Long-Term Thesis

Long-Term Thesis — Harsha Engineers

1. Long-Term Thesis in One Page

The 5-to-10-year thesis is that Harsha becomes the world's default outsourced cage supplier to a still-consolidating top-6 bearing-OEM customer set, lifting consolidated ROCE from 14% toward 18% as bushings + large-size cages + Advantek + China brownfield drive the mix shift, Romania is finally right-sized, and working capital compresses below 130 days. It is not a wide-moat compounder. It is a narrow-moat sticky sub-supplier whose qualification gate has now survived one documented stress test (FY24-FY25 European downturn, zero platform losses), with growth optionality concentrated in adjacent product lines rather than the cage franchise itself. The case fails if any of three things break — a top-6 OEM reverses outsourcing on a major platform, the Romania balance-sheet object takes a third impairment, or working-capital intensity proves structural and ROCE stays anchored at 14% through the next capex cycle. The most credible long-run risk is none of those — it is that the company remains a perfectly fine engineering business that earns its cost of capital and does not re-rate, because 23.9× P/E already prices in a convergence it has not yet earned.

2. The 5-to-10-Year Underwriting Map

The thesis is not one bet. It is six independent drivers, each with its own confidence level. The map below names them in order of how much value each one carries for a multi-year compounder thesis — and what evidence today suggests about whether they will hold for a full cycle.

The driver that matters most over a decade is row 1 — bearing-OEM cage outsourcing continuing. Everything else in this map (mix shift, ROCE expansion, capital-allocation discipline, even Romania resolution) is downstream of whether the top-6 keep externalising cage production. The outsourcing trend is the supply of platform wins that compound through the qualification gate; reverse it and Harsha is a niche specialist with no growth pull. The signal is external — top-6 OEM annual reports, capex announcements, investor-day commentary on captive cage capability — not internal Harsha disclosure. If the trend holds, drivers 2-6 are mostly execution. If it reverses, drivers 2-6 cannot save the thesis.

3. Compounding Path

The 5-to-10-year compound is not an EPS curve. It is a return-on-capital curve — because the central debate is whether ROCE drifts from 14% toward 18% through mix shift and Romania normalisation, or stays anchored at 14% by working-capital intensity and the no-aftermarket structural gap. The arithmetic that drives the thesis is below: a base case of mid-teens revenue compounding, 100-200 bps of EBITDA margin expansion off the FY26 base, and operating-leverage on a ₹615 Cr net block that management says can support ₹2,700-3,000 Cr of revenue at full utilisation.

The seven-year history is the underwriting baseline. Operating margin compressed from a 14.0% FY21 peak to a 12.3-12.4% FY24-FY25 trough as Europe rolled over, then re-accelerated to 14.8% in FY26 — the new high. ROCE moved with it, peaking at 18% in FY22 (pre-IPO equity base) and bottoming at 12% in FY24-FY25 as IPO cash sat idle and Romania bled. FY26's 14.3% is mid-cycle. The 5-to-10-year thesis says it drifts toward 18% over a full cycle; the bear case says working-capital intensity caps it permanently.

The base-case arithmetic is 15% revenue CAGR (slightly above management's "mid-teen" guide for India Engineering and matching FY26 actuals) and 230 bps of cumulative EBITDA-margin expansion off FY26 (within management's "100-200 bps over 2-3 years" guide and trending higher as Romania resolves). EPS roughly 2.5x over five years — compounding that supports the current multiple without requiring a re-rate. A bull outcome depends on ROCE clearing 18% within the window (Schaeffler India / NRB Bearings territory) and multiple expansion from 24× toward 28-30×, roughly doubling the return profile. A bear path leaves ROCE at 14-15% (working-capital intensity holds, mix shift slower than guided), EPS still compounds at 15% on volume, but the multiple compresses toward 18-20× (NRB ex-growth-premium level) — the same EPS path produces a meaningfully lower IRR.

Cash and balance-sheet capacity for the path. FY26 net cash position (–₹101 Cr net debt, ₹473 Cr treasury, ₹372 Cr borrowings) funds the FY27 ₹130-140 Cr capex guide without external equity. Management has stated explicitly that the ₹615 Cr net block can support ₹2,700-3,000 Cr of revenue at peak utilisation — i.e., the next 75% of revenue growth requires only maintenance capex (₹30-40 Cr/yr) plus the China brownfield ($9.94M, ~₹85 Cr total). That is the balance-sheet runway that lets cash conversion normalise above 60% of NI from FY28 onward — IF the working-capital cycle does not structurally widen.

