Kernex Microsystems (KERNEX) — Primaegis Equity Desk

A1–A2 · Business model, capabilities and the shape of the moat

Kavach (TCAS) is India's indigenous SIL-4, ETCS-aligned Automatic Train Protection standard, developed by RDSO with three OEMs and jointly patented by RDSO, Kernex, Medha Servo Drives and HBL Engineering[CARE 04-Nov-2025]. Kernex supplies both sides of the system — wayside (RFID tags, trackside equipment rooms, radio towers, OFC, interlocking interface) and on-board (loco/trainset cab equipment that reads tags and applies brakes automatically). Every contract is the same bundle: supply, installation, testing and commissioning.

Is the moat real? It is a licence-to-bid with a regulator-owned barrier. The barrier bites — the v3.2→v4.0 migration cost Kernex a ₹2,041 Cr order outright. But it confers no pricing power (CARE: "tender-based L1 procurement limiting pricing flexibility"), and the IP is co-owned with the customer's own standards body, so the count of three OEMs is an administrative decision RDSO can widen, not a technology cliff. Within the three, Kernex is the smallest: HBL Engineering runs 7.7x the revenue at a higher ROCE.

The consortium economics problem

Kernex acts as lead member with joint-and-several liability, raises the entire invoice to the customer, banks it through escrow, and recognises revenue only to the extent of its own scope[FY25 Key Audit Matter]. Its economic share is disclosed on only 3 of ~8 consortium orders (WR 70%, SR 60%, SCR-VRRC JV 80%), leaving ₹771.70 Cr at an unknowable haircut — on contracts where it carries 100% of the performance liability.

Derived unit economics: loco sets price at a near-identical ~₹81.5 lakh incl-GST across three separate awards (₹81.66 / ₹81.54 / ₹81.42 lakh) — evidence that price is set by the tender structure, not the OEM. Ex-GST realisation ≈ ₹69.1 lakh/set.

A3–A4 · Opportunity, operations and the execution record

Kernex's own last-published TAM framing: High Density Network 11,000 km, Highly Used Network 24,230 km, entire IR network 85,000 km targeted 2030–32[IP 24-Mar-2025]. Rollout status and the 20,000-loco tender programme are SECONDARY web aggregation, unverified against any Indian Railways primary document — flagged rather than relied upon.

Why now: the certification gate opened on 14-Oct-2025 when RDSO approved Kavach v4.0, unblocking deliveries against ₹3,346 Cr of accumulated orders. Every large on-board order Kernex has won post-dates it. And the field proof is independently verifiable — 572 of 652 NCR v4.0 route km, 160 kmph demonstrated, ISA-certified by Italcertifier of Italy.

The recurring failure mode: certification slip

Guided (24-Mar-2025)	Delivered	Slip
TCAS 4.0 ISA by April 2025	RDSO v4.0 approval 14-Oct-2025	~6 months
CLW delivery from end-May 2025, install from Jun-2025	Not complete by 05-Dec-2025; extension refused 09-Jan-2026	Order lapsed
NCR-A (GZB-CNB) by June 2025	207 km commissioned 02-Jul-2026	~12 months

Capex was essentially nil. The 2.27x revenue ramp consumed just ₹7.57 Cr of investing cash; fixed assets moved ₹25 Cr → ₹34 Cr. The capital went into working capital instead — inventory ₹52.58 Cr → ₹371.32 Cr, receivables ₹23.65 Cr → ₹316.26 Cr. Operationally scalable, financially constrained, regulatorily gated.

Project tracker — 17 live projects, most with no disclosed progress

Project	Value	Awarded	Deadline	Progress
CLW on-board v4.0, 3,024 sets	₹2,465.71 Cr	15-Jan-2026	~15-Jan-2027	30% at 29-May-2026
CLW follow-on	₹475.21 Cr	27-May-2026	not stated	none
BLW on-board, 505 sets	₹411.17 Cr	11-Feb-2026	~Feb-2027	none
SER — KERNEX-MRT	₹325.33 Cr	27-Feb-2025	1,000 days	share undisclosed
SR — KERNEX-VRRC (60%)	₹311.03 Cr	13-Jun-2025	730/540 days	none
DFCCIL	₹209.82 Cr	29-Aug-2025	730 days	share undisclosed
WR — KERNEX-KEC (70%)	₹182.81 Cr	06-Jun-2025	730 days	none
WCR — KERNEX-KEC	₹151.41 Cr	19-Aug-2025	600 days	share undisclosed
ICF Chennai + warranty/CAMC	₹100.53 Cr	14-Jul-2026	31-Mar-2028	new, long runway
NCR wayside v4.0 (in KEC consortium)	—	2024-25	—	207 km commissioned, ISA-certified
Jindal Steel yard automation	₹15.90 Cr	28-May-2026	not stated	LOI, not a PO
KERNEX-BHEPL JV — moving block / ATO / CTC	—	07-Mar-2026	—	not commenced at 31-Mar-2026
Kernex TCAS JV (80%)	—	legacy	—	₹4.03 Cr provided; stalled
Avant-Garde Infosystems Inc, USA (100%)	—	legacy	—	impaired ₹3.61 Cr; UNAUDITED

Unreconciled: 30% of the CLW order ≈ ₹627 Cr ex-GST supplied by 29-May-2026, yet total FY26 revenue was ₹430.22 Cr including all wayside work. The filings do not reconcile supply against recognition — which is precisely why Q1FY27 is the cleanest test of the story.

