Tejas Networks — A Tata bet that ate one giant order and is now learning to eat smaller ones

The Pulse

Tejas Networks is India’s one serious home-grown telecom-equipment maker, owned by the Tatas, and it has just lived through the most violent boom-and-bust you’ll see on the exchange. A single mammoth order — building BSNL’s 100,000-site 4G/5G network — swelled its revenue almost ninefold to ₹8,923 crore in FY25 and produced a record ₹447 crore profit. Then the order ran out. FY26 revenue collapsed to ₹1,103 crore and the company posted a record loss of ₹909 crore across four straight loss-making quarters. The stock, which had touched roughly ₹1,495, fell to around ₹300 before settling near ₹576. The whole company now turns on three questions: whether the long-promised BSNL follow-on order finally arrives, whether ₹3,300 crore of stuck receivables actually turns into cash, and whether a genuinely new business — selling 5G radios internationally through partners like NEC — can grow fast enough to replace a customer it can no longer lean on. Management is contrite, still spending hard on R&D, and promising a turnaround “starting with FY27.”

The Business

Strip away the jargon and Tejas does one thing: it designs and builds the boxes that carry data across networks. Two families of product. Wireline — the optical and routing gear that moves traffic through fibre, from DWDM systems (think of cramming dozens of separate light-beams down one strand of glass — Tejas now pushes 1.2 terabits per second per channel) to the IP/MPLS routers that direct that traffic. Wireless — the radio and baseband equipment that turns a cell tower into a working 4G or 5G site. It sells to telcos, governments, utilities, railways and defence across 75-odd countries, though in practice India is 85–97% of revenue depending on the year.

What makes it genuinely unusual is that it is Indian and full-stack. Most countries buy their telecom gear from Ericsson, Nokia, or — until they were banned in many markets — Huawei. India had no real alternative until Tejas. The company likes to say India is now “one of very few countries with its own end-to-end 4G/5G stack,” and that is largely true because Tejas built it. That indigenous status is the moat: it brings government tailwinds (₹469 crore of production-linked-incentive money booked for FY25 alone), the “Atmanirbhar” sovereign-equipment positioning, and the Tata brand as a calling card for Tier-1 telcos abroad. It also brings a deep patent base — 676 filed, many of them standards-essential for 5G-Advanced and 6G — which matters less as a licensing cash-cow than as a defensive chip to trade in the cross-licensing poker that the telecom-standards game runs on.

The trouble is the shape of the business. This is not a steady-compounder churning out predictable boxes. It is lumpy, project-driven, and brutally exposed to operating leverage. When a giant order is flowing, margins look healthy (operating margins were 18–22% in its pre-boom heyday, ~14–19% at the FY25 peak). When the order stops, the fixed costs — heavy R&D, ~₹100 crore a quarter of depreciation, interest on a now-large debt pile — don’t stop with it, and margins invert violently: operating margin hit minus 112% in one FY26 quarter. The BSNL project was a one-customer, one-country event so extreme that in FY25 a single customer accounted for ₹7,329 crore of revenue. Living off one giant is glorious on the way up and gutting on the way down.

How Management Thinks

The team running Tejas — led by Arnob Roy, elevated from COO to MD & CEO in April 2026 after the CEO seat sat empty for nearly a year — is a deep-tech management, and it behaves like one. Its governing belief, repeated almost as a creed, is that you must keep investing through a downturn: “being in a situation where you don’t invest in technology evolution is a bigger risk for the future of your business than the other way around.” So through a year of record losses they did not cut R&D; they spent ~₹700 crore building intangibles (largely a ₹550 crore technology licence from NEC), funded mostly with debt, while the business generated only ~₹135 crore of cash. That is conviction. Whether it is discipline or stubbornness is exactly the question a patient owner has to sit with — and to their credit, management has now conceded the balance sheet is stretched and promised “far tighter” investment from FY27.

On candour, there are two faces. On the facts of the pain, management has become genuinely straight. By the April 2026 call Roy was apologising — agreeing openly with a shareholder who pointed out the stock had fallen from ₹1,495 to ₹300, saying flatly “it’s been disappointing for us also,” no spin. That is rare and worth respecting. But on numbers that bind them — breakeven revenue, deal sizes, margins, a timeline to profit, market-share targets — they refuse, citing a long-standing no-guidance policy (one analyst disputed it even exists). The result is an awkward straddle: they will promise a turnaround “starting FY27” and call PAT-positivity in FY28 “the goal” in the same breath as “we don’t give guidance.” Analysts have pressed hard on this, on the multi-year unprofitability, and on a working-capital cycle so long — debtor days of 1,077, inventory days of 1,160, a cash-conversion cycle of roughly 2,000 days — that one investor asked whether shareholders are, by design, never going to see a return. Management’s answers there were thinner.

The credibility scorecard is genuinely mixed. They delivered the BSNL build — 100,000 sites in about 18 months, one of the largest single-vendor RAN deployments anywhere — which is a real, hard thing they said they’d do and did. But the promise that has slipped repeatedly is the BSNL ~18,000-site add-on order: across all four calls it was “expected within FY26,” and as of April 2026 it still had no purchase order, only an advance intimation sitting with TCS. The delay isn’t their fault (it’s BSNL’s site-readiness), but the prediction was wrong four quarters running, and that is the kind of thing a careful reader files away. The other tell is the smart-money exit: foreign institutions roughly halved their stake from 10.9% to 5.3% over three years, right through the boom, while retail piled in and the shareholder count more than doubled to 3.8 lakh. The professionals left as the crowd arrived.

