary high-count cable, China is the swing supplier and prices are set globally. So the analogy holds more than the bull would like: most of STL’s volume is still the commodity, and only a thin top layer is the genuinely differentiated, defensible product. The moat is real but narrow.
The named competition / threat:
| Threat | Who / backed by | Posture | Proof point |
|---|---|---|---|
| Chinese fibre majors | Yangtze (YOFC), Hengtong, Futong | Largest, lowest-cost, can flood ex-China markets | Global OFC prices set by Chinese supply; the FY22–FY25 glut is what crushed STL’s margins |
| US/EU incumbents | Corning (US), Prysmian (Italy) | Local, tariff-advantaged in the US, deep hyperscaler ties | ”Demand outpacing domestic supply in North America” — but Corning is the default US hyperscaler partner |
| Indian peers | HFCL, Birla Cable, Tejas (adjacent) | Same domestic cycle, similar margins | All sub-7% RoE today — the whole Indian sub-sector earns poorly through the cycle |
The real-world precedent. The 2017–2019 Indian fibre boom is the precedent, and it’s STL’s own. Jio’s national rollout sent fibre demand and STL’s margins to euphoric highs; the stock and the sector were loved; then the orders dried up, China gluts hit, and STL spent four years in the wilderness with two loss years. The incumbent (STL) survived but its returns and share price were eviscerated for years. AI is a bigger, more global driver than Jio was — but the shape of the risk (a demand surge that lifts a commodity, attracts supply, then normalises) is identical, and we have already watched it play out once on this exact company.
The answered follow-on questions.
- Is the damage/upside to volume or to price? Both, but mostly volume. Management says recent growth is “a mix of both,” with volume “a significant part” and richer mix helping — but they won’t split it, and pricing they describe as merely “stable, no decline.” So the boom is largely a volume story on a roughly flat price — which is good for revenue but does little for the structural margin.
- Which segment is protected vs exposed? The leading-edge data-centre connectivity (multi-core, hollow-core, ultra-dense) is the defensible bit; plain OFC is the exposed commodity. Today the defensible bit is only ~20% of revenue (target 30% in 12–18 months) — so most of the business is still the exposed kind.
- Has anyone actually moved yet? Yes — the $1 bn hyperscaler order and the 40% order-intake growth are real, filed facts (HARD). That’s the genuine, encouraging part. But it’s a forward order book to be executed FY27–FY29; it has not yet shown up as durable high returns in the accounts.
- Who’s on the other side? A determined Chinese supply base and a 50% US tariff. The bet is partly against both.
Honest verdict on the crux: Genuinely two-sided, leaning sceptical on the price. The demand is real and STL is well-positioned to ride it — order book and the hyperscaler win prove that. But the moat is narrow, the product is still mostly a globally-priced commodity, the tariff is actively eating the new US margin, and we have a precedent — on this very company — of exactly this kind of boom reverting. I can take a view on the business (a Good-to-Gruesome cyclical with a real but unproven upgrade) but the ten-year return outcome at today’s price is close to “too hard”: it requires both a sustained super-cycle and a structural margin step-up and tariff relief, all at once.
The watch-list
- RoCE crossing 15% for two consecutive years — the single number that would prove the cycle has structurally improved (FY26: 7.75%).
- Enterprise & data-centre revenue mix moving from ~20% toward the 30% target on the promised 12–18-month timeline.
- Reported (not “underlying”) EBITDA margin climbing back above ~15% — i.e. the tariff hit being mitigated, not just explained away.
- The $1 bn hyperscaler order converting to actual purchase orders and revenue from FY27 — watch the order-book execution schedule.
- Net-debt/EBITDA continuing to fall below ~2× as revenue scales.
- An India–US trade deal / tariff resolution — a binary catalyst either way.