A5 · Financials — a genuine step-change, and what carried it

Revenue went ₹4 Cr (FY23) → ₹20 Cr (FY24) → ₹190 Cr (FY25) → ₹430 Cr (FY26). The step-change is real; the base was nearly nothing.

₹ Cr	FY21	FY23	FY24	FY25	FY26
Revenue	18	4	20	190	430
EBITDA	4	−17	−23	41	149
EBITDA %	22.2	−425.0	−115.0	21.6	34.7
PAT	2	−20	−27	50	88
ROCE %	4	−23	−21	24	48
Borrowings	20	11	28	40	162
Free cash flow	3	−37	−76	6	−107

Two things flatter the FY25 base against which FY26 growth is measured: a ₹17.54 Cr deferred tax credit (98.8% of it booked in Q4FY25, producing PAT ₹33 Cr on PBT ₹15 Cr) and ₹15.84 Cr of impairment the auditor said should have been taken and was not. Roughly ₹33 Cr of the ₹50 Cr FY25 PAT was non-operating. Use PBT growth — ₹32.51 Cr → ₹117.04 Cr, 3.60x — as the clean measure, not the +76% headline.

Where the FY26 margin actually came from. Quarterly OPM ran 22% → 30% → 23% → 41%. The reviewed nine months delivered 25.0%; Q4 alone contributed ₹105 Cr of the ₹149 Cr FY26 EBITDA. Q4FY26 is explicitly a balancing figure between the audited year and the limited-reviewed 9M[Note 9, FY26 results] — so the reported margin expansion rests on the one quarter no reviewer touched.

A6–A7 · Regulation, coverage and capital allocation

No broker covers this stock. No consensus, no forward estimates, for Kernex or any peer[raw.json analyst_consensus]. The only third-party analytical view is CARE BBB−/A3, assigned 04-Nov-2025 on ₹275 Cr of limits, covenants "Not applicable" — and it was struck on the FY25 balance sheet, before the ₹2,465.71 Cr order and the entire FY26 cash burn.

The rating is not merely stale — the company was unrated for ~42 months. Prior ratings were CRISIL's, last dated 27-Apr-2022; CARE's was an assignment, not an upgrade. A three-and-a-half-year gap in external credit surveillance, ending one notch above sub-investment grade, now sits over a balance sheet gone 0.26x → 0.66x geared with ₹1,000 Cr of borrowing authority in flight. [devil gate — missed by all five sections]

Capital allocation: zero dividend in every year FY2015–FY2026. Two preferential issues — 13,00,000 shares (2022) and 13,00,000 warrants (2023) — every allottee a non-promoter. CARE records ~₹100 Cr infused across FY23–FY24 "to fund operational losses and for working capital"; fixed assets over those two years moved ₹24 Cr → ₹24 Cr. Outside money funded the turnaround and the promoters were diluted for it; they have contributed no fresh equity in five years and bought no shares on the open market in twelve quarters.

⚠️ Aggregator warning: screener.in's own peer group for KERNEX returns textile companies (KPR Mill, Vardhman, Trident). BSE classifies it "Transport Related Services"; CARE classifies it "Electrical Equipment". All peer comparisons in this report use a hand-built comparable set.

A8 · Balance sheet and cash-flow fraud filter — the central tension

The single most important chart in this report: accounting profit and operating cash flow moving in opposite directions, at scale.

The FY26 cash bridge — it reconciles exactly, and it is the most informative object in the file

Operating profit before working-capital changes	+170.80
Δ Inventories	−322.48
Δ Trade receivables	−308.19
Δ Trade payables	+422.06
Other working capital	−58.27
Net cash from operating activities	−99.74
Bank working-capital borrowing raised	+108.23
Inter-corporate deposits raised counterparty not named	+14.65
Closing cash	1.46

Read plainly: the company raised ₹544.94 Cr of external funding in one year — ₹422 Cr from suppliers, ₹108 Cr from banks, ₹15 Cr from undisclosed ICDs — to carry a ₹630.67 Cr build in inventory and receivables, and finished with ₹1.46 Cr of cash. The ₹88.24 Cr of accounting profit is a residual of that machine, not an output of it.

Working-capital days: the cash conversion cycle "improved" only because payables (938 days) now exceed inventory (787 days). Strip payables and the gross cycle went from 252 to 1,055 days.

The twelve-year record puts FY26 in context. Cumulative FY15–FY26: accounting profit +₹10 Cr, operating cash flow −₹261 Cr. Kernex has never produced a positive cumulative CFO in any rolling multi-year window on file. FY26 is not an aberration in that record — it is its largest instance.

What the fraud filter actually found

Not fabricated revenue. The execution is physically verifiable and third-party certified; a government PSU confirmed 30% supply; the inventory's origin was disclosed four months before the balance sheet needing explanation; counterparties are sovereign; the aged receivable tail is ₹4.23 Cr with ₹3.10 Cr already provided; tax is paid at a normal 24.6%; consolidation adds only 1.0% to revenue (the JVs drag, they don't flatter); and no insider has monetised anything — no dividend, no buyback, no QIP, no promoter sale, no auditor resignation.

But four things stand unrebutted: Q4 carrying 59%/71%/77% of revenue/EBITDA/PAT in an unreviewed balancing quarter, two years running; ₹9.15 Cr of Avant-Garde carrying value that the FY25 auditor said "stands impaired fully" surviving into a clean FY26 opinion, supported by an unaudited subsidiary's return to profit; zero provision against ₹371.32 Cr of inventory built for a lapsed order, of unstated RDSO specification version, while warranty provisions were raised 3.4x; and 938-day payables funding the balance sheet.

