Guide

Upcoming Semiconductor Plants in India

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

Ten approved plants are still to reach production, and the useful question is not how many were announced but which ones have physical evidence behind them.

How this list is built

Every published list of upcoming semiconductor plants in India mixes plants with foundations poured, plants with nothing but a Cabinet note, and plants that exist only in an executive's remarks at a summit. Sorted together they produce a pipeline that looks enormous and predicts nothing. Sorted apart they produce a forecast a reader can actually use.

This page uses three buckets. Bucket one is under construction, meaning there is dated evidence of physical work from a government document or a named report. Bucket two is approved but without independent construction evidence in this record, which is a statement about what has been published rather than an accusation that nothing is happening. Bucket three is reported only, meaning no approval document exists and the claim rests on reporting or an announcement.

Entries move between buckets when evidence appears, and the date of the move is what this record scores. Nothing here is a forecast of success. A plant in bucket one can still slip, and several will.

Bucket one: under construction

The Tata Electronics wafer fab at Dholera is the most heavily documented construction site in the programme. Approved by the Cabinet on 29 February 2024 at Rs 91,526 crore, with the Fiscal Support Agreement signed on 5 March 2025, it is designed for up to 50,000 wafer starts per month on 300mm wafers with PSMC as technology partner. Civil work stood at about 50 percent as of mid-2026 with cleanroom fit-out under way REPORTED. The main fab foundations were redesigned after soil testing found the ground unsuitable, reported on 28 November 2025 REPORTED, and a special economic zone of 66.16 hectares was notified on or about 16 April 2026 DURABLE. Commercial production is guided to mid-2028 TARGET, and there is no supported first-silicon date.

The Tata Electronics packaging plant at Jagiroad in Assam, approved at Rs 27,120 crore for 48 million units per day using indigenous packaging technologies, is the second-largest approval in the portfolio and gives the group a two-state architecture: front end in Gujarat, back end in Assam. The CG Power venture with Renesas Electronics America and STARS Microelectronics at Sanand, approved at Rs 7,584 crore for about 15.07 million units per day, sits in the same bucket. Both were cleared by the Cabinet in February 2024, and the government's position on 1 April 2026 was that two plants beyond Micron and Kaynes were expected to begin commercial production during the year. It did not say which two. Placing Jagiroad and Sanand in this bucket is therefore an inference from approval age and that expectation, not an identification the government made, and this record flags it as an inference rather than dressing it as a finding.

Construction visibility across this bucket is uneven, and the unevenness is a reporting artefact rather than a signal about the projects. A wafer fab generates geotechnical reports, foundation contractors, cleanroom packages and a special economic zone notification, each of which leaves a document trail that outsiders can read. A packaging plant is a large industrial shed with fit-out inside it, and its progress usually becomes public only when the company reports it or a minister visits. The absence of construction coverage for Jagiroad or Sanand therefore proves much less than the presence of it at Dholera, and this record does not read quiet as trouble.

Bucket two: approved, awaiting visible progress

Seven approved units sit here, and the list is worth reading as a map of where the next production announcements will come from.

• Vama Sundari Investments with Foxconn, Jewar, Uttar Pradesh: Rs 3,706 crore for display driver integrated circuits using gold bump technology with chip probing and die processing, about 20,000 wafers per month and 36 million chips per month, technology from Hon Hai.
• Crystal Matrix, Dholera: approved 5 May 2026 for an integrated compound semiconductor fab and ATMP facility making mini and micro-LED display modules, with gallium nitride epitaxy on six-inch wafers, 72,000 square metres of panels per year DURABLE.
• Suchi Semicon, Surat: approved 5 May 2026 as an outsourced assembly and test plant for discrete semiconductors at 1,033.20 million chips per year for power electronics, analog and industrial customers.
• SiCSem, Odisha: Rs 2,066 crore for a silicon carbide fab with Clas-SiC Wafer Fab at 5,000 wafers per month, plus packaging at 8 million units per month.
• 3D Glass Solutions, Odisha: Rs 1,943 crore for advanced packaging including flip chip ball grid array, radio frequency and antenna-in-package products, glass interposers and three-dimensional heterogeneous integration.
• Advanced System in Package Technologies, Andhra Pradesh: Rs 480 crore with APACT of South Korea for about 96 million units per year.
• Continental Device India, Punjab: Rs 117 crore to expand output of high-power discretes including MOSFETs, IGBTs, Schottky bypass diodes and transistors in silicon and silicon carbide.

The two Gujarat approvals of 5 May 2026 carry a combined investment of around Rs 3,936 crore and a stated employment expectation of 2,230 skilled professionals. The four units in Odisha, Punjab and Andhra Pradesh were cleared during 2025; the government annexure that lists their investment values does not carry their individual approval dates, so this record dates them only to the year, which is as precise as the primary document allows.

Bucket three: reported only

Three categories of claim sit here, and none of them should be counted in a pipeline total.

The first is the modernisation of the Semi-Conductor Laboratory at Mohali. Press reporting during 2026, including The Tribune and Outlook Business, describes a programme of about Rs 4,500 crore, with Tata Semiconductor and Israel's Tower Semiconductor shortlisted for part of the revamp and a decision to keep the laboratory in government hands REPORTED. This record has not seen a primary notification carrying that figure or an award, so it stays reported.