QGLP scorecard (the Motilal Oswal lens) — the receipts
| # | Question | Score | Evidence |
|---|---|---|---|
| 1 | Large opportunity? | 1 | AI/data-centre/5G/FTTx fibre TAM is enormous and multi-year (concall) |
| 2 | Favourable industry structure? | 0 | Globally-priced commodity; China can flood it; OPM swung 23%→4%→12% — brutal structure |
| 3 | Defensible moat? | 0.5 | Backward-integrated low-cost + 780 patents, but RoE beat its cost of capital in only 4 of 10 years |
| 4 | Return ratios >15%? | 0 | RoE 2.24%, RoCE 7.75% (FY26); chronically <15% since FY21 |
| 5 | Asset-light? | 0 | Fixed assets ₹2,971 cr; capital-hungry manufacturer, not a See’s-type business |
| 6 | Favourable terms of trade (neg. WC)? | 1 | Working-capital days negative; debtor 82 vs payable 163 days — suppliers fund it |
| 7 | Unquestionable integrity? | 0.5 | No pledging; Big-4 audit; but Vedanta group, TP tax notice, US lawsuit, “minimum remuneration” in a loss year |
| 8 | Proven execution? | 0 | Revenue collapsed FY23→FY24; two loss years; guidance-vs-delivery weak |
| 9 | Growth mindset & vision? | 1 | Heavy R&D, 780 patents, hollow-core/multi-core, AI-DC portfolio |
| 10 | Superior capital allocation? | 0 | Levered up into a downturn; RoE destroyed; economic profit −₹221 cr — one-dollar test fails |
| 11 | Clear succession? | 0.5 | Ankit Agarwal MD; family + professional bench; some key-man/family-control risk |
| 12 | Minority interests protected? | 0.5 | No dividend 3 yrs; promoter trimmed 54%→44%; no pledge, no obvious leakage |
| 13 | Structural tailwind? | 1 | AI data-centre fibre demand projected to grow fast (CRU: NA ~13.7% CAGR to 2030) |
| 14 | Volume-led growth? | 0.5 | ”Mix of both”; largely volume on flat price — cyclical, not pure structural |
| 15 | Operating leverage? | 0.5 | Underlying margin recovering (~18%), but reported dragged to 10.3% by tariff |
| 16 | Manageable leverage? | 0.5 | Net debt ₹1,331 cr, D/E 0.87×, net-debt/EBITDA 2.58× — much improved, not yet low |
| 17 | Market-share gain potential? | 0.5 | ~8% ex-China share, recently flat-to-slipping (8%→7%) |
| 18 | Earnings >15% CAGR? | 0.5 | History awful (10-yr PAT CAGR negative); forward could clear it if the order book delivers |
| 19 | Relevant 10–15 yrs? | 1 | Fibre is the backbone of all digital infrastructure — durable demand |
| 20 | Extend CAP? | 0 | RoE below cost of capital today — there is no spread to extend |
| 21 | Sustain GAP? | 1 | Long runway: FTTx, DC interconnect, 5G densification |
| 22 | Geographic/product headroom? | 1 | Global (US/EU/India), plus copper, connectivity, digital optionality |
| 23 | Adaptive, resilient culture? | 0.5 | Survived a savage cycle and deleveraged — but it was a near-thing |
| Business-quality total | 12 / 23 | Quality 5/12 · Growth 3.5/6 · Longevity 3.5/5 | |
| Price pillar (separate) | 0 / 2 | PEG not meaningful (loss-making history); 5-yr payback ≈ 7–14× |
The pattern is unusually clean: Longevity and the size of the opportunity are real strengths; Quality is the glaring weakness — the return ratios, capital allocation and industry structure all score poorly because this is a low-return, capital-heavy, cyclical commodity. Growth scores middling: the tailwind is genuine but the company’s own delivery history is poor. A 12/23 says “mixed — a Good business that needs capital, or a quality that erodes in the trough.” The Price pillar is a flat zero, and that is the whole story of the share today.
