
Netweb Technologies · NSE NETWEB
AI server and high performance computing manufacturer, NVIDIA manufacturing partner · sourcing secondary, no filing cited
Netweb owns no megawatts. It builds the machines that fill other people's data centres, so it has no capacity ladder and no place on the Execution against Ambition plot, which is drawn in megawatts. The unit it is measured on is the order book, and the concentration that sits inside the other names' revenue sits inside its backlog instead.
Order book
2,507Rs cr
As at 2026-06-30.
IndiaAI Mission
1,734Rs cr
Supply of NVIDIA Blackwell powered servers. Awarded 2025-09, due H1 FY27.
Share of the book
69.2%one counterparty
Everything else, every other customer, is Rs 773 cr.
Trailing earnings
99x
Roughly 99 times trailing earnings. The brief's own verdict is genuinely executing, and priced for perfection.
Exhibit 1
69 per cent of the book is one counterparty, and it is backlog rather than revenue
Share of each company's own whole. Netweb, order book in Rs cr at the date stated. Sify, revenue from operations. The two wholes are different quantities and are labelled as such.
Order book, Rs 2,507 cr, as at 2026-06-30
Research note, no filing cited
Revenue from operations, latest filed period
Draft red herring prospectus, printed page 36
The order book is stated at 30 June 2026 and this order was awarded in September 2025 with delivery due in H1 FY27. Anything already delivered has left the book. The share is therefore a ceiling on the concentration at that date rather than a measurement of it, and it is drawn as one.
The two bars are close and they are not the same measurement. Sify's 67.0 per cent is revenue already earned. Netweb's 69.2 per cent is work not yet delivered, awarded 2025-09 and due H1 FY27. One is a record of who paid. The other is a statement about who is expected to.
Source: Blueprint research note, BharatAI Infra Scope, compiled from company filings, brokerage notes and press reporting. No Netweb filing is cited, so nothing here is traced to a primary source. Sify from its draft red herring prospectus, revenue by client.
Exhibit 2
AI systems are 64 per cent of the quarter and 48 per cent of the nine months holding it
Share of revenue, per cent. Each square is one per cent. The longer period contains the shorter one, so these are not two independent readings.
One quarter
Nine months
The quarter runs 16 points above the period that contains it. Netweb is described as an AI company on the strength of Q3 FY26, and the nine months that include it are still under half. Both readings are true. Only one of them is a year.
The earlier quarters are not backed out of the two figures here. Doing that needs an assumption that revenue lands evenly across quarters, which is exactly what a lumpy project business does not do, and the arithmetic would look more precise than the evidence is.
Source: Blueprint research note, BharatAI Infra Scope, compiled from company filings, brokerage notes and press reporting. No Netweb filing is cited, so nothing here is traced to a primary source.
Exhibit 3
The one name here with no filing cited is the only one that can price itself
Severity against likelihood, an analyst grading rather than the company's. A chip is filled where the magnitude is derived from the figures recorded for this name and outlined where the row is judgement. Valuation carries a row here and none on the Sify page: a multiple needs a price, and a draft prospectus carries no price band.
| Severity | Low | Medium | High |
|---|---|---|---|
| High | 0203 | 01 | |
| Medium | 0405 | ||
| Low |
Likelihood, across
Most of the order book is a single counterparty, and that counterparty is a government mission rather than a commercial customer.
High severity, high likelihood
69.17 per cent of the order book, one counterparty. IndiaAI Mission against the book at 2026-06-30.
The same shape as the client concentration on the Sify pages, in a different unit. There it is revenue that has been earned. Here it is work that has not.
Mitigant. A government award carries lower credit risk than a private buyer of the same size, and the order is already won rather than bid.
One order has a published delivery window and the rest of the book has none, so the revenue timing of most of the backlog is unknown.
High severity, medium likelihood
773 crore rupees of book with no published schedule. The order book less the single order whose delivery date is known.
The magnitude is the part of the book with no schedule, not a claim that it is late. Nothing here says when the remainder converts, because nothing read says so.
Mitigant. The named order is the largest single item and it does carry a window, so the biggest piece of the timing question is answered.
The multiple is paid today on earnings already reported, while the concentration sits in work not yet delivered. Those are different periods.
High severity, medium likelihood
99 times trailing earnings. Price paid today against earnings already reported.
A multiple is the only pillar this page can evidence without a filing, and that is a comment on the multiple as much as on the page.
Mitigant. The order book is real and dated, so the growth the multiple anticipates is at least written down somewhere rather than purely narrative.
The AI share that defines the story is one quarter, and the nine months containing that quarter are still under half.
Medium severity, high likelihood
16 points between the quarter and the nine months holding it. Q3 FY26 against 9M FY26.
The gap is drawn in points rather than as a ratio, because the longer period contains the shorter one and a ratio between them would imply they are independent readings.
Mitigant. Both readings are reported and both are true. The company is not choosing which one to publish; the market is choosing which one to quote.
The concentration share is a ceiling rather than a measurement, because the book is struck after the award and anything already delivered has left it.
Medium severity, high likelihood
No magnitude. This row is graded, and nothing on the page pretends otherwise.
Not measured, and marked as such. Measuring it needs the book split by customer and by delivery date, and the research note gives neither.
Mitigant. The direction of the error is known. The true share can only be lower than the figure drawn, never higher, so the exhibit cannot overstate by accident.
Pillar coverage
Revenue quality
3 risks
Cash conversion
no row, nothing read
Balance sheet
no row, nothing read
Governance
no row, nothing read
Business model
1 risk
Valuation
1 risk
Three of the six pillars carry no row. Cash conversion and the balance sheet need revenue, receivables and cash flow in rupees, none of which has been read, and receivables are exactly where a lumpy project business hides. Governance needs the annual report. Valuation is the one pillar this page can evidence without a filing, because a multiple needs only a price and a reported earnings figure, which is also why a multiple is the weakest kind of evidence on the list.
1 of 5 risks sit in the worst cell, and 4 of 5 carry a figure derived from the recorded numbers rather than a grade, and none rests on a printed page. Severity and likelihood are analyst gradings, not the company's. A magnitude marked measured is derived from the figures recorded for this name, and those figures come from a research note rather than from a filing. Derived is not the same as filed, and each row says which it is.
No row is graded low on either axis, so the bottom row and the left column of the matrix stand empty. That is a property of the register rather than of the company: these are the risks worth writing down, not a survey of every risk the filing lists.
Source: Blueprint research note, BharatAI Infra Scope, compiled from company filings, brokerage notes and press reporting. No Netweb filing is cited, so nothing here is traced to a primary source.
The price, and what sits behind it
Every figure above comes from a research note rather than from a Netweb filing, so there is no revenue in rupees, no margin, no cash flow and no receivables beside them. What is here is what the order book is made of, and what is being paid for it.
99 times trailing earnings
Roughly 99 times trailing earnings. The brief's own verdict is genuinely executing, and priced for perfection. The multiple is on earnings already reported. The concentration is in work not yet delivered. Those are different periods, and the price is paid today for the second one.
One delivery window
Rs 1,734 cr, or 69 per cent of the book, is due H1 FY27. The research note gives no schedule for the remaining Rs 773 cr, so no timeline is drawn for it here.
Sify is the name with the deepest disclosure here, and the difference in what can be said about the two is the difference between a headline and a filed statement.