The single disclosure that would settle it: an inventory ageing-and-specification note — the ₹371.32 Cr split by Kavach version (v3.2 vs v4.0), by ageing bucket, by earmarked purchase order, with the NRV test shown. One table resolves the obsolescence, margin-deferral and Q4 questions simultaneously, because inventory is the hinge all three turn on. Due in the FY26 annual report, ~Sep-2026.

The FY25 qualified opinion — cured, or absorbed into a bigger year?

FY25 carried a QUALIFIED opinion: profit and Other Equity overstated by ₹15.84 Cr because two impairments were not taken — Avant-Garde ₹12.76 Cr (which the auditor said "stands impaired fully") and Kernex TCAS JV ₹3.08 Cr. FY26's opinion is unmodified. But the provisions actually taken were ₹3.61 Cr + ₹4.03 Cr = ₹7.64 Cr.

₹15.84 Cr flagged. ₹7.64 Cr provided. ₹8.20 Cr difference, unreconciled in any filing. The JV leg reconciles properly. The Avant-Garde leg does not: a full impairment became a 28% impairment on the strength of a return to profit at an entity the auditor explicitly states is UNAUDITED — in the one year where a ₹9.15 Cr charge would have been trivially absorbable against ₹88 Cr of PAT. Partially cured, partially absorbed — and the wrong direction for a company that had just been qualified.

Counterweight, stated fairly: the auditor who issued that qualification was not replaced and signed the next year's accounts. Companies with something to hide change auditors after a qualification. Kernex did not.

A10 · Valuation — three lenses on the same price

The same price is 41.7x on audited full-year accruals, ~138x on the limited-reviewed nine-month run-rate, and infinite on cash. The 96-point gap between the first two numbers is the market's willingness to underwrite one unreviewed balancing quarter. That is not a rounding difference — it is the valuation.

Lens	Earnings base	Implied multiple at ₹2,194.90
Reported FY26	EPS ₹52.58	P/E 41.7x
9MFY26 annualised (strips the balancing quarter)	EPS ₹15.86	P/E ≈ 138x
Reviewed-margin EBITDA (25.0%)	₹107.6 Cr	EV/EBITDA ≈ 35.7x (vs 25.8x reported)
Cash earnings	OCF −₹99.74 Cr	No positive multiple exists

The crux: Kernex versus HBL Engineering

HBL is the only listed comparable in the same three-OEM Kavach oligopoly (Medha Servo is unlisted). Kernex trades at an 80% P/E premium to it.

Metric	KERNEX	HBLENGINE	Kernex premium
P/E	41.7x	23.2x	+80%
EV/EBITDA	25.8x	17.5x	+47%
P/B	14.83x	8.76x	+69%
ROCE	48.0%	58.4%	HBL higher
3y profit CAGR	86%	105%	HBL faster
Borrowings	₹162 Cr (0.66x geared)	₹67 Cr (~net cash)	HBL cleaner
Dividend yield	0.00%	0.28%	HBL pays

Peer-set correction — CG Power belongs in the Kavach conversation. The listed Kavach universe is small: Kernex, HBL Engineering, CG Power and a couple of other names (Medha Servo Drives is unlisted). CGPOWER was missing from the curated set used in this run and is added here: market cap ₹1,29,080 Cr, CMP ₹820, P/E 102x, P/B 16.2x, ROCE 26.7%, ROE 20.5%, TTM sales +21%, TTM profit +27% [screener, 30-Jul-2026].

But read it correctly: CG Power is a ₹1.29 lakh Cr diversified industrial — motors, transformers, railway signalling and semiconductors/OSAT — at 35x Kernex's market cap, with Kavach immaterial to its valuation. Its 102x multiple is driven by the semiconductor and T&D stories, not train protection. It is a theme peer, not a valuation comparable — and its rich multiple is NOT evidence that Kernex's 41.7x is cheap. HBL Engineering remains the closest true economic comparable, and the conclusion above is unchanged.

The premium is a Kavach-purity-and-optionality premium, not a quality premium. On every quality axis observable in this dataset, HBL is the better business and it trades at half the earnings multiple. Zero-premium reference band ≈ ₹1,220–₹1,454 — i.e. the market currently pays roughly ₹1,245–₹1,640 Cr of market cap above HBL parity for purity and order-book cover. That is the number the bull case has to earn.

What genuinely supports part of it: Kernex's order-book-to-market-cap of 0.96x (conservative ex-GST basis) against HBL's 0.30x — 3.2x the contracted forward revenue per rupee of market cap, an advantage that survives the GST strip and doesn't fully break until realised share falls below ~40%. ⚠️ HBL's book is a SECONDARY media figure

Peer positioning — Kernex's returns are strong, but so is what you pay for them.

Reverse-DCF — what the price already requires

At a 15% cost of equity, ₹3,680 Cr of market cap requires ~37.9% revenue compounding for ten straight years to roughly ₹10,800 Cr by FY36, at a 23% EBIT margin, with capital efficiency improving ~43% from FY26 actual. For scale: that is ~3x HBL Engineering's entire current revenue, and about 20x CARE's own positive rating trigger of "TOI above ₹500 Cr". Not achievable on the Kavach point-ATP opportunity alone — it requires Kernex to become a broad signalling and industrial-automation company via the BHEPL moving-block JV (not commenced at 31-Mar-2026) and the Jindal LOI (₹15.90 Cr). The price capitalises those optionalities as if they were executing.

