Netweb Technologies — India's only homegrown AI-server maker
Netweb Technologies India Ltd
Snapshot
Netweb designs and builds high-end computers in India — supercomputers, private-cloud boxes, and (now mostly) the big GPU servers that run artificial intelligence. Think of it as the country’s only homegrown maker of the heavy “engines” that power AI, instead of importing finished machines from Dell or HPE. Market cap is about ₹28,817 crore, the share trades around ₹5,061 (52-week range ₹1,733–₹5,244), on a P/E of 140 and nearly 40 times book value. Return on equity is a glittering 32.8% and return on capital 37.5%. In one phrase: a young, fast-growing, asset-light franchise riding a once-in-a-generation tailwind — priced as if nothing can go wrong.
As of 2026-06-20, from screener snapshot.
The verdict in two boxes — the business first, the price second
Box 1 — The business (durable):
| Lens | Result |
|---|---|
| Business-quality score | 17.5 / 23 (Quality 9.5/12 · Growth 5.5/6 · Longevity 2.5/5) |
| Buffett rubric | 6.5 / 10 PASS |
| Business bucket | Great (high returns, light capital — a true cash-light compounder) |
| Wealth-creator type | Enduring-leaning · Volatile (lumpy, order-driven; only 7 years of public record) |
| Economic Profit | +₹150 cr (RoE 32.8% − CoE 12% on ₹723 cr net worth) — clearly creating value |
A Great business that is a wealth creator — it earns far more than the cost of its owners’ money, on very little capital, and has a long runway. The open questions are durability (will the boom last and the moat hold?) and that it has only a short public track record — independent of what the share costs today.
Box 2 — The price today (a current phenomenon):
| Reading | Result |
|---|---|
| CMP | ₹5,061 (as of 2026-06-20) |
| Price pillar | 0 / 2 (PEG ≈ 3.8x · 5-yr payback ≈ 10x) |
| Margin-of-safety band | ₹1,300–₹2,000 (where PEG approaches ~1x on 35–40% growth / payback gets sane) |
| Mr. Market’s mood now | Greedy / euphoric — priced for flawless multi-year AI growth |
| CMP vs the band | Demanding — roughly 2.5–4x above where the framework’s price discipline is satisfied |
Today the market is pricing it rich — a mood driven by the AI-infrastructure euphoria and Netweb’s “only Make-in-India full-stack” scarcity story, which can change in a quarter while the business above does not.
In plain English
Imagine the gold rush of AI. Everyone is buying shovels — the racks of computers stuffed with NVIDIA chips that train and run AI. Most countries buy those shovels ready-made from Dell, HPE, Lenovo or Supermicro. India decided it wants to build its own, on its own soil, owned and run at home — what the government calls “sovereign compute.” Netweb is the one listed Indian company that actually designs and manufactures those machines end-to-end: the boards, the boxes, the liquid cooling, and the software stack that ties it all together. When the Indian government’s AI Mission hands out money for home-grown AI computers, Netweb is the natural Indian maker to call. That is the whole story in a sentence.
Is it a good boat? Yes — a surprisingly good one. Netweb earns about ₹37 of operating return for every ₹100 of capital it employs, and it does this without swallowing cash: a single ₹35-crore factory, commissioned in May 2024, can support roughly ₹2,000–3,000 crore of sales. Management is blunt that it is a capability company, not a capacity one — it sells brains and design, not tonnage. The result is the rare combination Buffett prizes: high returns on little capital, and a balance sheet with no real debt. Sales nearly doubled to ₹2,184 crore in FY26, profit grew 81% to ₹206 crore, and the AI-systems segment alone exploded 460% to become 43% of the business. This is genuinely a Great business as the masters define it.