Buffett lens (the Berkshire-letters read)
| # | Test | Result | Evidence |
|---|---|---|---|
| 1 | Good boat? (business > management) | FAIL→PARTIAL | Capital-hungry cyclical; “Good” in good years, “Gruesome” in bad — RoCE 7.75% |
| 2 | Moat / franchise / pricing power | FAIL | Price-taker on most volume; “prices stable, no uptick”; RoE beat CoE 4/10 yrs |
| 3 | See’s test (high returns, little capital) | FAIL | Heavy fixed assets + CWIP; growth eats capital; this is the opposite of See’s |
| 4 | One-dollar test (capital allocation) | FAIL | Levered into a downturn; RoE collapsed; economic profit −₹221 cr |
| 5 | Owner-oriented, candid management | PARTIAL | Articulate, no pledging — but won’t disclose volumes/utilisation/pricing |
| 6 | Integrity / no “credit P&L, debit B/S” | PARTIAL | OCF backs profit reasonably; but Vedanta group, TP notice, US lawsuit, promoter trimming |
| 7 | Circle of competence / predictability | PARTIAL | The product (fibre) is durable and simple; the earnings are wildly unpredictable |
| 8 | Mr. Market — gift or trap now? | FAIL | P/E ~670, P/B 14×, up 8× in a year — priced for euphoria, not fear |
| 9 | Patience / compounding runway | PARTIAL | Long demand runway, but no high-RoE base to compound from yet |
| 10 | The honest red flag | (see below) | — |
| PASS count | ≈ 3 / 10 | Two FAILs softened to PARTIAL; the spirit is “not in the temple at this price” |
The See’s test, spelled out. See’s Candies earned a fortune on almost no reinvested capital — that’s a Great business. STL is the mirror image: to grow it must keep building plants and buying germanium-grade glass kit, and in a bad year it spends just to survive. Free cash flow has been positive in some years (FY24 ₹544 cr, FY26 ₹346 cr) but negative or thin in others — there is no reliable cash fountain. It fails See’s clearly.
The one-dollar test, spelled out. Over the last decade STL retained nearly all its earnings (no dividend for three years) and reinvested heavily. Did each retained rupee create a rupee of market value for the business? On fundamentals, no — net worth barely grew from ~₹1,720 cr (FY19) to ₹2,268 cr (FY26) while two of those years lost money, and economic profit is negative. The share price has soared, but that is Mr. Market’s mood, not retained-earnings compounding. By Buffett’s strict test — book and economic value created per rupee retained — this fails.
Test 10 — the honest red flag. The single strongest reason this is not a wealth creator: it is a globally-priced commodity manufacturer that has already shown, on its own numbers, that its high returns evaporate the moment the cycle turns — and the market is currently paying a 670× earnings multiple as if that will never happen again. The numbers support this bear point, not refute it: RoE 2.24%, RoCE 7.75%, two recent loss years, negative economic profit. The bull must argue that AI changes the cycle’s nature permanently — a claim the company’s own “hollow-core is years away, no standard yet, prices stable” commentary only half-supports.
The framework metrics
- Economic Profit = ₹2,268 cr × (2.24% − 12%) = −₹221 cr. Destroying value: STL earns far less than the cost of its owners’ money. (Even on a generous normalised RoE of ~15%, EP would be only ~+₹68 cr — barely above the hurdle.)
- Terms of Trade ≈ debtor days 82 / payable days 163 = ~50% — favourable; working-capital days are negative. A genuine plus.
- 5-yr Payback = ₹31,943 cr ÷ projected 5-yr cumulative PAT. Even assuming a normalised ₹340 cr starting PAT growing 15–25%, cumulative 5-yr PAT ≈ ₹2,300–2,800 cr → payback ≈ 11–14×; a heroic ₹500 cr start at 30% still gives ≈ 7×. The multi-bagger bar is < 1×. (Assumes margin recovery to ~20% EBITDA — itself optimistic.)
- PEG = not meaningful — earnings were negative in FY24–FY25, so the P/E of 671 has no honest growth denominator.
- RoE − CoE spread = 2.24% − 12% = −9.8% today; RoE exceeded 15% in only 4 of the last 10 years (all FY17–FY20).
