Netweb Technologies — India’s home-grown AI-supercomputer maker at full sprint

The Short Version

Netweb designs and builds high-end computers in India — supercomputers, AI servers packed with graphics chips, and private-cloud systems — for governments, research bodies and large companies. It is, in its own words, India’s only full-stack domestic maker of these machines, and that positioning has put it squarely in the path of the country’s biggest technology spending wave: the national push to build sovereign AI computing power. The year just ended (FY26) was a landmark — revenue nearly doubled to ₹2,184 crore and profit rose ~81% to ₹206 crore — driven by AI systems, which exploded from ~15% of sales to 43% in a single year. It carries zero net debt, earns exceptional returns on capital (~37%), and walked into the new year with an order book already larger than everything it billed in all of last year. The caveats are the flip side of the same story: the growth leans heavily on lumpy, government-linked mega-orders that distort quarters and carry thinner margins, and it depends on the pace of India’s AI spending continuing.

What This Company Actually Does

Most computing you use runs on machines someone else built — but at the very high end (supercomputers that model weather or crash-test cars, AI servers that train large language models, the private clouds that big institutions run in-house) the hardware is specialised, expensive, and usually imported. Netweb, founded in 1999 and listed in 2023, designs and manufactures these machines in India. Its supercomputers have appeared 15 times in the global Top 500 ranking — a genuine engineering credential, not a marketing line.

The business stands on three pillars, each more than a decade old and together ~95% of revenue:

• HPC (high-performance computing / supercomputers) — the traditional core, sold mostly to government research bodies and increasingly to industry (oil & gas, automotive).
• Private Cloud / HCI — the in-house “data-centre in a box” systems large enterprises run instead of renting public cloud.
• AI systems — servers built around graphics chips (GPUs, mostly NVIDIA’s latest) for training and running AI models. This is the rocket.

The crucial thing to understand about how Netweb makes money is that it is “capability-driven, not capacity-driven.” It doesn’t run big factories churning out volume; it designs the full machine — the compute, the interconnections, the storage, and a software layer (its Skylus.ai platform) that orchestrates the GPUs — and assembles to order. It gets early access to chip-makers’ roadmaps (~18 months ahead, on NVIDIA’s newest Blackwell and upcoming Vera Rubin), sells a complete system as a one-time capital purchase (no ongoing cloud rental), and keeps its margins above rivals by adding its own design and software rather than just “pushing boxes.” It runs zero net debt, spends ~3% of revenue on R&D, and earns strikingly high returns (FY26 return on capital ~37%).

A note on ownership worth getting right: the founding Lodha family’s stake has fallen from ~75% to ~67% since listing — but the latest step-down (a ~4% sale in February 2026) was a deliberate move to improve the stock’s thin trading liquidity, accompanied by a public commitment to no further selling for 12 months. The remaining drift reflects a young listing’s float gradually opening up to a tripling retail shareholder base.

The Long Game

Netweb’s long game is simple to state: it is the domestic supplier of choice for India’s build-out of sovereign AI computing — the national project to own AI infrastructure rather than rent it from American hyperscalers. Management’s framing is that India is “a sovereign-driven market, not a hyperscaler-driven market,” and Netweb is “India’s only full-stack domestic provider for high-end computing systems.” That’s the whole thesis.

The demand behind it is concrete. The government’s IndiaAI Mission has scaled its GPU ambition from 10,000 to 25,000 to talk of up to 100,000 chips, with management noting “budget is not the constraint — the speed of spending is.” It flows through two channels: renting GPU capacity via cloud providers (which has started, and is where Netweb’s big orders sit), and government building its own on-premises infrastructure (slower, just beginning). Alongside sits the National Supercomputing Mission 2.0 feeding the HPC pillar, and India’s large IT-services firms (TCS, Infosys, Wipro) building their own AI stacks as customers.

Management guides to 35–40% organic revenue growth for the next couple of years at a 13–14% operating margin — and crucially, that guidance excludes the large “strategic” AI-mission orders, which are upside on top. AI is expected to settle around 35% of the mix, with HPC and private cloud splitting the rest. The demand horizon management keeps citing is “18–24 months” of unabated AI appetite. The reason to take the guidance seriously: Netweb has reported every quarter since its 2023 IPO without missing its 35–40% growth promise.

The Story So Far

The spine of FY26 is the arrival of two enormous “strategic” orders — national AI-infrastructure contracts worth ~₹2,184 crore combined — landing on top of a steadily compounding base business. (The quarterly figures below are management’s own reported numbers from each call; the screener snapshot, re-fetched June 2026, matches them line for line.)