Where is the moat? It is a young one, and it is mostly made of two things. First, Make-in-India procurement rules and the politics of sovereign compute — a government building its own AI brain does not want to import the engine from a foreign giant, and that preference is a real wall around the ~70% of revenue that is sovereign and large-enterprise. Second, fifteen years of design know-how with NVIDIA, AMD and Intel — Netweb sees chip roadmaps 18 months early and ships home-grown servers on the latest Blackwell platform at the same time the world’s biggest OEMs do. That is not nothing. But it is not the unbreakable kind of moat either. Lenovo has already announced it will build AI servers in India under the same PLI scheme — the clearest sign a global giant is trying to climb the very wall Netweb stands behind.
What’s happening right now is a flood. India’s AI Mission is spending over ₹10,000 crore, scaling toward tens of thousands of GPUs and a nine-gigawatt datacentre ambition by 2030. Netweb walked into FY27 with a ₹2,400-crore order book — more than all of last year’s sales — plus a ₹4,400-crore pipeline. Management guides 35–40% organic growth for two-to-three years and means it. The catch is two-fold and both are real: a worldwide memory-chip shortage (DRAM prices roughly doubled in early 2026) is squeezing the cost of every machine Netweb builds, and margins already slipped — the big “strategic” government orders run about two percentage points thinner, and the stock fell 7% on its strong-but-lower-margin Q4 print.
And that is the one-line tension. This is a wonderful little business — but you are being asked to pay 140 times earnings and 40 times book for it. At that price, you are not buying the company that exists; you are pre-paying for years of flawless 40% growth that has not yet happened. The business and the price are two completely different conversations, and this report keeps them apart on purpose.
Sitting down with the management
Dear reader — if you sat across the table from Sanjay and Navin Lodha, you’d meet builders, not financiers. Sanjay started Netweb as a one-man proprietorship in 1996 and incorporated the company in 1999; he and his brother have run it for nearly three decades. On paper they hold commerce and management degrees, not engineering ones — but the company they built is deeply technical, and the fifteen-year working relationship with NVIDIA is the proof. When an analyst asked Sanjay on the January 2026 call what could go wrong, he answered, candidly, “I’ll have to spend a night to think about it” — then named black-swan events but no business risk he was willing to own. That is the one thing that gives me pause about the tone: a little too sunny, a little too sure. Genuinely candid managers can usually name their own weak spot without a night’s notice.
How have they spent the owners’ money? Well, and sparingly. The signature decision was a single ₹35-crore factory (May 2024) explicitly sized to support ₹2,000–3,000 crore of revenue with no further big capex — capital-light to the bone. They pay a small dividend (₹3 a share for FY26, an ~8% payout) and reinvest the rest at 30%-plus returns, which is exactly right while the runway is this long. There are no value-destroying acquisitions, no pledged shares, and no net debt (a ₹282-crore year-end borrowing is short-term working capital for a large order, and it sits against more cash). On the one-dollar test — has each rupee retained created at least a rupee of value? — the answer is a clear yes: book value has compounded and so has the market’s regard for it.
Do they talk straight on the numbers? Mostly. The CFO voluntarily walks through hedging gains and mark-to-market losses, explains why short-term borrowings spike against RBI loan-covenant rules, and admits the strategic orders carry ~200 basis points lower margin. That is owner-style disclosure. The forensic red flags I look for — profit that never becomes cash, a ballooning balance sheet, auditor games — mostly do not fire. Operating cash flow was a healthy ₹171 crore against ₹206 crore of profit in FY26. The honest caveat: in two of the last six years operating cash flow was negative because this is a lumpy, working-capital-heavy order business — so cash conversion is good on average but bumpy year to year. Watch it, don’t fear it.
The two real concerns are these. One — promoters trimmed near the top. In February 2026 all four Lodha promoters sold a combined ~3.65%, taking the family from ~70.6% to ~67.0%. They say it was to add float and liquidity to a thin stock, and they’ve committed to no further selling for twelve months. That’s a plausible reason at a 140 P/E, but four insiders selling at record highs is always worth a note in the ledger. Two — key-man risk. This is materially a Sanjay-Lodha-led company; the bench is thinning out (a CFO, a sales head, a strategy officer were formalised in FY26), but the vision and the customer relationships still run through one man.
