India Has 5,865 Industrial Automation Startups. The Machines Still Come From China.

Robotics and industrial automation is India's largest deep-tech sector by company count — 5,865 tracked startups, more than health and life sciences, more than double semiconductors and computing combined. It is also one of the few sectors where Indian inventors out-file foreigners by a wide margin: 4,233 domestic patent filings against 1,484 foreign in 2024.

By the usual measures, this is the sector that is working.

Now here is the import bill it sits next to. India bought $19.5 billion of machinery in this category from China in FY2024-25. Over eight years, imports from China doubled — $9.5 billion in 2017 to $19.2 billion in 2024 — and China's share of the category rose from 29.4% to 39.0%.

The most startups. The most domestic patents. And a dependence that grew by ten percentage points while all of that was being built.

What India actually buys

The specific products are the least glamorous imaginable:

Embroidery machines at 91.4%. Cranes at 95.9%. Compressors at 69-74%. These are not frontier technologies. They are the mechanical guts of textile mills, construction sites and cold chains — the equipment that other Indian industries buy in order to produce anything at all.

The startups are somewhere else, mostly

Look at what the sector's patents are about and the shape of the mismatch appears.

Top technologies: AI-based imaging (2,254 patents), microcontroller control (1,883), robotic arm (987), ultrasonic sensor (884), machine learning (852), computer vision (567).

Top-funded companies: Addverb ($132M, warehouse automation), ZEDEDA ($127M, edge intelligence software), Infinite Uptime ($68.7M, predictive maintenance AI), Detect Technologies ($43.3M, AI workplace safety), QpiAI ($38.5M, quantum computing).

Then a control test. We searched the platform's entire technology taxonomy for "compressor." Zero results. In a registry of roughly 1,050 technology categories built from what Indian inventors file, the single largest imported machine category in this sector has no corresponding technology at all.

The patent filers tell the same story from the other direction. India's top domestic filer in robotics and industrial automation is Tata Consultancy Services — an IT services company — followed by Tata Steel and BHEL. The foreign filers are the actual machine builders: JFE Steel, TMT Machinery, Murata Machinery, Maschinenfabrik Rieter, Siemens, ABB, Bosch. Rieter and Murata make textile machinery. India imports 91.4% of its embroidery machines from China while Swiss and Japanese textile-machinery firms file the patents here.

Where we'd push back on our own argument

That framing is too clean, and the data doesn't fully support it.

The sector's top two patent sub-themes are textile machinery (154 patents) and construction automation (153) — which map directly onto the embroidery-machine and crane import lines. Ethereal Machines ($52M) builds in-house 5-axis CNC machines for aerospace and healthcare, which is real precision hardware, not software wearing a robotics label. Favo ($37M) builds brick-masonry construction robots. The alignment exists.

It is just very small, and very thinly funded. Across 513 funded companies in this sector, the median raise is $984,500. Twenty-six companies have reached Series A. Nine have gone beyond it. Against $19.5 billion of annual imports, 154 textile-machinery patents and a scattering of sub-million-dollar cheques is not a substitution pipeline — it's a signal that a few people have noticed.

So the accurate version is not "Indian startups are working on the wrong things." It is that the ones working on the right things are capitalised at roughly a thousandth of the scale of the problem, while the capital concentrates in the software-adjacent layer where returns arrive faster.

The tariff nobody argues about

One more structural fact, from a source we haven't used yet.

Running World Bank WITS tariff data by sector, India's average applied MFN tariff on this machinery category is 7.87% — among the lowest of any sector we track. For comparison: mobility and transport, 27.4%. Agri-food, 35.3%.

India maintains a substantial tariff wall around cars and food, and leaves the capital goods that build its factories almost open.

There is a coherent argument for exactly this — cheap machinery lowers costs for every downstream manufacturer, and taxing the equipment your industrial policy depends on is self-defeating. Plenty of economists would defend the structure as written.

But it is worth stating the trade-off out loud, because it rarely is: the current settings optimise for cheap imported capital equipment over a domestic capital-goods industry, and a decade of data suggests that is precisely what they delivered. Import substitution in this sector is being asked to happen against a price signal pointing the other way.

What this means if you're allocating

Three things the data supports:

• The gap is mechanical, not digital. Compressors, gears, transmission components, lifting machinery. Anyone pitching "industrial automation import substitution" with a software product is not addressing this import bill, whatever the deck says.
• The few hardware companies here are unusually exposed and unusually under-funded. Ethereal Machines and Favo are building physical machines in a sector where the median cheque is under $1 million. That is either a mispricing or a signal that hardware returns in this category are genuinely difficult — worth diligencing directly rather than assuming either.
• Policy is not currently helping. A 7.9% tariff on the imported alternative is a real headwind for any domestic machine builder, and no amount of startup formation offsets it.

The encouraging fact remains true: this is the one major sector where Indians file most of the patents. The frustrating fact is that it hasn't yet touched what the country actually buys.

Methodology

Import figures are FY2024-25 from India's Trade Intelligence & Analytics Portal at HS8 level, mapped to the robotics and industrial automation sector via HS chapter 84 (machinery and mechanical appliances), with headings 8471 (computers) and 8486 (semiconductor equipment) reassigned to computing and semiconductors respectively. Chapter 84 is broad by construction — it spans everything from boilers to bearings — so this category should be read as "industrial machinery" rather than "robots" specifically.

Eight-year trends are from CEPII BACI V202601 (India as importer, HS6, 2017-2024). For this sector BACI and TIA agree closely on China share (39.0% versus 37.6%), so the trend is treated as reliable. Tariff figures are simple averages of applied MFN rates across HS6 lines in the sector from World Bank WITS TRAINS, 2023 — unweighted by trade volume, so a category with many low-value high-tariff lines can read higher than its trade-weighted exposure.

Patent, startup and funding figures are Deeptech Navigator's registry as of 2026-07-22. The "zero results for compressor" finding reflects the platform's technology taxonomy, which is derived from filed patents; it indicates an absence of Indian patent activity tagged to that term, not proof that no Indian company makes compressors.

In our data

Sources

• Capital goods imports test India's self-reliance push — Policy Circle ↗
• China supplies over 30% of industrial goods; overdependence critical: GTRI — Business Standard ↗
• CEPII BACI V202601 ↗