The second is the pipeline that Semicon 2.0 is expected to generate. The Cabinet approved the second phase on 15 July 2026 with an outlay of Rs 1,27,500 crore across six pillars including machines and materials DURABLE, which is a framework rather than a set of projects. Until specific units are approved under it, the correct count of Semicon 2.0 plants is zero, however large the outlay.

The third is the recurring talk of additional fabs, including a second commercial silicon fab in Gujarat. Expansion discussion is normal around a working cluster and occasionally turns out to be accurate, but no approval document supports a second silicon fab as of 25 August 2026, and this record will not list one until a Cabinet release does.

How to read a capacity number in this pipeline

The capacity figures attached to these plants are not comparable with one another, and adding them together produces a number that means nothing. The government annexure quotes Micron in millions of units per week, the Assam plant in millions of units per day, Kaynes in millions of chips per day, Suchi Semicon in millions of chips per year, the Andhra Pradesh unit in millions of units per year, the Jewar unit in both wafers per month and chips per month, Dholera in wafer starts per month, and Crystal Matrix in square metres of display panel per year. That is eight plants and nine different ways of stating output.

Each denominator is correct for the process it describes. A packaging plant counts finished devices because devices are what it ships, and a device can be a tiny discrete or a large processor package, so a unit count says nothing about value or silicon area. A wafer fab counts wafer starts because everything downstream depends on how much silicon enters the line, and a single 300mm wafer can yield anywhere from a few hundred to several thousand dies depending on die size. A display plant counts area because area is what a panel buyer purchases.

Three rules follow. Never compare a wafer number with a unit number, because they describe different stages of the same chip. Never sum unit counts across plants making different products, because the total will be dominated by whichever plant makes the physically smallest device. And read every figure as a design or approved capacity rather than an output forecast, because plants ramp over quarters and rarely run at nameplate in their early years. Applied to this pipeline, those rules leave one honest headline: twelve approved units and about Rs 1.64 lakh crore of approved investment, which is why this record tracks money and dates rather than an aggregated chip count.

What has already left this list

Two plants graduated out of the upcoming category, and noting that is as important as tracking what remains. Micron's assembly and test plant at Sanand, approved at Rs 22,516 crore for DRAM and NAND products at around 14 million units per week, and Kaynes Technology India's plant in Gujarat, approved at Rs 3,307 crore for wire bond interconnect and substrate-based packages at over 6.33 million chips per day, have both commenced commercial production DURABLE.

The intervals are the useful part. Micron's approval dates to 2023 and Kaynes's to September 2024, and both had commenced commercial production by the government's statement of 1 April 2026, which is roughly nineteen months from approval in one case and under three years in the other. That is the benchmark to read the rest of the list against: a packaging plant approved in 2026 that begins production in 2028 is running to a normal schedule rather than slipping. A wafer fab is a longer proposition by design, which is why the Dholera guidance sits at mid-2028 against a February 2024 approval, an interval of more than four years.

The delivery record this list is measured against

India's semiconductor pipeline has a history of announcements that did not become plants, and the discipline of this page comes from that history. The clearest documented case is the Vedanta and Foxconn joint venture, which signed with Gujarat in September 2022 and collapsed when Foxconn withdrew on 10 July 2023 DURABLE. Nothing was built, and the widely repeated project value was never committed capital.

That episode carries a second lesson about how slowly a collapse becomes visible. Roughly ten months separated the state signing from the exit, and for most of that period the venture appeared on published pipelines as a live project with a large headline value attached to it. A list that removes an entry only when a withdrawal is announced will overstate the pipeline for as long as the silence lasts, which is why the buckets on this page are built on evidence of progress rather than on the absence of bad news.

The structural difference in the current wave is not enthusiasm, it is instruments. Approvals now come with a Fiscal Support Agreement that converts an approval into an enforceable commitment, disbursal is milestone-linked so money follows verified progress, and the technology partner is named in the approval document rather than left to a future negotiation. Those three features are why this wave has produced operating plants and the earlier ones did not. They are not a guarantee, and this record will score any slip against the dates above rather than quietly restating them.

The gates an announcement has to pass

When a new plant is announced, five questions place it in a bucket within minutes. Is there a Cabinet approval naming the company, the state and the investment? Is there a signed fiscal support agreement or only an approval? Is there land, with a notification or allotment reference? Is there a named technology partner, or does the project still need to find one? And has any third party observed physical work, a foundation, a contractor mobilisation, a cleanroom package, an equipment order?

A project that answers yes to the first four but no to the fifth is real and early. A project that answers no to all five is a press release. The difference between them is not a matter of opinion, and it is why the date scoreboard and the official progress record are kept separately from the announcement flow.

Dates that are scoreable in the next two years

Four dates can be checked rather than argued about. The government stated on 1 April 2026 that two further plants were expected to start commercial production during the year, which is scoreable by 31 December 2026 TARGET. Construction evidence at Crystal Matrix and Suchi Semicon, approved on 5 May 2026, should appear during 2026 and 2027 if those projects are moving. The first project approvals under Semicon 2.0 would confirm that the equipment and materials pillar is operational rather than aspirational. And commercial production at Dholera is guided to mid-2028, with cleanroom completion and tool move-in as the intermediate markers that would either support or undermine that guidance long before it arrives.

Two negative findings are worth restating because they will keep being contradicted in coverage. No semiconductor plant below 28nm is approved in India, and no memory fab or display panel fab is approved either. The catalyst calendar carries the dated watch list, and every entry above moves buckets only when a document says it should.