Guide

Semiconductor Fabs in India

Published 25 August 2026. Facts verified to 25 August 2026 unless dated otherwise.

One fab runs today, one is being built, two compound fabs are approved, and nine plants that headlines call fabs are not fabs at all. This page separates them and shows the working.

What counts as a fab, and why the definition decides the answer

A fab is a wafer fabrication plant. It takes blank silicon wafers and builds transistors and interconnect into them layer by layer through lithography, etch, deposition, implantation, planarisation and cleaning, repeated across hundreds of steps, in an environment engineered to remove particles, vibration and electrical noise that would be invisible anywhere else. When the wafer leaves, the circuits exist. Everything that happens afterwards, separating the dies, mounting them, wiring or bumping them, sealing them and testing them, is packaging and test, and it happens in a different kind of building at a fraction of the capital cost.

That single boundary settles almost every dispute about how many fabs India has, because Indian coverage routinely applies the word fab to plants that will never fabricate a wafer. The government's own categorisation, published in the Union Cabinet release of 15 July 2026, sorts the twelve approved units under the India Semiconductor Mission into one silicon fab, one silicon carbide fab, one integrated gallium nitride micro-LED display fab and nine packaging units DURABLE. That is three fabs and nine packaging plants, it is the government's own count, and it matches the definition above.

One more plant sits outside that portfolio entirely, because it predates the mission and is a government laboratory rather than an approved commercial project. It is also, at the time of writing, the only place in India where wafers are actually being fabricated. The fab desk explainer covers the engineering in detail; this page is the register.

The fab that already runs: SCL Mohali

The Semi-Conductor Laboratory at Mohali fabricates wafers today and has done so for years, which is why the common claim that India has never made a chip is false as stated. The government's Lok Sabha reply of 1 April 2026 records that of 211 chips taped out by 75 Indian academic institutions using state-provided design tools, 149 were fabricated at 180nm at SCL Mohali, with 62 sent to overseas foundries. That is a working line with real output, used heavily by the country's universities.

Its scale and purpose are the point. Press reporting during 2026, including in The Tribune, describes an eight-inch line running 180nm CMOS at around 600 wafer starts per month, serving space, defence and medical applications, alongside a modernisation programme of about Rs 4,500 crore and a decision to keep the facility in government hands rather than privatise it REPORTED. This record has not seen a primary notification carrying that modernisation figure, so it stays at reported tier until one appears. Outlook Business has reported that Tata Semiconductor and Israel's Tower Semiconductor were shortlisted for part of the revamp.

Put SCL's numbers next to Dholera's and the difference between a strategic line and a commercial fab becomes concrete: about 600 wafer starts per month against a design capacity of up to 50,000, on eight-inch wafers rather than 300mm, at 180nm rather than a stated range of 28 to 110nm. SCL is not a small version of the commercial wave. It is a different instrument, built for sovereign requirements where supply security matters more than cost per die.

The commercial silicon fab under construction: Dholera

The Tata Electronics fab at Dholera is the only commercial silicon wafer fab approved in India. The Union Cabinet cleared it on 29 February 2024 at Rs 91,526 crore, the Fiscal Support Agreement committing fifty percent central support was signed on 5 March 2025 DURABLE, and the plant is designed for up to 50,000 wafer starts per month on 300mm wafers with PSMC of Taiwan as technology partner. A special economic zone for Tata Semiconductor Manufacturing Private Limited covering 66.16 hectares was notified on or about 16 April 2026 DURABLE.

On progress, the sourced position is that civil work stood at about 50 percent as of mid-2026 with cleanroom fit-out under way REPORTED, after the main fab foundations were redesigned following soil testing that found the ground unsuitable, reported by The Economic Times on 28 November 2025 REPORTED. Commercial production is guided to mid-2028 by the union minister, speaking on 17 July 2026 TARGET. No primary source publishes a first-silicon date, and this record does not print one.

The node is genuinely unresolved rather than merely unclear. Tata's stated range is 28 to 110nm and the Tata Sons chairman's FY25 letter described starting at 28nm, while Bloomberg on 17 July 2026 and TrendForce on 20 July 2026 reported that the plant will open mostly at 90nm. The ASML memorandum announced by both companies on 16 May 2026 names 28, 40, 55, 90 and 110nm, which is consistent with either version. The node question is tracked as an open contradiction because both statements remain on the record and neither party has withdrawn one.

The two compound fabs

Compound semiconductor fabs use materials other than plain silicon, and India has two approved. SiCSem Private Limited is approved for a silicon carbide fab in Odisha at Rs 2,066 crore, in technology partnership with Clas-SiC Wafer Fab, with a stated fab capacity of 5,000 wafers per month and packaging capacity of 8 million units per month, per the annexure of 1 April 2026. Silicon carbide devices matter for power electronics, where they switch at higher voltages and temperatures than silicon can manage efficiently, which is why electric vehicles and grid equipment pull that demand.

Crystal Matrix Limited was approved by the Cabinet on 5 May 2026 for an integrated compound semiconductor fabrication and ATMP facility at Dholera making mini and micro-LED display modules, including gallium nitride foundry services with epitaxy on six-inch wafers, at a stated annual capacity of 72,000 square metres of display panels and 24,000 sets of red, green and blue epitaxy wafers DURABLE. The Cabinet described it as the country's first commercial mini and micro-LED display facility based on gallium nitride technology.