⚠️ Devil-corrected: at the solution point the explicit-period FCFF is negative and terminal value is ~104% of EV — this reverse-DCF is 100% terminal value.

Order-book sensitivity — what realised economic share does to the multiple

Base ₹3,517 Cr = headline ₹4,150 Cr ÷ 1.18 (GST rate derived from the company's own two disclosures of the same CLW order). Converting over 3 years at the reviewed 25% margin:

Realised share	Annual revenue	Implied EPS	P/E at CMP
100%	₹1,172 Cr	₹98.7	22.2x
70%	₹821 Cr	₹68.2	32.2x
50%	₹586 Cr	₹47.8	45.9x
35%	₹410 Cr	₹32.5	67.5x

At the current price the market underwrites roughly 60–75% realised economic share converting over three years at reviewed margins, for a ~30x forward P/E. The "order book exceeds market cap" framing is not a valuation argument — it compares a GST-inclusive, partner-inclusive, multi-year gross contract value to a net present equity value, and survives neither the GST strip (0.96x) nor the company's own scope definition (0.89x).

Quality scores

Piotroski F-Score 3/9 (weak). The diagnosis is in the composition: every point Kernex loses is a cash-flow or balance-sheet point; every point it wins is a P&L point. A company with 48% ROCE scoring 3/9 is not a contradiction — the income statement and the balance sheet are describing two different companies.

Altman Z 4.62 — nominally "safe", but 71% of it is the share price. X4 (market value ÷ liabilities) contributes 3.29 of the 4.62. The Z-Score says Kernex is safe largely because the market says it is worth ₹3,680 Cr, which is circular. At half the current price Z falls to 2.97 — the grey zone. True accumulated retained earnings are only ~₹35 Cr; the ₹231 Cr of "reserves" is mostly securities premium.

A11 · Order book — five disclosures, three different bases

As of	Value	Basis
Mar-2025	₹2,124.16 Cr	Total; only time the own-scope / with-partners split was published
30-Sep-2025	₹2,563 Cr	Outstanding (CARE)
14-Oct-2025	₹3,346.35 Cr	Cumulative since Apr-2024, incl. GST
31-Dec-2025	₹3,268 Cr	Kernex own scope, EXCLUDING GST — the only defensible entitlement figure
29-May-2026	₹4,150 Cr	Aggregate outstanding, incl. GST, consortium-wide; CLW 30% supplied

Devil's finding — the order book's own composition note is arithmetically impossible on its stated date. Note 7 says ₹3,268 Cr own-scope ex-GST at 31-Dec-2025, "of which ₹2,500 Cr relates to Loco TCAS from CLW and BLW". But at 31-Dec-2025 the only live loco order was CLW's ₹1,730 Cr ex-GST — BLW's first loco order did not exist until 11-Feb-2026. Either the figure is dated at the filing rather than the stated date, or there are undisclosed BLW awards. Every section of this analysis anchors its conservative case on a figure whose own composition does not survive its stated date.

The January 2026 CLW sequence — read it as one event

CLW refused the extension and let ₹2,041.40 Cr lapse (09/10-Jan-2026), then five days later awarded a 21% larger order on the same 12-month clock (15-Jan-2026). Both halves matter: the relationship is intact and the company's "eligibility unaffected" claim was validated by the counterparty within five days — strong evidence the constraint was RDSO's certification clock, not Kernex's capability. But CLW did not grant an extension when it could have. It re-tendered.

The devil's reading, which no analyst priced: the replacement order is 21% larger in value and 21% larger in units — ₹81.54 lakh/set against ₹81.66 lakh. Kernex re-won the same work at a fractionally lower unit price, on the clock it had just failed, with no escalation clause disclosed. That is equally consistent with a re-tender Kernex had to win at L1 to avoid stranding ₹371 Cr of inventory — i.e. with the customer capturing the negotiating leverage the lapse created.

A12 · Walk vs talk — and the sixteen-month silence

There has been no investor presentation and no concall since 24-Mar-2025 — sixteen months. Nothing for Q1FY26, Q2FY26, Q3FY26 or the FY26 audited results. Inside that silence: revenue quadrupled against last-presented figures, the largest order in company history lapsed and was replaced, operating cash flow swung to −₹99.74 Cr, a qualified audit opinion was issued and only partly resolved, ₹1,000 Cr of borrowing authority was sought, and market cap tripled. Not one of these was ever explained in a forum where a shareholder could ask a question — and no broker covers the stock.

Said (24-Mar-2025)	Delivered	Verdict
TCAS 4.0 ISA by April 2025	RDSO v4.0 approval 14-Oct-2025	❌ ~6 months late
CLW delivery from end-May 2025	Extension refused; order lapsed	❌ ₹2,041 Cr lost
NCR-A by June 2025	Commissioned 02-Jul-2026, ISA-certified	❌ ~12 months late, but delivered
Capacity 450 Kavach units/month	Never re-disclosed	⚠️ gap
Order book ₹2,124 Cr	₹4,150 Cr at 29-May-2026	✅ beat
New products (NMS, radio modems, pulse generators)	No update in any filing	⚠️ gap

Net pattern: Kernex consistently WINS more than it guides and EXECUTES slower than it guides. Certification dependency is the recurring slippage vector, and it cost them ₹2,041 Cr.