- Consistent vs Volatile test = FAIL → Volatile. Net profit fell more than 10% in multiple years and went negative twice; this is a value-on-P/B, not value-on-P/E, business — and it trades at 14× book.
Peer comparison
| Company | Mkt cap (₹ cr) | CMP (₹) | P/E | P/B | RoE | RoCE | OPM (latest yr) | Sales (₹ cr) |
|---|---|---|---|---|---|---|---|---|
| Sterlite Tech (STLTECH) | 31,943 | 654 | 671 | 14.1× | 2.24% | 7.75% | 12% | 4,745 |
| HFCL | 32,083 | 210 | 103 | 6.6× | 6.95% | 10.9% | ~10% | — |
| Birla Cable | 682 | 227 | 40.4 | 2.4× | 6.33% | 8.99% | — | — |
| Tejas Networks | 11,246 | 632 | n/a (loss) | 3.8× | −26.8% | −14.6% | loss | — |
The peer table tells you two things. First, the whole Indian optical/telecom-equipment sub-sector earns poorly — every name here is sub-7% RoE, confirming this is a structurally low-return, cyclical patch of the market, not a one-company stumble. Second, on every value yardstick STL is the most expensive of the lot — 671× earnings and 14× book against HFCL’s 103×/6.6× and Birla Cable’s 40×/2.4×, despite worse current returns than HFCL. STL has the strongest franchise of the group (largest, lowest-cost, most patents, the marquee hyperscaler order) — so a premium is fair — but the size of the premium is extreme even relative to its own asset class. The relative read and the absolute read agree for once: dear.
Latest quarter & what’s happening now
Q3 FY26 (reported 23 Jan 2026): revenue ₹1,257 cr (up YoY), EBITDA ₹129 cr at a 10.3% margin — dragged down ~760 bps by the US tariff; underlying margin ~18%. Profit before exceptional items turned positive (₹9 cr) vs a ₹78 cr loss a year earlier. FY26 full year (Mar 2026): sales ₹4,745 cr, net profit ₹56 cr — back in the black after two loss years. (Q3 FY26, reported 2026-01-23; FY26 from screener.)
Concall takeaways: (1) order intake up 40% YTD to ₹4,263 cr; open order book ₹5,325 cr; (2) the over-$1 bn hyperscaler AI-datacentre supply deal (announced 22 May 2026, FY27–FY29) is the catalyst behind the share’s vertical move (HARD — filed); (3) management targets the enterprise/DC mix rising from ~20% to 30% in 12–18 months and an eventual 20% EBITDA margin “at 70%+ utilisation” (MEDIUM — guidance); (4) tariff mitigation via US local production and contract repricing is “2–3 quarters” away (MEDIUM).
Where the two lenses agree — and disagree
For once, the two lenses agree almost completely — and that agreement is itself the signal. QGLP scores it 12/23 (mixed, Quality the weak link); Buffett passes only ~3/10. Both flag the same things: a low-return, capital-heavy, cyclical commodity (Q2/Q4/Q5 fail; Buffett tests 1–4 fail), with a real opportunity and runway (Q1/Q19/Q21 pass; Buffett test 9 partial). There is no interesting divergence on quality — both say “Good-tilting-Gruesome, not a compounder yet.” The only place they could be said to part is optimism about the future: QGLP gives partial credit for the growth tailwind (Q13/Q18), while the Buffett predictability test (7) is harsher because the earnings are so unpredictable even if the product is durable. When both your quality lenses land in the same unflattering spot, the burden falls entirely on price — and price fails too.
The price as a current phenomenon
This section judges the price, not the business — the business verdict above is already settled. Here we only ask: what is Mr. Market charging today, and is that a passing mood?
The margin-of-safety band: QGLP’s price pillar wants a 5-yr payback ≤ 1× or PEG ≤ 1×. PEG is meaningless (no clean earnings base). On payback, even a generous normalised-earnings projection (₹340–500 cr PAT growing 15–30%) only justifies a market cap of roughly ₹6,000–12,000 cr for a 1× payback — i.e. a share price band of very roughly ₹130–₹230. Today’s ₹654 is three to five times above that band. (Assumptions: CoE 12%; normalised EBITDA margin ~18–20%; these are themselves optimistic.)