Would Buffett and Agrawal shake hands on this management? Probably yes — they are competent, honest-enough technologist-owners with skin in the game and a clean balance sheet. What would change their mind: a second round of promoter selling inside the no-sell window, or any sign the candor curdles into spin as growth inevitably slows.
What’s on the horizon (live-issues tracker)
1. The India AI Mission order wave — the crux (interrogated below). 🟢 On track.
2. The memory-chip shortage squeezing margins. 🟡 Mixed / live. DRAM and high-bandwidth-memory prices roughly doubled in early 2026 as the whole world raced to build AI machines. Memory is a big slice of what goes inside a Netweb server, so this directly lifts the cost of every box. Management says its long supplier relationships give it priority access and that it can pass costs on to new orders — and so far it has delivered without disruption. But this is the concrete reason margins are under pressure now, and it is outside Netweb’s control. Watch: EBITDA margin holding the guided 13–14% through FY27. (HARD — TrendForce/CNBC, Jan 2026; management commentary, Jan & May 2026 calls.)
3. Margin trajectory as the mix tilts to government. 🟡 Early warning. Adjusted operating EBITDA margin slipped from 13.8% (FY25) to 13.3% (FY26), and the stock fell ~7% on the Q4 print specifically on margins despite 90% revenue growth. Management is honest that strategic government orders run ~200 bps thinner at the profit-before-tax line. At 13% Netweb still earns more than global box-makers (Dell/Supermicro server margins are ~5–10%), so it has a real premium — the risk is that premium erodes as it chases low-margin sovereign tonnage. This is the single most watchable line in the whole thesis. (HARD — company filings & Business Standard, 4 May 2026.)
4. Exports and quantum — optionality, not yet real. 🟡 Too early. Management openly says exports stay at ~5% because domestic demand is “unabated” and they can’t get to it; a quantum-computing effort exists but is undisclosed “for competitive reasons.” Genuine future runway, zero current contribution. Don’t pay for it yet.
The crux — interrogated, not just named
The crux in one sentence: Netweb’s investment case works if and only if it can hold its place as India’s go-to homegrown AI/HPC server maker as the India AI Mission spends — without global giants undercutting it, the AI capex proving a bubble, or margins collapsing on low-margin government orders.
The mechanism, in plain words. Why does anyone buy a Netweb server instead of a cheaper Dell? Two reasons, and they’re different in kind. The first is a rule — Indian public-procurement preference (“Class-I local supplier” / Make-in-India) and the political logic that a country building its own AI brain wants the engine made and owned at home. The second is skill — Netweb actually designs and integrates the full system, not just screws a box together. Here’s the test of how strong that is: a rule-moat can be legislated away or climbed over by a rival who localises; a skill-moat cannot. Netweb has both, but its biggest, fattest slice of revenue (the ~70% sovereign book) leans on the rule. That’s the part to worry about.
The competition, by name.
| Threat | Who backs it | India posture | Proof point |
|---|---|---|---|
| Lenovo | Global #3 server OEM | Announced AI-server manufacturing in India under PLI + an India AI R&D lab | The clearest active attempt to climb the Make-in-India wall (HARD — AICerts, 2025) |
| Dell / HPE / Supermicro | Global #1/#5/#2 | Sell into India but import finished boxes; thin 5–10% margins | Win commercial/enterprise, weak on sovereign preference (HARD — IDC shares, 2025) |
| Cloud players (Yotta, E2E, NxtGen, Jio) | Empanelled in India AI Mission | Customers, not rivals — they rent GPUs; Netweb builds the servers they run | A common misread; they sit one layer above Netweb (HARD — Inc42/eGov, 2025) |
The reframe that matters: the headline “India AI” winners (Yotta, E2E) are buyers of the layer Netweb sells, not competitors. Netweb’s real contest for share is Lenovo localising, not the cloud providers. As of mid-2026, Netweb is still winning the marquee tenders — a ₹1,734-crore sovereign AI compute order in September 2025 is the proof point that it’s holding. (HARD — Business Standard, 3 Sep 2025.)