The counterweight, stated fairly. Where Kernex does speak, it speaks well. The 02-Jul-2026 filing is specific and independently verifiable by third parties — 39 stations, 10 WAP-7 locos, 4,600+ RFID tags, 17 towers, 48-core OFC, ISA by Italcertifier, Hitachi interlocking integration. And the 10-Jan-2026 filing disclosed a lapsed ₹2,041 Cr order promptly, with the reason stated — bad news released fast; a company managing a narrative does not file that. The problem is frequency and forum, not integrity.

A13 · Risks, ranked by what actually kills the thesis

#	Risk	Severity	What confirms / refutes it
1	A second missed 12-month clock. ₹1,726 Cr remaining on the CLW order in ~7.5 months ≈ ₹230 Cr/month, against FY26 revenue of ₹430 Cr for the entire year — roughly 6.4x the FY26 monthly rate	🔴 High	Refuted by Q1FY27 ≥₹250 Cr and ≥50% supplied by Sep-2026. Confirmed by any Reg-30 extension request
2	Funding cliff. Cash ₹1.46 Cr; borrowings 92.6% of the ₹175 Cr sanctioned line; payables ₹442.56 Cr at 938 days. Total identifiable liquidity ≈ ₹14.4 Cr	🔴 High	The ₹1,000 Cr borrowing ballot is the company's own admission. Refuted by enhanced limits + positive Q1FY27 CFO
3	Inventory obsolescence. ₹371.32 Cr built for the lapsed order, RDSO spec version never disclosed, zero provision — while warranty provisions rose 3.4x	🔴 Medium	The FY26 annual report inventory ageing-and-specification note (~Sep-2026)
4	Earnings quality. Q4 = 59% revenue / 71% EBITDA / 77% PAT in an unreviewed balancing quarter — two years running	🔴 Med-High	Refuted by Q1FY27 ≥₹200 Cr at ≥30% OPM. Confirmed by a print near Q2/Q3FY26's ₹47–73 Cr
5	Single sovereign counterparty. CLW ≈ 53% of the outstanding book; first non-railway order ever was 28-May-2026 at ₹15.90 Cr	🔴 Medium	Non-railway intake reaching double-digit % of revenue
6	Consortium joint-and-several liability on ₹771.70 Cr of undisclosed economic share	🟡 Medium	Per-order share disclosure in the FY26 AR
7	Hidden leverage in payables. ₹28.15 Cr finance cost on ~₹101 Cr average debt = ~28% implied, ~2.5x a BBB− rate	🟡 Medium	The Ind AS 7 supplier-finance note, mandatory in the FY26 AR (G.S.R. 549(E), verified)
8	Governance process. Retrospective ratification of ~10 months of related-party pay; independence at exactly the 33.3% LODR floor	🟡 High	FY26 AR related-party note; the postal-ballot result
9	Volatility, not illiquidity. 3.14% average daily move; 57% of sessions ≥3%; −19.9% limit-down band; ASM LT Stage 1	🟡 Current	Position-sizing constraint — see Part B

Risks 1–4 are not independent — they are four views of one object. The inventory exists because execution slipped; the funding gap exists because the inventory and receivables were built; the Q4 print is the accounting expression of the same ramp. A single Q1FY27 disclosure package — revenue, operating cash flow, and a supplied-% update — moves all four at once.

A15 · Ownership — flow in, but not smart money

Promoter 28.73%, FII up 8x in one quarter to 2.48%, DII a completed exit to 0.05%.

Cohort	Position	Read
Promoter	28.73%	Down 41bps over 8 quarters. Only 19bps reconcile to the Jun-2025 reclassification and ESOP dilution; ~37,000 shares (0.22% of equity) genuinely left the count — immaterial, below the SAST 2% threshold, but not nil
FII	0.30% → 2.48%	~₹91 Cr, 3.7% of free float — but ten unnamed entities averaging 0.25% each, 1.38% of it in the lower-diligence Category II bucket, and this line has liquidated to 0.07% twice in eight quarters. Built into a 115% quarterly move: confirmation, not anticipation
DII	1.11% → 0.05%	A completed domestic institutional exit. Built around the Dec-2024 order and the last-ever investor presentation, liquidated over four quarters into weakness, gone before the doubling
Retail	34.56% in sub-₹2L tickets	More than the entire promoter stake. Holder count 2.85x since Sep-2023 while public % stayed flat — average public holding fell 62%, the textbook signature of distribution into retail at prices from ~₹390 to ₹2,195

Devil's finding — 85% of the promoter block is non-resident, and no analyst section noticed. Of 48,74,581 promoter shares, 12 of 16 holders are foreign individuals / NRIs holding 24.82% of equity, against 4.27% held by Indian individuals and HUFs. Set beside an MD who signs the FY26 audited results from Singapore, a 100%-owned UNAUDITED US subsidiary sitting in the component-procurement path, and stated operations in Egypt with no Egyptian entity anywhere in the disclosed group structure. Individually explicable; together they describe a control structure whose economic centre of gravity may sit outside India. Not an allegation — a connected set of facts the desk should have surfaced.

Pledge — get the denominator right. The FY25 AR figure is 4,28,297 shares = 60.07% of ONE promoter's individual holding, which is 8.79% of the promoter block and 2.55% of equity. The pledgor is the 4.25% holder, not the 14.80% principal promoter. The pledge does not touch the control block. ⚠️ But a fresh encumbrance filing by the largest promoter exists as an unreadable scanned image — and the devil found its own filename embeds 23-04-2026 and "Reg11", not the 05-Jun-2026 / Reg 31(4) three sections attributed to it. Unverified in date, regulation and content.