Both are fabs in the strict sense: material goes in, devices are built into it. Both are also small next to Dholera in capital terms, which is normal for compound work, where a viable plant costs a fraction of a leading silicon fab and serves narrower markets.

The nine plants that are not fabs

The remaining nine approved units assemble, package and test chips that were fabricated somewhere else, and calling any of them a fab misstates what they do. They are Tata Electronics at Jagiroad in Assam at Rs 27,120 crore, Micron at Sanand at Rs 22,516 crore, the CG Power venture with Renesas and STARS at Sanand at Rs 7,584 crore, Vama Sundari Investments with Foxconn at Jewar at Rs 3,706 crore, Kaynes Technology India at Sanand at Rs 3,307 crore, Suchi Semicon at Surat, 3D Glass Solutions in Odisha at Rs 1,943 crore, Advanced System in Package Technologies in Andhra Pradesh at Rs 480 crore, and the Continental Device India expansion in Punjab at Rs 117 crore.

This is not a demotion. Packaging employs large workforces across the global industry, it is where several advanced technologies now live, and 3D Glass Solutions in particular is approved for flip chip ball grid array assembly, radio frequency and antenna-in-package products, glass interposers and three-dimensional heterogeneous integration, which is genuinely frontier work. The point is only that these plants cannot make a wafer, so they cannot substitute for one. The packaging and test explainer sets out what they physically do.

The confusion has a practical cost. Readers who believe India has ten fabs conclude the country has solved a problem it has not solved, and readers who then hear that India has one fab conclude the programme failed. Both readings come from the same missing definition.

The Jewar unit, and why it is the hardest case

The unit approved for Jewar in Uttar Pradesh is where the fab boundary is most often crossed in print. Its approved scope is display driver integrated circuits using gold bump technology, along with chip probing and die processing services, at about 20,000 wafers per month and 36 million chips per month, with technology from Hon Hai of Taiwan and Foxconn India as joint venture partner. The wafer figure is what causes the trouble: a plant that measures its throughput in wafers per month sounds like a fab.

It is not one. Gold bumping deposits contact bumps onto wafers that arrive already fabricated, and probing tests dies while they are still on the wafer. Both are back-end operations performed on incoming wafers, which is why the government's own categorisation counts this unit among the nine packaging units rather than among the three fabs DURABLE. Display drivers are chips, so this plant is part of India's display supply chain, but it does not fabricate the driver, and no display panel fab has been approved in India.

The fab claims that died

Any list of Indian fabs that has been circulating for more than three years probably still carries the Vedanta and Foxconn venture. Foxconn withdrew from it on 10 July 2023 DURABLE. The frequently quoted headline value was a joint venture project figure that never became committed capital, no plant was built, and the venture belongs to the history of Indian fab attempts rather than to its pipeline.

The broader lesson is procedural. Announcements are not approvals, approvals are not contracts, and contracts are not construction. Between a memorandum signed at an investment summit and a working fab lie a formal application into the incentive scheme, a technology licensing agreement with a process owner, a final investment decision by every shareholder, a fiscal support agreement, land, foundations, structure, cleanroom, hookup, tool move-in and qualification. A venture can pass the early gates loudly and fail a later one quietly, which is exactly what happened in 2023.

Why the list is this short

Four fabs for a country of India's size looks thin until the economics are laid out. A modern 300mm plant is among the most expensive industrial buildings anywhere, and Rs 91,526 crore is what a mature-node fab of this capacity was approved at. Most of that money is process equipment rather than structure, which is standard for wafer fabs, the tools come from a small supplier base with lead times measured in quarters, and the plant earns nothing until yield climbs to a commercially viable level well after the first wafer runs.

Then there is demand. A fab is only viable if someone qualifies its output, and qualification for automotive or industrial customers is a multi-year process involving reliability testing the customer controls. That is why fabs cluster in places with existing customers, and why a new entrant needs both a technology partner and a subsidy to be financeable at all. Choosing mature nodes is the strategy that makes the arithmetic work, not a consolation prize.

Time compounds the capital problem. Process tools are ordered many months before they are needed, from a supplier base narrow enough that a global demand surge lengthens everyone's queue at once, and they cannot be installed until the cleanroom that houses them is sealed, filtered, levelled and vibration-controlled to specification. That sequence is unforgiving: a delay in the building pushes tool move-in, tool move-in gates qualification, and qualification gates the first revenue. Nothing in it can be compressed by spending more, which is the single most common misreading of fab schedules by people used to other kinds of construction.

The result is that fabs are built by a small number of organisations with the balance sheet to wait, and countries acquire them one at a time. Four in India, three of them approved within the past thirty months, is quick work in an industry that adds plants one at a time.

What would add a name to this list

Three developments would change the register. A Cabinet approval for a second commercial silicon fab, in Gujarat or elsewhere, would take the count from four to five, and expansion talk has circulated for some time without an approval document to support it. An approval for a memory fab or a display panel fab would open a category India does not have at all today. And a modernisation contract at SCL Mohali that converts the reported programme into a signed award would change what the existing fab can do rather than how many exist.

Until one of those is documented, the list is: one operating, one building, two approved. Anything longer is counting packaging plants as fabs, and the recurring Dholera and Sanand mix-up is a version of the same confusion applied to geography instead of to process.