Part A — Fundamentals

A5. Financials — audited consolidated, ₹ Cr

Q1 FY27 — and the margin decomposition that actually matters

Consolidated, ₹ Cr: Revenue 360.7 (+130.4% YoY, +17.9% QoQ) · Gross profit 164.2 (45.5%) · EBITDA 85.1 (23.6%, +199.7%) · Other income 7.9 · D&A 9.7 · Finance costs 15.8 · PBT 67.4 (+355.0%) · Tax 17.2 · PAT 50.2 (13.9%, +364.5%) · EPS ₹16.33 consolidated / ₹16.42 standalone. Reg-33 + press release 29-Jul-2026

Read: the entire Q1 FY27 margin expansion is fixed-cost absorption on 2.3× volume, not better unit economics. Gross margin deteriorated ~870 bps YoY and sits below both the FY26 full-year 47.7% and FY25's 49.4%. Whether Q1's gross profit itself contains a WIP credit cannot be determined — the Q1 expense detail was never extracted. GAP

Cash. FY26 CFO ₹196.90 Cr (from ₹101.25 Cr); capex ₹133.60 Cr; FCF ₹63.30 Cr on this desk's gross-capex definition. But CFO includes a +₹210.73 Cr swing in customer advances — strip that single line and CFO is −₹13.83 Cr and FCF is −₹147.43 Cr. Receivables absorbed ₹116.69 Cr and inventories ₹154.47 Cr in the same year. No dividend since FY2023. comparatives.md §E

A6. Regulatory — the tariff exposure, sized

A9. Quarterly P&L — last 8 quarters

Screener quarterly series, ₹ Cr, integers, standalone basis (proven on EPS: Q1 FY27 screener 16.42 = standalone; consolidated 16.33).

The cause of the step-change is identified, and it is not organic operating leverage alone: the ₹386 Cr Bloom Energy order of 10-Sep-2025, with ₹205 Cr executable by Mar-2026 and ₹181 Cr by Jun-2026. Reg-30 10-Sep-2025 Note that Sep-2025 (OPM 13%, PAT ₹5 Cr) is the trough that inflates every subsequent YoY comparison and drags the TTM multiple. Presentation defect: the Q1 FY27 deck's comparison column is labelled “Q4 FY25” but carries Q1 FY26 values.

A11. Order book

Order book up ~5.5× in twelve months, and Q1 book-to-bill was 8.03×. The build-up arithmetic is disclosed and self-consistent — a genuine positive disclosure marker in a file with several negative ones.

A12. Track record — walk vs talk

FY26: guidance raised mid-year, then missed on three of four metrics. EBITDA 19.5% against a 21% guide; order book ₹2,581.9 Cr against ₹2,800 Cr; revenue ₹876.2 Cr against a revised “₹900 Cr plus”. Management then raised FY27 revenue guidance 50% → 80% and capex ₹250–300 Cr → ₹500 Cr within ten weeks. Q1 delivered ₹360.7 Cr = 22.9% of the implied ₹1,577 Cr, with management guiding H2 > H1.

Charts — built from the audited series

Rendered by the desk from computed.json. The revenue, EBITDA and PAT series are the audited consolidated figures — FY26 ₹876.21 Cr and FY25 ₹676.00 Cr tie to the tables above exactly. An earlier build of these charts ran on screener's rounded standalone integers because the audited override silently failed to fire; the adapter now measures the filing's own units against the screener series and rescales, and refuses the override outright rather than publishing a chart it cannot reconcile. The return-ratio and working-capital panels remain on the screener series — computed.json carries those as NaN across all 13 periods — and show shape and direction only.

A8. Balance sheet + cash flows — the fraud filter

The 12-point checklist

2 flags · 5 cautions · 5 clean · 8 sub-tests formally withdrawn as unrunnable. No evidence of fabrication was found anywhere — every reconciliation attempted tied, several to the paisa. This is an earnings-quality and liquidity-structure problem, not a fraud problem, and the desk will not blur the distinction.