The Mr. Market read: the crowd is greedy on this name right now, and the reason is identifiable — a single >$1 bn AI-datacentre order plus the broader AI-capex frenzy that is bidding up anything with “fibre” and “AI” in the same sentence. That is a narrative re-rating, not an earnings re-rating: the company earned ₹56 cr and is valued at ₹31,943 cr. Narrative moods can persist for a while in a hot theme — but they can also halve in weeks on a single soft quarter, a tariff surprise, or a Chinese price move, without one fact in Box 1 changing.
The plain statement of the tension: a Great business can sit at an unwonderful price, and a Gruesome one can be a bargain. This is the rarer third case: a Good-to-Gruesome business sitting at a price built for a Great one. The quality doesn’t justify the multiple, and the multiple doesn’t leave a margin of safety if the cycle behaves the way it always has. This reading can change next week without a single thing in the business above changing — that is exactly why we keep the two boxes apart.
Conviction texture
The bull case, at its strongest: AI data centres are a structural, multi-year demand shock that genuinely needs far more fibre (up to 36× per rack), STL is India’s lowest-cost integrated maker with a real technology lead in dense and multi-core cable, it has just won a marquee >$1 bn hyperscaler order, it has deleveraged from its near-death balance sheet, and margins are recovering underneath the tariff noise. If this is a true super-cycle and STL structurally moves up the value chain to 30%+ high-margin mix, the company could earn far more than its trough numbers suggest — and the stock would then look forward-cheap rather than absurd.
The bear case, at its strongest (and test 10’s red flag): this is a globally-priced commodity manufacturer that has already demonstrated its high returns evaporate when the cycle turns — RoE went from 34% to negative inside five years — and the market is paying 670× earnings and 14× book as if that history is irrelevant. Most volume is still commodity fibre where China sets the price; the 50% US tariff is actively eating the new American margin; and the genuinely differentiated products (hollow-core) are years from scale. The precedent for “fibre boom on this exact company” ended in four lost years.
What the numbers actually support: a real, encouraging operational recovery on a structurally low-return, cyclical base — and a share price that has run far ahead of it. The business is back to thin profit; the price assumes durable, high-return compounding that the accounts have not yet shown. The three things to watch that would tip it: RoCE crossing and holding above 15%; the DC/enterprise mix actually reaching ~30% on schedule; and the $1 bn order converting into reported revenue and reported (not “underlying”) margins above 15%. Until those land, the gap between Box 1 and Box 2 is the entire story — and it is wide.
No buy/sell/hold. The deliverable is the quality verdict, the price band, and both sides of the argument; the reader decides.
Sources
- Screener.in: https://www.screener.in/company/STLTECH/consolidated/ (snapshot fetched 2026-06-20)
- Q3 FY26 earnings call transcript, 23 Jan 2026 (BSE filing); Q2 FY26 call, 6 Nov 2025
- Annual Reports FY25 and FY24 (chairman/board profiles, remuneration, segment, RPT/governance sections)
- STL >$1 bn hyperscaler AI-datacentre supply deal — STL press release & Business Standard, 22 May 2026 (HARD)
- “Up 351% in 2026… on $1 bn AI order,” Business Standard, May 2026 (price-run context)
- Sterlite Power demerger (2016) & Vedanta/Anil Agarwal debt history — Business Standard, Wikipedia, M&A Critique
- “Ankit Agarwal confirms no encumbrance on STL shares in FY26” — ScanX (no-pledge confirmation)
- Peer snapshots: HFCL, Birla Cable, Tejas Networks (screener.in, fetched 2026-06-20)
- Assumptions: Cost of equity 12%; normalised EBITDA margin 18–20% and PAT ₹340–500 cr growing 15–30% for the payback band — all optimistic, stated to be transparent.