The real-world precedent — and it cuts both ways. China’s Inspur is the near-perfect parallel: sovereign procurement built it into the world’s #3 server maker. That’s the bull case made real — the model works while the home government keeps spending. But Inspur’s lesson for the bear is sharper: it was a domestic champion built on imported foreign silicon, and when the US blacklisted it in 2023 it was cut off from the very NVIDIA/Intel chips it integrated and had to rebrand. The vulnerability wasn’t competition — it was being a middleman dependent on someone else’s chips. Netweb’s analogous tail risk: any tightening of US export rules toward India, or NVIDIA favouring bigger global OEMs when chips are scarce, hits the integrator hardest. (HARD — DataCenterDynamics / Washington Post.)
The follow-on questions, answered. Is the AI demand a bubble? As of mid-2026 the India AI Mission spend is not slowing — but credible voices (Michael Burry, May 2026) warn the global buildout understates depreciation and overstates earnings, and circular financing among NVIDIA/OpenAI/Oracle is real. The nearer, more concrete hit is the memory shortage, not a demand collapse. Which segment is hit first? Enterprise (non-government) demand would soften first in a bubble pop; the government book is the most protected. Is anyone actually switching away from Netweb yet? No — it’s still winning sovereign tenders.
Honest view: This is not “too hard.” Netweb is genuinely holding its position on the strength of a real (if young) moat, and is winning the orders. The crux that will decide the next two years is not competition and not the bubble — it is margin discipline: can it grow 40% on government tonnage and keep EBITDA at 13–14%? That’s a knowable thing to watch, quarter by quarter.
The watch-list:
- EBITDA margin — does it hold ≥13% through FY27, or drift toward the global 5–10% band?
- A second promoter sale inside the Feb-2026-to-Feb-2027 no-sell window — would be a real tell.
- Lenovo (or any global OEM) winning an India sovereign/government AI tender — the first crack in the moat.
- India AI Mission Tranche-2 orders landing on Netweb’s books (the ~₹1,200-cr RFP) — confirms the runway.
- Operating cash flow staying positive and ≥70% of profit — proves the growth is real cash, not paper.
- Memory-cost pass-through — any quarter where margin breaks below guidance because of input costs.
QGLP scorecard (the Motilal Oswal lens) — the receipts
Score each line 0 / 0.5 / 1. RoE = profit earned per ₹100 of owners’ money; OPM = operating profit margin; PEG = price-to-earnings divided by growth rate.
| # | Question | Score | Evidence |
|---|---|---|---|
| Quality of Business | 5.0/6 | ||
| 1 | Large opportunity? | 1 | Indian HPC market US$644m (FY25)→US$919m (FY29); AI-server TAM far bigger and exploding. Sovereign compute a generational theme. (AR FY25) |
| 2 | Favourable industry structure? | 1 | Netweb is the only listed Indian full-stack server OEM; sovereign preference limits rivalry. OPM stable 13–16%. |
| 3 | Defensible moat? | 0.5 | RoE > CoE every year of its short record, but moat is young + part rule-based; Lenovo localising. Real but unproven over a cycle. |
| 4 | Return ratios > 15%? | 1 | RoE 32.8%, RoCE 37.5%; RoCE 31–64% every year FY20–FY26. (ratios) |