Q1 FY26 (reported August) — a strong, ordinary quarter

Revenue ~₹301 crore, up 102% on a soft year-ago base, with AI systems already 29% of revenue (up 300%). Management held its 35–40% growth and ~14% margin guidance, raised its expected AI share to ~22% of the year, and was explicit that the IndiaAI Mission was not yet in its pipeline — a notable line given what came next. The CMD reframed how to read the company: don’t watch the order book (it converts in 8–12 weeks), watch the ~₹4,000 crore pipeline.

“I have been categorically saying that IndiaAI mission is not included in my pipeline.” — Sanjay Lodha, CMD (Q1)

Q2 FY26 (reported November) — the mega-orders land

The base business grew a steady 21%, but the headline was two strategic orders worth ~₹2,100 crore, of national significance, to be executed by FY27 — Netweb’s entry into the IndiaAI compute build-out (via empanelled cloud providers, not directly from government). Management was careful to manage expectations: the orders carry margins ~150–200 basis points lower at the profit level (large scale, thinner cut), are fully backed by letters of credit and customer advances (“credit risk almost zero”), and need no new capex. Organic guidance stayed unchanged — the strategic orders were framed as pure upside.

“There is no revision in the guidance. We are maintaining 35% to 40% growth for the regular business and strategic orders are being informed separately.” — Ankit Singhal, CFO (Q2)

Q3 FY26 (reported January) — a record, and the lumpiness shows

One of the strategic orders (~₹450 crore) executed in the quarter — earlier than the guided Q4 — and revenue leapt to ₹805 crore (up 141%), with AI systems a remarkable 64% of the quarter. But the quarter also exposed the model’s character: with the mega-order in the mix, the operating margin dipped to 12.2% (as flagged), and the organic business looked optically soft quarter-on-quarter. Management’s repeated defence — that large orders make quarter-to-quarter comparison meaningless and that nine-month organic growth was still >30% — is the honest frame for reading Netweb.

“In our kind of business, it becomes very difficult to measure anything quarter-on-quarter.” — Sanjay Lodha, CMD (Q3)

Q4 and full-year FY26 (reported May) — the landmark closes

Revenue ₹774 crore for the quarter and ₹2,184 crore for the year, up 90%, with profit up 81% to ₹206 crore, return on capital at ~37%, zero net debt, and a ₹3/share dividend. AI systems ended at 43% of revenue (up 460%). The standout: Netweb entered FY27 with a ~₹2,100 crore firm order book (₹2,400 crore including near-certain wins) — more than its entire FY26 billing — on day one. Guidance held at 35–40% organic growth (strategic orders on top) and 13–14% margin, with the remaining strategic order to execute over the next three quarters.

“Day one of financial year ‘27, we already have on our books more than what we had billed for all of last year.” — Sanjay Lodha, CMD (Q4)

The through-line a long-term reader should hold: this is a genuinely differentiated, high-return, debt-free company executing flawlessly into a once-in-a-generation national tailwind — and a company whose growth now rides on a small number of very large, government-linked orders that make any single quarter noisy, dilute margins when they’re heavy, and will need to be continually replaced by fresh wins once executed.

Where Things Stand

Netweb enters FY27 in an unusually strong position: record FY26 behind it, an order book bigger than last year’s revenue already in hand, a demand wave management sees running 18–24 months, no debt, ~37% returns on capital, and a near-unique position as India’s full-stack domestic AI/HPC builder at the centre of the sovereign-compute push. The base business alone is guided to grow 35–40%, with the strategic AI-mission orders as upside — and a pipeline of ~₹4,000 crore beyond the order book.

The honest watch-items, for a patient owner, are the texture of that growth rather than its direction. The strategic mega-orders are lower-margin and lumpy — they’ll keep distorting quarterly optics and pulling the blended margin toward the bottom of the 13–14% range when they’re heavy — and they are, by nature, one-off project sales that must be won again and again to sustain the pace. The whole AI leg depends on the speed of India’s government and enterprise AI spending, which management itself says is the real variable (not budget). Global memory and component shortages are a live risk that management says it’s managing through supply-chain planning. And the price already knows most of this story: at the June 2026 snapshot the stock sat at ₹4,349 — roughly 120 times trailing earnings and 34 times book value, after nearly tripling from its 52-week low — so a richly-priced stock leaves little room if the AI capex cadence slows.

