Every defence deck doing the rounds this month opens with the same stat — 86% of India’s FY26 defence export growth came from public-sector units. Bulls read it as proof the sector is on fire. Bears read it as proof the private-sector story has stalled. Both are reading the same number, and both are wrong. This issue is about why, and about the leg of the trade that actually holds up when you pull it apart.

If you only take one line from this — stop scoring the private defence thesis on the export headline. Score it on the domestic order book and production share. That leg is 4.6 times larger, it is cleanly attributable to the firm that builds the value, and it is where the mispricing sits.

The number everyone is quoting

Here is the FY26 export split, from the PIB release in April 2026. Total defence exports came in at ₹38,424 cr. Public-sector units (DPSUs) did ₹21,071 cr of that, up 151% from ₹8,389 cr. Private did ₹17,353 cr, up just 14% from ₹15,233 cr. Of the ₹14,802 cr total increase, DPSUs contributed ₹12,682 cr — roughly 86%. Private’s share of exports fell from 64% in FY25 to 45% in FY26.

On the face of it, that looks like the state sector pulling away and the private sector fading. It is neither.

Why the number is noise, not signal

Two reasons, and they compound.

First, it is lumpy. The DPSU jump is not broad-based ecosystem strength. It is a BrahMos delivery ramp — the Philippines deliveries that began in 2024, the Indonesia order worth around ₹3,800 cr, amplified by the battlefield exposure from Operation Sindoor. That is a handful of government-to-government deals maturing in the same year, not a structural export win you can extrapolate.

Second, and this is the part almost nobody models — it is mislabelled. India counts exports by exporter-of-record. When a private firm’s avionics or propellant sits inside a BrahMos, an Akash or a Pinaka that a DPSU ships abroad, that value is booked as a DPSU export, not a private one. So the ₹17,353 cr private line mostly captures direct component exports to foreign primes, and it structurally understates how much private industry is actually participating. The split cannot be read as a private-versus-public scoreboard. It was never built to be one.

The leg that actually is the trade

Now look at domestic production, which is where we would spend the analytical time.

FY26 production was ₹1.78 lakh cr, up 15.6% from ₹1.54 lakh cr in FY25. That is 4.6 times the size of total exports. And unlike exports, production is credited to the firm that builds the value. So when private share of production climbs — 21% in FY24 to 23% in FY25 — that is a real economic win, not an accounting placement.

This is the leg where “private sector rising” is measurable. It is bigger, it is attributable, and it is quieter — which is exactly why it is less priced-in than the export headline that gets the airtime.

The share is gaining at roughly 2 percentage points a year. The popular “30%+ private share by FY29” target needs that pace to accelerate. Watch the rate, do not assume it.

Three domestic clocks striking at once

The FY27 Union Budget (Feb 2026) reserved 75% of the capital-acquisition budget — ₹1.39 lakh cr of ₹1.85 lakh cr — for domestic sourcing. The Defence Acquisition Council cleared ₹3.3 lakh cr of Acceptances of Necessity in FY26, about twice the annual capital outlay. And 61% of the 5,012 items on the Positive Indigenisation List were indigenised by December 2024. All three feed the domestic order book.

An Acceptance of Necessity is an intent to buy, not a signed contract. The gap between AoN, signed order and actual delivery is exactly where the sector’s earnings risk lives. A big pipeline number is bullish for the narrative and roughly neutral for next year’s earnings. Watch conversion, not the pipeline.

The beneficiary map

Bharat Electronics (BEL) — the cleanest domestic compounder. Electronics, radar and electronic warfare are the highest-value-add layer of indigenisation. BEL pairs the largest domestic electronics book with real direct exports, and it is the DPSU least dependent on lumpy platform deals. Order book ₹73,882 cr (Apr 2026), around ₹30,000 cr of fresh orders in FY26 including roughly USD 346 mn of exports. Re-rated to a rich premium, priced for sustained mid-teens growth. Direction supported, margin of safety thin. A quality hold, not a mispricing. Risk is program concentration — one slip in a large program dents multiple years.

Bharat Forge — the under-priced overlap. ATAGS artillery, armoured platforms and the KSSL defence arm give private-margin domestic exposure, plus direct component exports to Western primes — which is the private export that actually gets counted. Defence order book ₹9,400 cr (Q2 FY26). Because it is bundled with a cyclical auto and industrial base, on a sum-of-parts basis the defence optionality is the least priced-in name on this list. The most interesting risk-reward if the mix keeps climbing. Risk is defence is still a minority of consolidated revenue, so the cyclical base can mask the ramp.

Hindustan Aeronautics (HAL) — the visibility leader, execution-gated. Tejas and platform indigenisation give it the longest domestic revenue runway in the sector — an order book of ₹2.54 lakh cr (Mar 2026), roughly 7 to 8 years of revenue. But that is 7 to 8 years of work, not of growth. PSU revenue recognition lags booking by 18 to 36 months, so delivery, not intake, is the binding constraint. Priced for near-flawless execution. Risk is single-customer (IAF) concentration and Tejas Mk1A tied to GE engine supply.

Solar Industries — the embedded-supplier angle, expensive. Migration from commercial explosives into rockets, missiles, loitering munitions and Pinaka, at private margins of 18 to 20%, and a likely embedded supplier to the very export platforms that inflated the DPSU line. Defence order book estimated at ₹6,000 to 8,000 cr and scaling. The theme is aggressively priced in. Justified only if defence share and export contracts keep compounding, with little cushion if either stalls.

Data Patterns — the purest and highest-beta. Pure-play defence electronics (radar, avionics, C4I) at the highest margins in the group, around 22 to 25%. Order book ₹673 cr (Q2 FY26). Direct leverage to the electronics layer, but a small and lumpy book means a single order delay swings a full year. Priced as a high-growth compounder on very rich multiples. The highest-beta way to play the theme, and the least room for error.

For the students: how we pulled this apart

The method here is worth more than the answer. When a single statistic is being used to argue two opposite conclusions, that is usually the tell that the statistic is measuring the wrong thing. Our steps this week:

One, we traced the number to its primary source (the PIB April 2026 release) instead of the decks quoting it. Two, we asked how it is constructed — and the exporter-of-record accounting rule is what broke the “private-versus-public” reading. Three, we looked for the bigger, cleaner variable in the same dataset — which was domestic production at 4.6x the export line. Four, we separated intent from contract (AoN versus signed order) so the demand pipeline did not get double-counted as earnings. The habit to build: when a number is doing too much narrative work, find the quieter number next to it that actually attributes cleanly.

What we are watching

Private share of the domestic order book and production — not exports. If private production share keeps climbing toward 30% and private domestic intake accelerates, the investable thesis confirms.

Execution on HAL Tejas Mk1A (GE engine deliveries) and MDL P-75I. Continued slippage invalidates the assumption, baked into current multiples, that visibility converts cleanly into earnings.

AoN-to-signed-contract conversion on the ₹3.3 lakh cr of FY26 AoNs. If materially less than half converts to signed orders within about 18 months, the demand pipeline is softer than valuations assume.