Where It’s Going

The strategic pivot is the real story now, and it is sensible. Having ridden one government giant up and down, Tejas is deliberately rebuilding around things that aren’t BSNL — and notably, its order book of ₹1,514 crore (up ~49% year-on-year) contains zero BSNL content. Three legs hold up the next chapter. BharatNet Phase III — the rural-broadband build — where Tejas won 7 of 12 packages and is deploying 60,000+ routers; this is the working engine today, in wireline, where margins are better. International wireless — the genuinely new beachhead: a first order from Japan’s NEC to build 5G massive-MIMO radios for a global customer (revenue lands in FY27), a 4G expansion order in South Asia, Open-RAN trials with Rakuten, and pilots across the Americas. Management is honest that none of this will match BSNL’s 100,000-site scale — international markets are simply smaller — but it diversifies the customer base and, crucially, carries higher margins than the cut-throat Indian market, which lowers the revenue Tejas needs to break even. The AI-era optical bet — selling the high-bandwidth gear that connects AI data centres, where the company already won its first sovereign data-centre networking deal and a nationwide multi-terabit build for a hyperscaler.

The tensions are equally clear, and a patient owner should hold them in plain sight. The balance sheet is the pressing one: ₹3,258 crore of receivables (about three times annual revenue), ₹2,438 crore of inventory pre-bought for a BSNL order that hasn’t been placed, ₹3,531 crore of net debt, and only ₹505 crore of cash. Management insists the inventory is redeployable (the radios work for international 4G too) and that BSNL collections will land in FY27 — but until receivables convert and the order book turns into revenue, this is a company burning roughly ₹150 crore of EBITDA a quarter and funding its ambitions with borrowing. The bet management is asking the market to make is that FY26 was a trough between two waves — that the new, more diversified order book converts in FY27, margins lift on international mix, and the business climbs back toward profit without another single-customer rescue. They have built the products and won early references to make that case. They have not yet shown they can turn this order book into cash on a predictable cadence. FY27 is when the words get tested against the numbers.

The Four Checks

1. Quality and moat. A strategically important business, but not a structurally protected one. The real assets are indigenous status — India’s only end-to-end 4G/5G equipment stack, with the policy tailwinds that brings (₹469 crore of PLI money in FY25 alone, BharatNet set-asides) — plus Tata parentage and a 676-patent base built for cross-licensing defence, not licensing income. None of this confers pricing power: the Indian market is cut-throat, the global incumbents (Ericsson, Nokia) are vastly larger, and the revenue line depends on lumpy government orders Tejas does not control. EPS has been negative in roughly half of the last twelve fiscal years, which is what a contestable, order-cycle business looks like over time. The edge is real but it is a door-opener, not a moat — and it makes the remaining checks heavy going.

2. Returns on incremental capital and runway. Poor, and not just at the trough. ROCE peaked at 15% in the FY25 super-cycle year and sits at −14.6% in the June 2026 snapshot; three-year average ROE is −4%. The deeper problem is that even the “profitable” boom consumed capital — operating cash flow was −₹2,036 crore in FY24 and −₹491 crore in FY25, because every rupee of BSNL revenue had to be financed through a working-capital cycle now running roughly 2,000 days (1,077 debtor days, 1,160 inventory days). A business whose best year in a decade earned 15% on capital while bleeding cash, and whose normal years earn nothing, has no demonstrated ability to compound retained capital. The runway story (international 5G, AI-era optics, BharatNet) is plausible but entirely unproven at the returns level.

3. Capital allocation for the stage. Conviction-driven and, on the record so far, costly. Management took a near-debt-free balance sheet to ₹4,177 crore of borrowings in three years — partly to finance the BSNL working-capital ramp, partly to fund ~₹700 crore of intangible investment (including a ₹550 crore NEC technology licence) in a year the business generated only about ₹135 crore of cash. They pre-bought ₹2,438 crore of inventory for a BSNL add-on order that still has no purchase order, paid a token dividend in the peak year (10% payout, FY25), and have never bought back stock. The deep-tech logic of investing through the trough is coherent, and they did deliver the 100,000-site build they promised — but debt-funded R&D atop stuck receivables is aggressive allocation for a loss-making stage, and management itself has conceded the point, promising “far tighter” discipline from FY27. Judged on actions, this is a 3-out-of-10 record with a stated intention to improve.

4. Price. Demanding to the point of speculative. As of the June 2026 snapshot the stock trades at ₹557 — a ₹9,902 crore market cap, no P/E because there are no earnings, 3.31 times a book value the business is currently eroding at a −26.8% ROE, with ₹3,531 crore of net debt on top. The market is paying over ten times FY26 revenue for Tata parentage and the FY27 turnaround promise: BSNL collections landing, the ₹1,514 crore order book converting, international margins lifting. Every leg of that has to come through for the price to be justified by economics rather than hope, and the professionals voting with their feet (FIIs halved to 5.3% as retail doubled to 3.8 lakh shareholders) suggest the people who do this for a living are not waiting to find out. This is a demanding price for an unproven recovery.

Sources

• Earnings calls read: Q1 FY26 (14 Jul 2025), Q2 FY26 (17 Oct 2025), Q3 FY26 (9 Jan 2026), Q4 FY26 (15 Apr 2026). The Oct-2025 disclosure was a PPT-only quarter with no transcript and is not separately covered.
• Annual reports read: FY24, FY25, FY26 (trimmed high-signal sections).
• Financial snapshot: screener.in (consolidated), fetched 2026-06-08 (logged-out/public).
• Research dumps (not published): vault/Sources/Earnings/Tejas Networks Ltd/.
• Gaps to flag: the FY26 annual-report extract was thin — the Chairman’s and MD’s letters and most of the MD&A were stripped in trimming, so FY26 full-year figures lean on the concalls and snapshot rather than the AR. Headline FY25/FY24 numbers come from the chairman’s letters, not clean financial tables. Quarterly and balance-sheet figures are from the screener snapshot and the call transcripts.