Smart-money verdict: NEUTRAL. The flow arrow points in; the quality and timing of the money do not support calling it smart. No mutual fund, insurer or named FPI appears in any bulk or block deal in the stock's entire history — every identifiable counterparty across 2010–2026 is a proprietary desk, a broking entity, an HNI individual or a private company. The only well-timed large trade on record belongs to an individual: 90,000 shares bought at ₹910 on 27-Mar-2026, near the low, two months before the catalysts — by the same name that sold 167,800 shares at ₹1,383 six days after the previous mega-order.

A16 · Scenarios

Scenario	Prob.	FY27 rev	FY27 EPS	24m reference	vs CMP
Bear — second execution slip	30%	₹380 Cr	₹12.11	₹650 – ₹1,000	−70% to −54%
Base — delivery with drag	45%	₹700 Cr	₹61.21	₹2,374	+8%
Bull — book converts, cash turns ⚠️	25%	₹900 Cr	₹101.05	₹5,418	+147%

Expected-value check — it does not clear the required return. Probability-weighted: −14% over 12 months and +21.7% over 24 months (+10.3% annualised), against a 15% cost of equity. The weights that would flip this are roughly bull 35% / base 45% / bear 20% — i.e. the bull case would have to be more likely than the bear, which requires believing ~₹230 Cr/month of CLW delivery is achievable.

⚠️ Devil's flag on the bull case — the funding bridge is unsourced. Bull assumes ₹900 Cr of FY27 revenue and operating cash flow turning positive and interest falling. At the desk's own sales-to-invested-capital of 1.05x, ₹470 Cr of incremental revenue needs ~₹448 Cr of incremental capital — against ₹1.46 Cr of cash and 92.6% of the sanctioned line drawn. The bull case contradicts the mechanic the valuation section itself established: "the growth and the funding requirement are the same event." Read it as conditional on the ₹371 Cr of inventory converting to cash, or on the ₹1,000 Cr borrowing authority being drawn.

All price levels are research reference levels derived from stated assumptions. They are not targets and not forecasts. No company guidance exists — there has been none since 24-Mar-2025 — so every forward number here is the desk's own.

Part B · Technicals — Stage 2 intact, but no entry

B0 — Stage 2 advance, confirmed and structurally intact. The 30-week moving average has risen sixteen consecutive weeks (₹1,144.55 → ₹1,454.93); price has held above it for fifteen weeks since the 17-Apr-2026 cross; the breakout was volume-confirmed at 5.2x the base. It cannot mechanically roll into Stage 3 before roughly end-December 2026 at a flat price, because the weekly closes rolling out of the window are all ₹900–1,500 for the next five months.

B1 — but there is no setup. Price sits mid-box at 42.8% of a 17.9%-wide range, at the 95th percentile of its own one-year close distribution, +50.9% above the 30-week MA, with 14-session momentum at 1.11:1 and the SMA12 fractionally below the SMA26. Nearest structure is 4.5% above and 7.1% below — both further than the entire risk budget. ₹2,194.90 is not an entry.

B2 · Momentum — decaying on the near horizon

Window	Return
12 months	+110.8%
3 months	+62.7%
1 month	+1.6%

⚠️ RSI, MACD and TradingView TA are UNAVAILABLE (MCP servers not loaded — verified, not skipped). No OHL data in the series, so no candle, gap or intraday-range analysis and no VCP confirmation. No indicator values were invented.

B3 · Levels

Level	Distance	Meaning
₹2,447.00	+11.5%	52-week high — overhead
₹2,403.00	+9.5%	Box ceiling. Made 06-Jul on light volume and failed −15.2% within two sessions
₹2,194.90	—	CMP — mid-box, no pivot
₹2,038.60	−7.1%	Box floor. Break voids the setup thesis
₹1,990.90	−9.3%	50DMA. Loss damages the Stage-2 uptrend
₹1,757.25	−19.9%	Limit-down floor — no exit available at this price
₹1,454.93 – ₹1,497.22	−31.8% to −33.7%	Formal Stage-2 invalidation (vendor-anchored)

B4 · Relative strength — outright leader, fading

+111% over one year against HBLENGINE's +22% and a peer median near −21% — #1 of 7, at 5x the next name. But the return decay above says RS is fading on the near horizon. no index series available — sign of 1-month RS vs an index is unknown

B5 · Risk : reward — the instrument fails the user's own framework at this price

The disqualifier is volatility, not thinness — and the common objection is the wrong one. NSE liquidity is fine: ₹28 lakh of exposure is 0.59% of the 220,820-share daily average (₹48 Cr/day), and volume expanded on the two worst down days. What kills it is that the average session moves 3.14%, 57% of sessions move ≥3% and 81% move ≥1%, on top of a −19.9% limit-down band that no stop can survive and an ASM Long Term Stage 1 flag.

A 3% stop here is not a risk-control device — it is a 57%-per-session lottery ticket. Full-capital deployment into this name is a large-stop trade wearing a small stop's clothing, and the gap risk is uninsurable at any stop size.

⚠️ Devil adjudication: the "~11,000 shares/day, genuinely thin" figure repeated in three analyst sections is BSE-only, and BSE is ~5% of this stock's volume. This report carries the consolidated NSE figure.

