Methodology, data caveats & who we are

Methodology

This report merges two Dart Consultants primers built independently on the same first-principles method: identify the one real structural asymmetry inside the product, derive the value ladder and moat mechanism from it, then apply that framework to the listed companies. Research was conducted via public sources only — company filings and investor presentations, credit-rating-agency press releases (CRISIL, ICRA, CARE, India Ratings), stock-exchange disclosures, company websites, Nvidia's own certification- programme documentation, industry trade press (Digitimes, Tom's Hardware, CBRE, JLL), Indian government sources (PIB, MeitY, ISM, IndiaAI Mission), and market-research summaries. The compute-layer research (original 8 companies) was conducted 18-19 September 2026; the seven GPU-cloud/telecom/conglomerate additions and all twenty physical-layer companies were researched 19-21 September 2026. Company financials and stock data are dated individually throughout each report, primarily to 18-20 September 2026 (screener.in).

Data caveats

• Constructed/indicative figures are labelled as such wherever used — step counts in §4, the non-margin portions of the value-ladder chart in §5, and forward-EPS growth assumptions in each company's valuation exhibit. These illustrate a structural relationship or a stated assumption; they are not measured from a disclosed company forecast unless explicitly cited as management guidance.
• Reported (not indicative) figures used without a qualifier include hyperscaler capex guidance, Nvidia/Micron/SK Hynix reported margins, the white-box ODM assembly-margin compression data, and each company's own disclosed revenue/margin/order-book figures — drawn directly from disclosures and trade-press reporting.
• Derived figures are marked with a tilde (~) and explained in the source line. Undisclosed figures are shown as a dash, never estimated. Several companies (Kaynes, LTTS, Cyient, Sasken, Syrma, Blue Star, Voltas, Thermax, GE Vernova T&D India, L&T) do not disclose a standalone AI/semiconductor or data-centre revenue percentage — this report does not manufacture one on their behalf.
• Market-size and forecast disagreements are shown as ranges or multiple bars rather than resolved to a single number — see §6's AI-market and power-demand dispersion charts.
• Several company-level financial figures showed material, sometimes irreconcilable, disagreement between sources. Handled explicitly rather than silently reconciled: Sterlite Technologies' Q1 FY27 PAT is reported three ways (₹125cr screener.in, ₹197cr company press release, a ₹17cr loss per a third aggregator) — its valuation is constructed specifically to avoid taking a side; Siemens Ltd's reported quarterly PAT (₹2,122-2,143cr) appears to include a ~₹1,800cr unexplained exceptional item against an "ordinary activities" run-rate closer to ₹343cr/quarter — its valuation departs from this report's usual two methods for exactly this reason; Netweb, Kaynes, Dixon and Cyient each showed material market-cap/P/E disagreement between aggregators, resolved in favour of the screener.in figure, dated, in each case; Voltas's revenue is reported differently by Trendlyne (₹10,837cr) and screener.in (₹14,244cr), likely a standalone-vs-consolidated difference this report states rather than resolves.
• ABB India reports on a calendar fiscal year (January-December), not the April-March year used by every other company in this report — its own company report labels periods "CY2025," "Q2 CY2026," etc. throughout to avoid a dating error against the other thirty-four companies.
• Corporate-structure disambiguations stated explicitly in the relevant company reports and summarised in §9: Schneider Electric Infrastructure Ltd (listed, switchgear/transformers) versus Schneider Electric India Pvt Ltd (unlisted, holds the APC/UPS brand); Siemens Ltd versus Siemens Energy India (separate companies post-demerger, June 2025); Kalpataru Projects International versus the unlisted "Kalpataru Limited" real-estate entity under the same promoter family; Siemens Energy India versus the unrelated, similarly-named "SEIL Energy India Limited"; and Tata Communications' divested data-centre real estate versus its separate, live NVIDIA AI Cloud partnership.
• Two claims required explicit correction or qualification before use: the "India allowed to import ~50,000 H100-class GPUs through 2027" figure is single-source aggregation, not a confirmed primary regulatory document (§8); and Adani Group's reported "$100bn AI data-centre" figure is the weakest-sourced hyperscaler-India-capex data point gathered and is not relied upon anywhere in this report's valuation or thesis.

About Us

Dart Consultants is a market intelligence and technology service provider. We are not a SEBI-registered Investment Adviser or Research Analyst. This document, and the company reports that follow it, are educational material only — not investment advice, and not a recommendation to buy or sell any stock. BUY/HOLD/SELL labels used in this series are an educational device for summarising publicly available information, not a regulated recommendation. The analyst(s) preparing this report hold no position in, and have no banking, advisory or brokerage relationship with, any company named in it, and have received no compensation from any of them. Readers should note that seven companies in this report carry above-average, recently-materialised risk factors independent of this report's eventual rating: Kaynes Technology (SEBI settlement, RPT controversy, guidance miss), CG Power (pre-2020 fraud history, now under different ownership), Cyient (a pending, unclosed acquisition), ESDS Software Solution (a ~2-week-old listing under exchange surveillance), Anant Raj (an unresolved Enforcement Directorate probe), Adani Enterprises (a $275 million US OFAC sanctions settlement and unresolved PIL litigation), and Sterlite Technologies (a genuinely unreconciled, three-way conflict in recent quarterly profit figures). Readers should treat these names as carrying above-average risk independent of this report's rating.