| 5 | Asset-light? | 1 | One ₹35-cr factory supports ₹2,000–3,000 cr sales; “capability not capacity.” Gross fixed-asset turn 33x. (May 2026 call) |
| 6 | Favourable terms of trade? | 0.5 | ToT ≈ 70% (Debtor 112d / Payable 161d) is favourable, but inventory days rose to 170 and cash cycle to 84–121d — working-capital-heavy. (ratios) |
| Quality of Management | (within Quality 12) | ||
| 7 | Integrity / clean accounts? | 1 | No pledging, OCF ₹171cr vs PAT ₹206cr FY26, no auditor flags, no forensic concerns found. (web) |
| 8 | Proven execution? | 1 | 90% revenue growth FY26; delivered to guidance, preponed a strategic order from Q4 to Q3. (calls) |
| 9 | Growth mindset? | 1 | Heavy R&D, NVIDIA Blackwell day-one, new facility, quantum optionality. |
| 10 | Superior capital allocation? | 1 | No dilution post-IPO, no bad M&A, reinvests at ~37% RoCE, tiny capex. One-dollar test passed. |
| 11 | Succession plan? | 0.5 | Senior layer formalised FY26 but materially a Sanjay-Lodha company. Key-man risk. (web) |
| 12 | Minority interests protected? | 0.5 | Small dividend, fair pay, no leakage — but Feb-2026 promoter sale near highs is a watch-item. |
| Growth | 5.5/6 | ||
| 13 | Structural tailwind ≥1.5× GDP? | 1 | AI/sovereign-compute growing far faster than nominal GDP; India AI Mission >₹10,000cr. |
| 14 | Volume-led (durable)? | 1 | Growth is unit/order-led on new AI systems, not price; AI segment +460%. |
| 15 | Operating leverage? | 0.5 | OPM broadly flat 13–14% (slipped FY26) — leverage masked by lower-margin strategic mix. |
| 16 | Manageable leverage? | 1 | Zero net debt; borrowings are transient working capital. (balance sheet) |
| 17 | Market-share gain? | 1 | Gaining share as the default domestic AI/HPC OEM in a booming category. |
| 18 | Earnings growth > 15%? | 1 | PAT CAGR ~64% (3yr) / ~91% (5yr); guided 35–40% forward. (profit_loss) |
| Longevity | 2.5/5 | ||
| 19 | Relevant in 10–15 years? | 1 | Compute demand is structural; the form (which chips, training vs inference) will change but the need won’t. |
| 20 | Extend Competitive Advantage Period? | 0.5 | Skill-moat can widen with R&D; rule-moat (procurement) is policy-dependent and contestable (Lenovo). |
| 21 | Sustain Growth Advantage Period? | 1 | Huge TAM, low penetration, multi-year order book — long runway. |
| 22 | Geographic / product headroom? | 0.5 | Exports ~5% (untapped), quantum optionality — real but not yet delivering. |
| 23 | Adaptive culture through cycles? | 0 | Only ~7 years of public record (IPO 2023) and no downturn tested. Cannot yet credit resilience. |
| Business-quality total | 17.5/23 | Quality 9.5 · Growth 5.5 · Longevity 2.5 | |
| Price (reported separately) | 0/2 | ||
| 24 | Valuation reasonable (PEG)? | 0 | PEG ≈ 140 / 37 ≈ 3.8x. (ratios + guidance) |
| 25 | Margin of safety (PEG<1 or payback<1)? | 0 | 5-yr payback ≈ 10x; both bars failed badly. |
| Canonical QGLP /25 (for fidelity) | 17.5/25 | Headline is the 17.5/23 business score; Price 0/2 reported apart. |
Pillar pattern: Quality and Growth are the strength — this is a textbook high-return, light-capital, fast-growing franchise. Longevity is the genuine gap, and honestly so: a 7-year-old listed company in a boom it hasn’t seen end, with a young and partly policy-based moat, simply hasn’t earned the durability points yet. Price is the only thing standing between a 17.5 and a clear buy-zone — and right now it stands a long way out.