The reconciliation. The ≤3% stop rule and this instrument are compatible at exactly two prices — and nowhere else on the chart.

Trigger	Entry	Stop	Risk	R:R	Confirmation required
1 — Pullback (preferred)	₹2,020–2,060	₹1,985	−2.70%	6.6:1	Contracting volume into the zone; delivery ≥47% on the turn; daily close back above ₹2,060 — not the touch
2 — Breakout retest	~₹2,410	₹2,350	−2.49%	~4:1	Daily close >₹2,403 on volume >220,820, then a retest holding ₹2,380+. Light-volume breakouts on this stock have failed 100% of the time (n=1)

Delivery % — the distribution warning fired in June, not now

Month	Delivery %	Read
April	45.2%
May	43.9%
June	38.9%	Price rose +126% on falling delivery — genuine churn, and it preceded the July stall
July	42.9%	Reversed, on contracting volume; the month's best day carried its best delivery at 56%

Counter-signal: July down-day volume (235,272) exceeds up-day volume (209,981) — a ratio of 0.89 versus June's 1.25. Net: two-sided consolidation, no clear accumulation, no clear distribution. Critically, today's tape matches none of the three legs of the January-2026 distribution template (3,240,816 shares / 21% delivery / at the high, then −36.5% over ten weeks) — the strongest thing that can be said in the setup's favour.

Action: watchlist, two alerts, no position. ₹2,040 and ₹2,403. Everything else on this chart fails the arithmetic.

Live chart · NSE:KERNEX

⚠️ Static TradingView captures were unavailable this run — the desktop app's chart tabs were occupied by intraday scanner overlays and the display mapping in the project mandate no longer matches this machine's monitor arrangement. The live widget below and the data-derived charts above stand in for them.

Part C · Consolidated VIEW

C1 · Investment case

Kernex is one of only three RDSO-authorised Kavach OEMs, and in FY26 it converted that licence into a genuine operating step-change: revenue ₹430.22 Cr (+127%), PAT ₹88.24 Cr, ROCE 48%, from ₹20 Cr of revenue two years earlier. The execution is real and independently verified — 572 of NCR's 652 Kavach v4.0 route km, 207 km commissioned, ISA-certified by Italcertifier of Italy, clean Hitachi interlocking integration. You cannot fake commissioned, ISA-certified track, and the counterparty is sovereign.

But the same year produced operating cash flow of minus ₹99.74 Cr and closing cash of ₹1.46 Cr, because ₹630 Cr went into inventory and receivables funded by ₹422 Cr of supplier credit at 938 days and bank borrowing now at 92.6% of the sanctioned line. Over twelve years Kernex has produced ₹10 Cr of cumulative accounting profit and consumed ₹261 Cr of operating cash — FY26 is not an aberration in that record, it is its largest instance. And 59% of revenue, 71% of EBITDA and 77% of PAT landed in a Q4 the company itself discloses as a balancing figure, whose reviewed nine-month margin was 25%, not the headline 35%.

C4 · Alignment — fundamentals and technicals AGREE

This is not a diverging setup, and that convergence is the finding. Fundamentals say: real business, real order book, but accrual-only earnings, a supplier-financed balance sheet, and a price demanding a decade of near-impossible compounding. Technicals say: genuinely Stage 2 and genuinely intact, but extended, momentum decaying, and no entry that survives a tight stop.

Both point to the same action — wait, with a dated trigger. When a richly-valued fundamental picture and an extended-but-unbroken technical picture agree that the next move is to do nothing, the probability that "nothing" is correct is materially higher than when they conflict. The convergence also tells you what changes the answer: a pullback to ₹2,020–2,060 fixes the technical objection without fixing the fundamental one; a strong Q1FY27 fixes the fundamental objection and would likely remove the technical entry at the same time. The cleanest outcome for a buyer is the rarer one — a soft tape into ₹2,040 followed by a Q1FY27 beat.

⚔️ Devil's Objections — the adversarial gate rejected 4 of 5 analyst sections

The desk's red-team recomputed every load-bearing number independently. It found 11 claims supported, 7 weak, 5 unsupported, and forced corrections across four sections before this report could ship. The five bear points below survived that process and are carried verbatim.

85% of the promoter block is non-resident, and nobody in five sections noticed. Of 48,74,581 promoter shares, 12 of 16 holders are foreign individuals / NRIs holding 24.82% of equity, against 4.27% held by Indian individuals and HUFs. Set that beside an MD who signs the FY26 audited results from Singapore, a 100%-owned unaudited US subsidiary in the component-procurement path, and stated operations in Egypt with no entity for it anywhere in the group structure. Falsifiable: FY26 AR related-party and geographic-segment notes, ~Sep-2026.
The order book's own composition note is arithmetically impossible on its stated date. Note 7 says ₹3,268 Cr own-scope ex-GST at 31-Dec-2025, "of which ₹2,500 Cr relates to Loco TCAS from CLW and BLW" — but BLW's first loco order did not exist until 11-Feb-2026. Every section anchors its conservative case on a figure whose own composition does not survive its stated date. Falsifiable: FY26 AR order-book note.
The bull case needs ₹450 Cr of working capital it never sources, and the report says so itself. ₹900 Cr of FY27 revenue at the desk's own 1.05x sales-to-invested-capital requires ~₹448 Cr of incremental capital against ₹1.46 Cr of cash. The valuation section's own words: "the growth and the funding requirement are the same event." The bull case is the one scenario that contradicts the mechanic the same document established.
CLW's re-award in five days is being read as a relationship signal when it is equally a re-pricing signal. The replacement order is 21% larger in value and units — ₹81.54 lakh/set against ₹81.66 lakh. Kernex re-won the same work at a fractionally lower unit price, on the clock it had just failed, with no escalation clause disclosed. No section priced the possibility that this was a re-tender Kernex had to win at L1 to avoid stranding ₹371 Cr of inventory — i.e. that the customer captured the leverage the lapse created.
The rating is not merely stale — the company was unrated for three and a half years. Prior ratings were CRISIL's, last 27-Apr-2022; CARE's BBB−/A3 was an assignment, not an upgrade, on 04-Nov-2025. A ~42-month gap in external credit surveillance, ending one notch above sub-investment grade, now sits over a balance sheet gone 0.26x → 0.66x geared with ₹1,000 Cr of borrowing authority in flight and covenants recorded as "Not applicable." Falsifiable: CARE surveillance ~Nov-2026.