A10. Valuation

The earnings base — core-adjusted

On the incremental-strip basis FY26 core EPS is ₹18.05 against a reported ₹30.57, and core EPS grew ~5%, not ~78%. The trailing P/E on that base is 385.5× rather than 227.6×. On the complete bound (EBITDA ₹16.74 Cr) EV/EBITDA is 1,286× — published only to show that the EV multiple is not robust to the earnings-quality question at all.

A10.2 Peer table

Source asymmetry, stated: all peer figures are screener.in WEB SCRAPE — the screener MCP timed out on all four calls. MTAR's own figures are screener-auth MCP + audited filing. MTAR's numbers are better-sourced than its peers'.

Is the premium earned? Metric by metric, and the answers differ.

• P/E +48.9% vs the n=4 median — probably not a real premium at all. On annualised Q1 FY27 earnings (16.42 × 4 = 65.68) MTAR trades at 106.0×, essentially exactly at the peer median. Three qualifiers travel with that and none alone is honest: 16.42 is standalone; annualising one quarter assumes non-seasonality that is not established (management guides H2 > H1, which cuts the other way); and Q1 FY27 earnings quality is untestable — no quarterly inventory breakdown exists.
• P/B +112.2% — NOT earned. MTAR pays 26.0× book for a 12.5% ROE while Data Patterns earns a higher 15.2% ROE at 14.2× book.
• P/S −20.8% — NOT cheapness. MTAR has the lowest operating margin in the set ex-Dynamatic (21.1% vs 35.25%). A 21% revenue-multiple discount against a 40% margin discount is, if anything, still generous.
• EV/EBITDA +31.1% — partially explained, not earned.

Quality scores

Piotroski F-Score = 5 / 9 — and 3 / 9 advance-adjusted. The four failures are not random: #5 leverage, #6 current ratio, #8 gross margin and #9 asset turnover — the entire operating-efficiency and balance-sheet-discipline block failed in the year that reported a 78% EPS increase. The three passes that carry the score are ROA level, ROA direction and CFO — and tests 2 and 4 pass only because of customer advances. Strip that line and CFO is −₹13.83 Cr and the score is 3/9. Both are published; the desk does not resolve it.

A16. Scenario analysis

Provenance. The 1-year case is the built model's output, not the desk's: scenarios_2026-08-03.json, ok: true · confidence high · tier stated · band_source own · eps_base 30.99 (standalone) · guided_growth 142.7% (call 20-May-2026). Delivery multipliers are the store's own measured distribution — low 0.578 / mid 1.000 / high 1.359 on n=153 (76 annual guided-vs-filed pairs + 77 near-complete quarterly years), not chosen numbers. delivery_is_own: false — MTAR's own record is not in the set. The store built at ₹5,726; implied values are price-independent and unchanged, only the price comparison is restated to ₹6,959.

These are research reference levels produced by arithmetic on management's own filed guidance and this company's own filed-year valuation band. They are NOT price targets, NOT a recommendation and NOT a forecast. Do not read the +96.4% without the box above: the only row the model itself produced is the −4.0%.

The sensitivity that matters more than any of the above — the exit multiple

On base FY29 EPS of 162.30 (a DESK EXTENSION — the model publishes no FY29 EPS):

Read it both ways, because both readings are true. If you believe the desk-extension FY29 EPS of 162.30, the price must de-rate to 42.9× — below MTAR's own six-year post-IPO minimum — merely to stay flat. If you think ~40× is right for a matured, post-inflection MTAR, the base FY29 case is worth ₹6,492 — below today's price after three years of ~47% compounding.

No conventional DCF is published. Four required inputs are absent; the desk declines rather than approximates.

A13. Issues & risks

Severity 🔴 = can break the thesis · 🟡 = can dent it. Likelihood is desk judgement, labelled as such. Every early-warning marker is something an analyst can actually observe on a known date.