The Four Checks

1. Quality and moat. A genuinely good business with a real but narrow and partly borrowed moat. The concrete edges: it is, by its own repeated and so-far-uncontested claim, India’s only full-stack domestic designer-and-manufacturer of high-end computing systems; it gets chip-makers’ roadmaps ~18 months early (already building on NVIDIA Blackwell, set up for Vera Rubin); it has 15 Top-500 supercomputer appearances and a 25-year engineering record; and its own software layer (Skylus.ai) plus full-system design keeps it above “box pushers” on margin. But the moat leans heavily on policy — the sovereign-compute preference for a domestic builder — and the underlying market is contestable: the irreplaceable technology in every machine is NVIDIA’s, not Netweb’s, and global systems vendors are vastly larger. Call it niche dominance in a protected lane, durable as long as the lane stays protected and the execution stays clean.

2. Returns on incremental capital and runway. The engine is capital-light growth at very high returns. FY26 delivered ROCE of 37.5% and ROE of 32.8% (ROCE was running at 41.3% at nine months), on a gross fixed-asset turnover of 33x — and the screener record says this isn’t a one-year artefact: three-year average ROE is 29%, profit has compounded at ~90% over five years, and working-capital days (on screener’s measure) have collapsed from 62 to 14 — the company calls itself “capability-driven, not capacity-driven,” and management says no capex beyond a routine ₹20–25 crore is needed in FY27 despite revenue having doubled to ₹2,184 crore. The real capital the business consumes is working capital (cash conversion cycle 84 days, guided 90–110), funded from a debt-free balance sheet. The rate of the loop is excellent; the question is the runway: guidance is 35–40% organic growth “for the next couple of years,” a ~₹4,000 crore pipeline sits beyond the order book, and management’s own demand horizon is 18–24 months. High returns with a real but not yet proven-long runway.

3. Capital allocation for the stage. Close to textbook for a company at full sprint. Management is feeding the engine where it actually needs feeding — R&D (~3% of revenue, team grown from 240 employees at IPO to 600+), a new end-to-end manufacturing facility (May 2024) and a 15,000 sq ft liquid-cooling line — while keeping zero net debt and using only short-term, self-liquidating borrowings to fund the working capital of the mega-orders. A ₹3/share dividend (payout has run at 5–15% of profit since FY23) is a token, which is the right size when incremental returns run near 40%. No acquisitions, no empire-building, no equity dilution (the promoters’ February 2026 ~4% sale was a secondary sale to improve float, with a 12-month no-further-selling commitment — it took nothing from the company). No buyback history exists to judge; at three years listed, none should.

4. Price. As of the June 2026 snapshot the numbers are in, and they are steep: ₹4,349 a share, a market cap of ₹24,763 crore, 120 times trailing earnings, 33.8 times book value, a dividend yield of 0.06%, against a 52-week range of ₹1,700–4,965 — the stock has nearly tripled from its low inside a year. The register tells the same story: shareholder count has tripled in two years to ~2.43 lakh and public holding has climbed to 19%, the classic footprint of a hot theme finding retail. Set the multiple against the economics and the arithmetic is unforgiving: even if profit doubles again on the order book and the strategic pipeline, the buyer is still paying ~60 times next year’s earnings for a business whose mega-orders are lower-margin, lumpy, and must be re-won every cycle. The quality is real, the growth guidance (35–40% organic, strategic orders on top) is credible, and the company has never missed it — but at 120 times earnings the market has prepaid for years of flawless execution and an unbroken AI capex cadence. Priced for perfection; the room for anything less has been sold.

Sources

• Concall transcripts (4): Q1 FY26 (Aug 1, 2025), Q2 FY26 (Nov 3, 2025), Q3 FY26 (Jan 19, 2026), Q4 FY26 + full-year (May 4, 2026) — BSE filings, converted to markdown. These carried the full quarterly financials, order book and guidance, and are the backbone of this chronicle.
• Annual reports (2): FY24 (first as a listed company) and FY25 sections — FY25’s flagged the new end-to-end manufacturing facility (commissioned May 2024), the Skylus.ai launch and the NVIDIA Blackwell/AMD Genoa server push. Both extracts were thin (CMD letters were image-omitted).
• Screener snapshot: re-fetched 2026-06-10 (standalone page; the earlier consolidated fetch returned empty tables). The fresh snapshot carries the full quarterly/annual financials, ratios, valuation and shareholding — it confirms the management-reported figures used in this chronicle, and supplies the valuation multiples quoted in the Price check.
• Research files: vault/Sources/Earnings/Netweb Technologies India Ltd/ — raw transcripts, AR sections, snapshot, per-document digests (not published).