Corrections forced before publication (all applied): loss-making years 10 → 8; negative-EBITDA years 10 → 9; CLW book concentration 71% → 53%; undisclosed consortium scope ₹792.73 Cr → ₹771.70 Cr (3 of ~8 disclosed, not 2); top-three promoter aggregate 23.13pp → 22.85pp, with the "not one share sold" claim withdrawn as circular; the "highest P/B in a decade" claim withdrawn (Dec-2024 was ~21.7x); reverse-DCF terminal value corrected (~37.9% CAGR, not ~37%); a "6.5x revenue ramp" figure deleted as underivable; and a BSE-only volume figure that three sections used to call the stock "thin" replaced with the consolidated NSE figure — which also removed the liquidity leg from the cost-of-equity build.

C5 · Catalysts, dated

Q1FY27 results — due by 14 Aug 2026 (not yet declared, no board meeting intimated). The single highest-information event on the calendar: it moves risks 1, 2, 3 and 4 simultaneously.
Postal ballot result — ₹1,000 Cr borrowing headroom and charge-creation authority; voting closed 29-Jul-2026.
FY26 annual report, ~Sep 2026 — the inventory ageing-and-specification note would settle the whole file.
Q2FY27, ~14 Nov 2026 — CLW should be ≥55–60% supplied on a linear glidepath.
CARE surveillance, ~Nov 2026 — first review since assignment, and the first ever on this balance sheet.
CLW clock expiry, ~15 Jan 2027.

C6 · Watchlist verdict — MONITOR (dated)

Not TRACK, because there is no entry within the user's risk framework at today's price and the decisive disclosure is fifteen days away. Not PASS, because the business is real, third-party certified, carries 7.6x revenue cover on a conservative order-book basis, and is one of three players in a mandated national safety rollout. Set two price alerts (₹2,040 and ₹2,403) and one calendar alert (14 Aug 2026). Re-rate on the Q1FY27 print, not before.

Primaegis Research Opinion · Internal Analyst View

NEUTRAL

Conviction: Moderate-High — unusually well-evidenced in both directions, which is precisely why the answer is "wait" rather than "act"

Horizon: Decision point 14 Aug 2026 (Q1FY27); full resolution ~Jan 2027 (CLW clock)

Risk profile: Very High / Speculative — ASM LT Stage 1, 3.14% average daily move, zero broker coverage, BBB− after a ~42-month unrated gap

⬆ Upgrade to ACCUMULATE if ALL of:

Q1FY27 (by 14 Aug 2026) revenue ≥ ₹200 Cr at ≥30% OPM, in a limited-reviewed quarter
Q1FY27 operating cash flow positive, or inventory down materially from ₹371.32 Cr
CLW supplies disclosed ≥50% complete by the Q2FY27 print (~14 Nov 2026)

⬇ Downgrade to REDUCE / SELL if ANY of:

Any Reg-30 filing seeking or reporting an extension, cancellation or LD claim on the ₹2,465.71 Cr CLW order — the exact sequence that preceded the January-2026 lapse
Q1FY27 revenue < ₹150 Cr, proving Q4FY26 was a pull-forward
Any inventory write-down, or Q1FY27 gross margin below the 22–30% of reviewed FY26 quarters
CARE downgrade below BBB−, or an "Issuer Not Cooperating" designation
Borrowings above ₹300 Cr at any reporting date without a corresponding revenue step-up

Rationale: The thesis is intact but fully priced. The desk's own probability-weighted expected return is −14% over 12 months and +10.3% annualised over 24 months against a 15% cost of equity — it does not clear the required return on our own weights. Meanwhile HBL Engineering, the only listed comparable in the same three-OEM oligopoly, earns a higher ROCE, grows profit faster, carries effectively no debt, pays a dividend, and trades at half the P/E. Buying here is a bet that ₹230 Cr per month of delivery happens on a clock the company has already failed once — and that bet is available fifteen days from now with far more information, and quite possibly at a better price.

Entry discipline, if and only if the fundamental gate is passed: ₹2,020–2,060 with a ₹1,985 stop (−2.70%), or a volume-confirmed break above ₹2,403 entered on the retest at ~₹2,410 with a ₹2,350 stop (−2.49%). No entry at ₹2,194.90 under any circumstances.

STRONG BUY

BUY

ACCUMULATE

▶ NEUTRAL

REDUCE

SELL

STRONG SELL

⚠️ SEBI: Internal research opinion only. Primaegis Research is not a SEBI-registered investment adviser or research analyst. Not investment advice. Research reference only.