Buffett lens (the Berkshire-letters read)
| # | Test | Result | Evidence / Buffett line |
|---|---|---|---|
| 1 | Good boat (Great/Good/Gruesome)? | PASS | Great: ~37% RoCE on a single small factory — high returns, little capital. |
| 2 | Moat + franchise + pricing power? | PARTIAL | Stable OPM + RoE > CoE every year, but moat young and part rule-based. “competitors will assault any castle earning high returns.” |
| 3 | See’s test (high returns, little capital)? | PASS | ₹35-cr factory supports ₹2,000–3,000cr sales; capability not capacity. |
| 4 | Capital allocation (one-dollar test)? | PASS | RoE sustained through reinvestment; no dilution, no bad M&A, tiny capex. |
| 5 | Owner-oriented, candid management? | PARTIAL | Honest CFO disclosure, but CMD’s “what could go wrong? I’ll think overnight” is a touch too sunny; Feb-26 promoter sale. |
| 6 | Integrity / no “credit P&L, debit balance sheet”? | PASS | OCF ₹171cr ≈ PAT ₹206cr FY26; no pledging, no auditor flags. (Caveat: cash flow lumpy historically.) |
| 7 | Circle of competence / predictability? | PARTIAL | ”if there’s lots of technology, we won’t understand it.” Chip cycles, ASICs, the bubble question make 10-yr shape harder to call. |
| 8 | Mr. Market — gift or trap now? | FAIL | PEG ~3.8x, P/E 140, ~40x book, euphoric. Priced for perfection. |
| 9 | Patience / compounding runway? | PASS | Large TAM, low penetration, multi-year order book — a long reinvestment runway at high RoE. |
| 10 | The honest red flag | (prose below) | See “Conviction texture.” |
Score: 6.5 / 10 PASS (PARTIALs at ½). A real business with real, identifiable gaps — chiefly predictability and (today) price.
The See’s test, spelled out. See’s Candy needed only $32m of reinvestment over 35 years and threw off $1.35bn. Netweb is cut from similar cloth: one ₹35-crore factory underwrites ₹2,000–3,000 crore of sales, and the company refuses to become a capacity-led “box pusher.” That is the asset-light engine Buffett loved — high returns that don’t get eaten by the capital needed to grow.
The one-dollar test, spelled out. Has each retained rupee created at least a rupee of market value? Emphatically yes — net worth compounded from ~₹50 crore pre-IPO to ₹723 crore, RoE held above 30% through that growth, and the market value sits at ₹28,817 crore. The danger isn’t the past dollar; it’s the next one: at a 140 P/E the market has already booked years of perfect future reinvestment.
The framework metrics
- Economic Profit = Net Worth ₹723 cr × (RoE 32.8% − CoE 12%) = +₹150 cr → strongly creating value (a top-quintile economic-profit signature). (CoE = 12%, the mid-point the studies use.)
- Terms of Trade = Debtors ÷ Creditors ≈ 112d / 161d ≈ 70% → favourable (suppliers fund more than customers owe), though inventory days (170) make the overall cash cycle heavy.
- 5-yr Payback = Mcap ₹28,817 cr ÷ projected cumulative 5-yr PAT (~₹2,764 cr at 35% growth from ₹206 cr) ≈ 10.4x → far above the 1x multi-bagger bar.
- PEG = P/E 140 ÷ forward growth ~37% ≈ 3.8x → price discipline badly failed (PEG < 1 is the bar).
- RoE − CoE spread = 32.8% − 12% = +20.8%, and RoE has been > 15% in every year of the public record (all 6 of the last 6) — a wide, but short-dated, “uncommon profit.”
- Consistent vs Volatile = PAT rose every year FY19→FY26 with no falls — looks Consistent, but the record is only ~7 years, the business is order-lumpy, and quarterly profit swings hard. Treat as Volatile until a cycle is seen (value it on growth-adjusted earnings, with humility).
