Twelve contract factories: earnings grew in both cohorts. The multiple moved in only one

Twelve companies, four plates. The first plate sorts them, the second shows what their prices did this year, the third draws the chain they all sit in, and the fourth shows which way each one is walking. The numbers come from the twelve company pages on this portal, read on 9 October, and from the deep-tracks that rebuilt nine of them over the last week.

The thing to hold on to: these companies all do roughly the same job, and the market spent 2026 treating half of them very differently from the other half.

1. One job, two cohorts

An electronics manufacturing services company is a contract factory. A brand designs a product; the EMS firm solders chips onto a board, puts the board in a box, tests it and ships it. The brand puts its name on it. Of the four steps on the strip above, EMS does the two in the middle.

The twelve names here split cleanly by who the brand is.

The AC makers build for Indian brands, in volume

Dixon, Amber, PG Electroplast, EPACK and Virtuoso build phones, TVs and room ACs for Voltas, Samsung and the rest. Thin margins (FY26 operating margins of 4–10%), seasonal, Indian customers, and the profit becomes cash: over five years the cohort produced ₹1.51 of operating cash for every rupee of profit. Dixon earns 3.8% on sales and still returns 29% on capital, because its inventory turns in a month.

The specialists build for industrial and defence customers, in small runs

Syrma, Kaynes, Avalon, Cyient DLM, Centum, Vinyas and Aimtron build boards and boxes for aerospace, defence, medical, industrial and semiconductor-equipment customers, much of it exported. Fatter margins (10–20%), order-book driven, and the profit has not become cash yet: the cohort's five-year operating cash is −0.18 of its profit. Receivables sit at 130 days against the AC makers' 60. Kaynes is the extreme, with ₹977 Cr of profit over five years against −₹615 Cr of operating cash.

Both cohorts have met about half of the financial guides they have closed (30 of 59 for the AC makers, 34 of 65 for the specialists). The AC makers slip dates more; the specialists keep their margins less.

2. Same multiple, reached two different ways

Both cohorts trade at about the same price for a rupee of earnings: a median P/E of 84 for the AC makers and 85 for the specialists. And both prices ask for about 25 points more growth a year than our base cases. On the multiple alone, the market makes no distinction.

It made one on the way there. Every price move on the plate is split exactly as earnings times multiple, so the teal and amber segments add up to the bar.

The specialists are 2.5x since January, and the P/E did most of it

Six of the seven specialists are in an up-phase that began between January and April 2026. The cohort's price is 2.53x on average: earnings 1.29x, multiple 1.89x. Cyient DLM is the purest case, 3.4x on earnings that did not move (0.98x) and a multiple that went 3.5x. Vinyas is the one name where earnings led. Kaynes is the exception the other way: its multiple halved between November and June, after its plant dates slipped and a broker questioned its accounts, and it has gone sideways since.

The AC makers sat it out

The AC cohort is 1.04x on average, 0.81x at the median. Dixon is up only as far as its earnings grew (multiple flat at 0.96x). Amber and PG Electroplast are falling; EPACK has lost 57% since September 2025. The cohort with the better cash record saw no change in its multiple at all.

So the sector's multiple did not move. A story's did. The higher multiple went to the names with a defence, export or semiconductor label, and the next two plates are about what that label means.

3. Where the money is in the chain

The strip in Plate 01 was four steps. Stretched out, it is six, and the margins are not evenly spread along it.

Making the chip is rich; using it is thin

Designing a chip and printing it on silicon carry the fat margins, and no Indian listed company does either. Packaging and testing the chip is the middle layer. Soldering packaged chips onto a board and building the board into a box, the two green boxes, is where every Indian EMS name in this piece earns its money, at 4–12% and 5–10%. The margin ranges per layer are typical industry figures, not from any filing; the two green boxes carry the cohort's own numbers.

The "semiconductor" names are mostly suppliers to the fab

Under the chip-making layers sit the companies that feed them: chemicals, equipment and parts, plant infrastructure. This is where most of the listed semiconductor story lives, and in every case it is one customer or about 1% of sales. Avalon, Centum and Cyient DLM each build parts for a single chip-equipment customer. Acutaas's chip chemicals are about 1% of revenue. Kaynes is the only EMS name that touches the chip itself, with a packaging plant at Sanand that expects to start billing in Q3 FY27.

4. Which way each specialist is moving

Every one of the seven specialists is building or filling a plant, and every one has a slide about moving to a higher-margin layer. They are not all moving in the same direction.

Two go backward, into the component

Syrma is building a bare-PCB plant in Andhra Pradesh, ₹1,500–1,600 Cr by FY30, with trial production due in Q3 FY27: the layer it currently buys from. Kaynes is going two layers back, into chip packaging and HDI boards, about ₹4,600 Cr of capex across the two. Its plant dates have moved five times, from FY25-end to Q3 FY27, and neither plant had billed by June 2026. These are the two largest companies in the cohort, and the two with the longest delays.

Three go forward, toward the system

Vinyas, Centum and Cyient DLM are moving the other way: from the board to the sub-system or system the board sits in, which they design. Vinyas has a 25,000 sq ft integration facility nearly built and its system orders went from ₹247 Cr to ₹324 Cr in ten months. Centum is furthest along, designing whole radars and satellite payloads, with about ₹520 Cr booked toward a ₹2,000 Cr FY28 goal. Cyient DLM is slowest: its design-led work was 6–7% of FY26 revenue, volume is promised for FY28, and the margin step to 11–13% is dated FY29. These are the three smallest companies here.

Two stay put and add capacity

Aimtron is filling a six-line greenfield and a US plant it bought, at the same layer. Avalon said no, out loud, to a component plant with incentive money on the table, and is instead taking its box-build model into new end-markets. Its new Chennai land has no size or cost attached yet.

One pattern across all seven: the plant comes first and the mix comes second. The design-led revenue is guided for FY28 almost everywhere, and almost nobody guides the FY27 margin up. Vinyas guides 11–12% against 12.5% delivered; Syrma 10.5–11% against 11.3%. The multiple in Plate 02 arrived before the margin in Plate 04.

What would make me wrong

• The cash gap is partly construction. A company growing 30–50% a year ties up cash in receivables and inventory by design; the specialists' negative operating cash says "not collected yet", not "never collected". If their order books convert on schedule, the cohort's cash record flips.
• The AC makers' flat multiple may be the market pricing a real problem: FY26 was a lost AC season (EPACK's revenue fell 12.7%, PG Electroplast missed its own guide by a fifth), and a cohort that lives on one Indian summer may carry a different multiple from one that lives on global defence budgets.
• The "same multiple" line hides a difference in what the P/E is on. Dixon's trailing profit carries fair-value gains on a stake, so its 41x understates it; Centum's loss years make its multiple hard to read at all.
• Plate 03's layer margins are typical industry ranges, and the actual spread between an Indian packaging plant and an Indian board plant may be narrower than the chart implies once incentives and utilisation are in.

The opposite error is to read the specialists' higher multiple as proof the move up the chain has already worked. On the filings, none of it has billed.

The seven specialists appear on this portal with every section rebuilt from their own filings; the one earlier story is Aimtron's bridge. Twelve factories, one job, and the market paid up for the seven that have a reason to be somewhere else in the chain by FY28. The plants exist. The mix does not, yet.

This is a summary of what twelve companies' filings, earnings calls and investor decks say. It is not investment advice or research.