A15. Ownership — smart money

Direction: OUT — but the composition of “out” matters more than the direction

The one-line read: every domestic constituency that could identify itself — promoters, insurers, named mutual funds — reduced into the June-2026 peak, and an anonymous, index-weighted, 281-name foreign book took the other side. That is a distribution pattern by composition, not merely by percentage. But direction from 01-Jul-2026 onward is UNKNOWN and will remain so until the 30-Sep-2026 register.

A12b. Governance — 3.0 / 5

The governance risk in MTAR is not expropriation. There is no RPT abuse, no dilution, no delisting play, no pay grab and no litigation overhang — those boxes are genuinely clean. The risk is that the control environment and the disclosure function did not scale with a business that grew revenue 130% YoY, order book 5.5× and unit count 9 → 16 in six months — and that the one external check that should catch it, the credit rating, loosened its own binding trigger in the same period.

Part B — Technical read

B0. Weinstein stage — Stage 3 top formation, in progress, and explicitly UNCONFIRMED

The section argues this down rather than asserting it. Weinstein's test is defined on the 30-week simple moving average of weekly closes — not on any 50- or 200-day line, simple or exponential. That line, built as a 150-session proxy, is ₹5,077.87 and rising +30.43 in a single session (~+3.0% per week), with price +37.05% above it. Weinstein requires a flat-to-declining 30-week MA for a genuine Stage 3. This one is climbing hard. The verdict is therefore carried as unconfirmed, with two-sided falsification stated up front.

The moving-average evidence is split, and the section counts it NEUTRAL rather than arguing it selectively. The SMA50 has rolled over (6,959.75 → 6,941.37, falling ~9.7/session) with price only +0.25% above it; the EMA50 has turned up (from a 6,425.56 trough on 04-Aug) with price +7.82% above it. Two legitimate averages of the same 50-period length pointing opposite ways. The section withdrew its own bear item rather than keep the half that suited it — the right call.

Falsification, stated both ways: the Stage-3 read is negated by a close above ₹8,374.50, and confirmed by a fourth lower high forming under ₹7,101.

B2. Key levels

Price is above every average on the chart. That is the constructive fact, stated without hedging. The qualifier that belongs beside it, not instead of it: price crossed above the two 50-period lines on 05-Aug and 06-Aug — within the last two sessions, on locked limit-up closes. The stock sits 31.6% of the way up the May–June distribution range — the lower third — sandwiched between two mechanical walls exactly 5% apart.

B4. Risk:Reward — does any entry survive a ≤3% stop?

B3 & B5 — relative strength and what would change the read

Extreme long-run RS with dead short-run RS is the canonical Stage 3 divergence — and it is the item that carries the verdict when the moving-average evidence is counted neutral.

Part C — Investment thesis

C4. Alignment — ALIGNED (negative)

C5. Catalysts — what to actually watch

• Q2 FY27 (~late Oct 2026) — the decisive print. Read the “changes in inventory of WIP” line first. If the credit is again the majority of the YoY profit delta, the margin is inventory-borne. Revenue below ~₹300 Cr is the bear trigger.
• A delivery schedule for the ₹3,100.09 Cr order. If the Q3 FY27 deck (~Feb-2027) still carries none, a blanket PO of 3.5× revenue that has stood 9+ months undated should be excluded from any revenue bridge.
• Interest coverage each quarter. Q1 FY27 is already 5.39× against ICRA's live <5.0× trigger — in the best revenue quarter the company has ever printed. Annualised finance cost is ₹63.2 Cr, 2.15× all of FY26.
• The FY26 annual report — the only route to the CARO inventory opinion, the KAM, the NRV policy note and the Sec-197(12) median-pay ratio. And whether it publishes in machine-readable form at all.
• Series reverting BE → EQ in the daily bhavcopy — the only thing that restores normal price discovery.

C6. Watchlist verdict — MONITOR

Not TRACK: no entry survives a 3% stop, the stock is administratively untradeable in a 5% band, and four independent lenses point the same way. Not PASS: the capability is scarce, the order book is real and verifiable, the board structure is genuinely above average for an Indian smallcap, and management has not diluted in five years.

C7. PRIMAEGIS opinion — REDUCE

Provenance — how this was built, and where it disagreed with itself