Peer comparison
There is no true listed twin — Netweb is India’s only full-stack high-end server designer-manufacturer. The closest valuation comparators sit in electronics manufacturing / design, so this table answers “how does the market price Indian electronics-manufacturing quality?” not “who are its product rivals” (its real rivals — Dell, HPE, Lenovo — are foreign).
| Company | Mkt cap (₹cr) | CMP (₹) | P/E | P/B | RoE | RoCE | OPM | FY26 Sales (₹cr) |
|---|---|---|---|---|---|---|---|---|
| Netweb | 28,817 | 5,061 | 140 | 39.8 | 32.8% | 37.5% | 13% | 2,184 |
| Dixon Technologies | 76,461 | 12,517 | 53 | 16.3 | 37.4% | 42.0% | 4% | 48,873 |
| Kaynes Technology | 21,807 | 3,253 | 60 | 4.6 | 9.6% | 13.2% | 16% | 3,626 |
| Syrma SGS | 25,724 | 1,334 | 80 | 9.0 | 13.9% | 16.7% | 11% | 4,819 |
All from each peer’s screener snapshot, FY26. Read it two ways. On returns, Netweb is the class of the field on margin quality — 13% operating margin with 37% RoCE beats the thin-margin contract manufacturers (Dixon runs 4% OPM at huge scale; Kaynes/Syrma earn single-digit-to-teens RoE). Netweb sells design and brains, not assembly tonnage — and the numbers show it. On price, Netweb is the most expensive of the lot — 140x earnings and ~40x book versus a 50–80x / 5–16x peer set that already isn’t cheap. So even relative to a richly-valued asset class, Netweb is the dearest, and the relative read does not rescue the absolute one here. Its distinctive edge — the only sovereign-compute pure-play — is exactly what the premium is paying for.
Latest quarter & what’s happening now
Q4 FY26 (reported 4 May 2026): revenue ₹773.7 cr (+86.6% YoY), PAT ₹70.6 cr (+65.7%), 9% PAT margin. Full-year FY26: revenue ₹2,184 cr (+90%), PAT ₹206 cr (+81%), RoE 32.9%, zero net debt, ₹3/share dividend. The defining feature was AI systems hitting 43.4% of revenue (+460% YoY). Two concall takeaways: (1) entered FY27 with a ₹2,400-cr order book (incl. L1) — bigger than all of FY26’s sales — plus a ₹4,400-cr pipeline (~60% converts over 18–24 months); guidance 35–40% organic growth, 13–14% EBITDA, no major capex. (2) Margins are the sore spot — the stock fell ~7% on the print despite the blow-out top line, because strategic government orders run 200 bps thinner and memory costs are climbing. Q4 FY26, reported 2026-05-04 — HARD. Live catalysts: India AI Mission Tranche-2 RFP (₹1,200 cr, open — SOFT/MEDIUM); strategic orders to execute over the next 3 quarters (MEDIUM).
Where the two lenses agree — and disagree
They agree on the big picture: a Great, asset-light, high-return business (QGLP Quality 9.5/12; Buffett tests 1, 3, 4 PASS) creating real economic profit — and a price that fails the discipline of both (QGLP Price 0/2; Buffett test 8 FAIL).
They diverge in one revealing place: QGLP’s checklist hands Growth a near-perfect 5.5/6, but Buffett’s predictability test (7) only PARTIALs. A checklist rewards the visible 40% growth and full order book; the letters ask the harder question — can you confidently say what this business looks like in ten years? And here you can’t, fully: the AI cycle could cool, the chip mix (GPUs → ASICs) could shift, and the moat’s policy-based half could be eroded by a localising Lenovo. The divergence is the signal: the growth is real and bankable for two-to-three years; the decade is genuinely less certain. Trust the Buffett caution on durability — it’s why Longevity scores only 2.5/5.
The price as a current phenomenon
This section judges the price, not the business. The business verdict above — a Great, value-creating compounder — does not move with what follows.