4. Durability and Moat Tests

A 5-to-10-year thesis lives or dies on whether the moat is durable in directions other than the one it has been tested in. The FY24-FY25 European downturn validated one specific stress case — cyclical demand shock, no platform losses. The tests below are the ones that have NOT been stressed yet, and where the thesis would either confirm or refute.

Tests 1 and 2 are the competitive stress cases that have not been tested in current configuration. Tests 3 and 4 are the financial durability cases — cash conversion is the most under-appreciated, Romania is the most disclosed. Test 5 is the long-tail technology risk that matters only over a full decade. The honest reading is that the moat has been documented against one demand-side stress (FY24-FY25 cyclical) but has not yet been tested against a competitive substitution stress (Chinese privates retake) or a financial intensity stress (CCC stays elevated through a full capex cycle). Investors who underwrite a wide moat are implicitly betting on all five — the file supports the bet at "plausible" confidence, not better.

5. Management and Capital Allocation Over a Cycle

The next decade is the next-generation team's decade. Vishal Rangwala (CEO, 48, joined 2007, 15+ years operating) and Pilak Shah (COO, 44, joined 2006) have been in operating roles for the entire post-IPO period and were formally re-appointed in December 2024 for five-year terms. Their fathers — Rajendra Shah and Harish Rangwala, both 77, both Lukhdhirji Engineering College mechanical engineers, both with 52 years in the business — remain on the board but have stepped back from day-to-day execution. The transition is essentially complete on operations; strategy and capital allocation still flow through the founders.

The capital-allocation track record is mixed, and the mix matters more for a 5-to-10-year underwriter than the absolute number. India-product-line bets have delivered: Bhayla greenfield (Harsha Advantek) commissioned within a quarter of guidance; China subsidiary turned PAT-positive; bushings compounded ~3.2x in two years; the IPO proceeds went into debt repayment exactly as promised in the RHP. International-subsidiary bets have not: Romania has been impaired twice in five years (₹95 Cr standalone FY25 + ₹27.68 Cr goodwill); HASPL Americas was wound up; the associate Sunstream solar stake was sold; the Solar EPC arm had to write off ₹20.6 Cr of "sticky" debt in FY25. The story tab calls this honestly: "trust India-product-line numbers, discount the subsidiary timelines by 12-18 months, and ignore aggregate ROE / margin glidepaths until a track record is rebuilt." That is the right frame for an underwriter looking out a decade.

What an investor is buying on capital allocation is founder skin-in-the-game — promoter group at the 75% SEBI cap, founder families own ~50% directly, zero pledges, zero dilution since IPO, promoter buying 0.39 percentage points to the cap in 2025. There is no realistic mechanism for the board to override the founders (5 of 10 directors are immediate family), but there is also no realistic mechanism for the founders to dilute or extract — alignment is structural. The two unforced governance errors on the file — the AIA Engineering Audit-Committee cross-directorship (₹15.11 Cr FY25 RPT sales approved by an Audit Committee chaired by an AIA executive) and the FY25 Rule 11(g) audit-trail deviation — are not fraud markers, but they are exactly the kind of footnotes that would be the first paragraphs of a short-report cover the day someone bothers to write one. The pre-IPO SEBI RPT investigation acknowledged in Whalesbook's May 2026 coverage is an open compliance overhang of unknown size and timing.

The pattern is honest: India bets work, international bets fail, governance hygiene needs work, alignment is structural. For a 5-to-10-year underwriter the single binary that matters is whether the next M&A or international bet repeats the Romania mistake — the founders have now demonstrated that they will impair before they pretend, which is rarer than it should be in Indian mid-caps, but the file does not yet show that they have learned to stop making the bet in the first place. The China brownfield is the first test of that lesson: it is local-for-local (defensible), customer-pre-qualified (de-risked), and small enough not to mortgage the franchise ($9.94M vs the 2016 Romania acquisition size).

6. Failure Modes

The thesis breakers below are not generic execution risks. Each is a specific, observable, multi-year failure mode tied to a number on the file. A 5-to-10-year underwriter should be able to recite all of them in a sentence.

7. What To Watch Over Years, Not Just Quarters

A 5-to-10-year underwriter does not watch the next print. The signals below are the multi-year markers that would update the thesis — each with a metric, a horizon, and what would validate or weaken the long-term case.