L&T Technology Services LTTS

Summary

LTTS is the engineering R&D services arm of the Larsen & Toubro group, reporting through three segments — Mobility, Sustainability and Tech — with semiconductor design, embedded software and, increasingly, AI-data-centre engineering all sitting inside the Tech segment (32-33% of FY26 revenue by one measure). The company entered chip design specifically through its 2018 acquisition of Graphene Semiconductor Services, which brought VLSI design and embedded-software expertise and a claimed roster of "10 of the world's top semiconductor fabrication and fabless companies" as customers — never named publicly. LTTS markets a dedicated "AI data stack" practice spanning power, cooling, compute and software engineering for data-centre clients, and opened a new AI- and defence-focused engineering design centre in Plano, Texas in mid-2025.

Financially, LTTS carries the strongest credit profile of any company in this report — CRISIL reaffirmed AAA/Stable/A1+ in June 2026, supported by a debt-free balance sheet (aside from lease liabilities) and ₹3,531 crore of cash and liquid surplus. Large-deal order intake has been consistently strong, with quarterly total-contract-value wins running $180-300 million across the past three quarters disclosed in this research. Against that, operating margin has compressed steadily from 22% (FY22) to 18% (FY26), and FY26 net profit growth was essentially flat versus FY25.

The core limitation for this report's purposes is disclosure: LTTS explicitly describes its semiconductor, data-centre-engineering and e-commerce-platform businesses as "newly incubated" and "yet to meaningfully contribute to overall revenue," with no standalone percentage disclosed for any of them. This report cannot verify how large the AI-compute-stack-specific opportunity actually is inside LTTS today — only that management describes it as still small relative to the whole.

Net: a HOLD. LTTS is the most conservatively financed, most creditworthy company in this report, trading at a materially lower multiple than its AI-server or OSAT-linked peers — but the margin-compression trend and the total absence of a disclosed AI/semiconductor revenue figure mean there is no verifiable basis for a more constructive call today.

• The strongest balance sheet and credit rating in this report. CRISIL AAA/Stable/A1+, nil utilisation of fund-based bank limits, and ₹3,531 crore of cash and liquid surplus as of March 2026 — categorically the safest name among this report's eight companies.
• The cheapest valuation multiple among the AI-compute-stack peer set covered here — ~26x trailing earnings versus 42-110x for several peers, despite a genuine (if undisclosed-in-scale) semiconductor and AI-data-centre engineering practice.
• Consistent, large-scale deal-win momentum. Quarterly large-deal TCV of $180-300 million across the three most recently disclosed quarters, including individual deals as large as $100 million.
• A genuine, decade-old semiconductor design practice, not a recent pivot — the Graphene acquisition dates to 2018, giving LTTS a longer track record in chip-design services than most peers attempting to enter this space now.
• A dedicated AI-data-centre engineering practice spanning power, cooling, compute and software — directly relevant to this report's own value-chain framing of the "hard half" engineering problem in §2-3.
• L&T group parentage provides financial and reputational backing that most peers in this report, as standalone entities, do not have.

• No disclosed AI/semiconductor/data-centre revenue figure exists — management explicitly describes these as sub-scale, "newly incubated" businesses, which means this report cannot verify how large or fast-growing the AI-compute-stack-specific opportunity inside LTTS actually is.
• Operating margin has compressed for four straight years, from 22% (FY22) to 18% (FY26) — a trend this report's research could not attribute to any single, disclosed cause.
• FY26 profit growth was essentially flat versus FY25 (₹1,281cr vs. ₹1,264cr, +1.3%), despite double-digit revenue growth and strong deal-win momentum — suggesting cost pressure is offsetting top-line gains.
• No named semiconductor or hyperscaler client was found for the Tech segment's marquee work — only the unnamed "10 of the world's top" claim.
• The current statutory auditor is unclear — conflicting names (Khimji Kunverji and Co LLP vs. MSKA & Associates) surfaced in this research and were not resolved.
• Named brokerage coverage is genuinely split — from a "Sell"-skewed 41-analyst compilation to a Buy-rated ₹4,850 target (Choice Institutional Equities), a wider disagreement than this report found for several other companies.

• Any future disclosure of a standalone AI/semiconductor/data-centre revenue percentage — the single most valuable disclosure this company could make for this report's thesis.
• Whether the four-year margin-compression trend stabilises in FY27.
• Confirmation of the current statutory auditor.

• Margin risk (dominant). Four consecutive years of operating-margin decline is a trend, not noise, and this report could not identify its root cause from public disclosures.
• Disclosure risk specific to the AI/semiconductor segment, which limits how confidently this report — or any external analyst — can size the opportunity.
• Client-concentration risk implied by the unnamed "10 of the world's top" semiconductor-client claim.
• Valuation-multiple risk is comparatively low given the current cheap multiple, but a further margin miss could still compress it further.

FY26 PAT of ₹1,281cr on an implied share count of ~10.54cr (market cap ÷ CMP) gives FY26 EPS of ~₹121.5. Applying an indicative 10% forward EPS growth assumption (reflecting the recent flat profit trend, not management guidance) gives a constructed FY27E EPS of ~₹133.7. We apply a target multiple band close to the current trailing multiple, reflecting a stable, well-financed but not currently accelerating business:

Base case rounded to ₹3,476. Named brokerage targets found in this research span a wide ₹2,900-4,850 range, itself a sign of genuine analyst disagreement; our base case sits within that range but does not adopt either extreme.

Source: screener.in (18 Sep 2026).