The margin-of-safety band. The framework’s price discipline (PEG ≤ 1x, payback ≤ 1x) is satisfied nowhere near today’s quote. To get PEG toward ~1x on a generous 35–40% forward growth, the P/E would need to fall to roughly 40–50x — implying a share price band of about ₹1,300–₹2,000 (close to where it traded just a year ago; the 52-week low is ₹1,733). That is not a target or a forecast — it’s simply the arithmetic of where a patient QGLP/Buffett buyer’s price rule turns green. Today’s ₹5,061 sits roughly 2.5–4x above that band.
Mr. Market’s mood. Right now the crowd is greedy on this name, and for understandable reasons: it is the only listed way to own India’s sovereign-AI build-out, the order book is bursting, and scarcity plus a great story is catnip in a bull market. That is precisely the condition Buffett warned about — “be fearful when others are greedy.” The euphoria is real, the business behind it is also real, and both things are true at once.
The plain tension. This is the classic case of a wonderful business at an unwonderful price. Nothing in Box 1 says “avoid”; everything in Box 2 says “expensive.” A 140 P/E prices in years of flawless 40% compounding with no margin slip, no bubble, no competitive crack — a lot of perfection to pre-pay for. And remember: this reading can flip next week without one thing in the business changing — a market wobble, a soft quarter on margins, or a sector de-rating could hand a patient owner a very different price for the same fine company.
Conviction texture
The bull case, at its strongest. India has decided to build its own AI brain, will spend tens of thousands of crores doing it, and Netweb is the single homegrown company that designs and makes the machines — protected by procurement rules and 15 years of NVIDIA-grade design skill. It earns ~37% on capital with almost no capital, carries no debt, has an order book bigger than last year’s sales, and guides 40% growth for years. This is a Great business at the start of a generational tailwind, run by honest builder-owners. Compounders like this are rare and the market pays up for them for good reason.
The bear case, at its strongest (the honest red flag). You are paying 140 times earnings and 40 times book for a 7-year-old listed company whose fattest revenue slice depends on a government’s spending mood and procurement preference — a moat a localising Lenovo is already attacking. Margins just slipped, memory costs are surging, the AI capex cycle has credible bubble warnings, and the integrator-on-imported-chips model has a real precedent (Inspur) in how it can go wrong. Promoters sold near the highs. At this price, any stumble — a soft quarter, a margin miss, a cooling of the AI frenzy — could re-rate the stock hard, and the business wouldn’t even have to break for that to happen.
What the numbers actually support: a genuinely excellent, value-creating business (the 17.5/23 and +₹150 cr economic profit are not flukes), trading at a price that has already paid itself forward for several years of perfection. The quality is durable; the price is a current mood. The three things to watch that would tip it: (1) EBITDA margin holding ≥13% through FY27; (2) any second promoter sale or a global OEM winning an Indian sovereign tender; (3) operating cash flow staying real (≥70% of profit). No buy/sell/hold — the reader decides where the price meets the patience.
Sources
- Screener.in — Netweb Technologies (FY26 financials, ratios, shareholding). Fetched 2026-06-20.
- Netweb Q4 FY26 earnings call transcript, 2026-05-04 (BSE filing); Q3 FY26 call, 2026-01-19.
- Annual Reports FY24 & FY25 (MD&A, segment, governance sections).
- Business Standard — ₹1,734-cr sovereign AI order (2025-09-03); stock −7% on Q4 margins (2026-05-04).
- TrendForce / CNBC / IEEE Spectrum — DRAM/HBM shortage, Jan 2026.
- AICerts / Mordor — Lenovo India AI-server manufacturing under PLI, 2025.
- DataCenterDynamics / Washington Post — Inspur (IEIT) US Entity List precedent, 2023–25.
- Inc42 / Elets eGov / Entrepreneur — India AI Mission GPU tenders & scale, 2025.
- SEBI takeover disclosure (scanx/investywise) — promoter sale 70.64%→66.98%, 2026-02-12.
- Assumptions: Cost of Equity 12%; forward earnings growth 35% (management’s 35–40% organic guide, conservative end). Peer table from each peer’s screener snapshot, FY26.