Guide

India's semiconductor history: the dates that disagree, the fabs that were never built, and what changed after 2021

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

Most accounts of India's chip history are confident about dates that the underlying record does not settle, so this page prints the disagreements instead of smoothing them.

What this page establishes, and the rule it follows

India's semiconductor history is retold more often than it is checked. The same handful of anecdotes circulate with different years attached, a fire is described alternately as an accident and as sabotage, and two rounds of failed fab approvals get compressed into one. Because the value of this record is that its numbers survive scrutiny, this page follows a narrower rule than most histories do: a date or a figure appears here only if a named source with a date carries it, and where two credible accounts disagree, both appear and the disagreement is named as a disagreement.

That rule costs the page some tidiness. Several figures that circulate widely, including the combined investment totals attached to the 2013 and 2014 fab approvals, are not printed below, because this desk has not examined the underlying decisions and has found the reported totals quoted in inconsistent forms. An honest blank is more useful than a plausible number, particularly on a page whose whole purpose is to explain how India's chip ambitions failed for forty years.

The argument the page makes is structural rather than patriotic. India did not lack engineers, ambition or announcements in any decade since the 1960s. It lacked a financing instrument that survived a change of government, a technology partner with a running process, and a site with power and water already in the ground. What separates the current wave from its predecessors is not enthusiasm. It is that those three things exist at once, and can be checked.

The 1960s and 1970s: a base that genuinely existed

The starting position was better than the later record suggests. In its June 2023 analysis of India's semiconductor past, written by Trisha Ray and published on 22 June 2023, the Observer Research Foundation records that a handful of Indian companies were producing germanium semiconductors in the 1960s, that Fairchild Semiconductor considered India for its first Asian unit, and that Bharat Electronics, the defence ministry undertaking, acquired both germanium and silicon device technology. The same account notes Metkem Silicon establishing polysilicon facilities at Mettur in Tamil Nadu with support from Bharat Electronics.

Those are small facts with a large implication. Device manufacture in India predates the industry's move to advanced nodes, and it predates the East Asian foundry model entirely. The country was not a latecomer in the 1960s. It became one, and the interesting historical question is what mechanism produced that gap, not whether the gap exists.

The mechanism visible in the record is a policy environment built for import substitution and licensing rather than for scale manufacturing. Semiconductor plants are capital instruments before they are engineering projects, and they require a decade of stable, unglamorous capital expenditure to pay back. An industrial policy that rationed imports and allocated licences could produce a plant. It could not produce the second, third and fourth reinvestment cycles that keep a plant near the technology frontier.

Semiconductor Complex Limited: the founding dates that disagree, and the fire that ended it

The centrepiece of the early record is the public-sector plant at Mohali, near Chandigarh, originally incorporated as Semiconductor Complex Limited. Its founding is where the retold accounts first diverge, and the divergence is instructive rather than trivial.

One line of accounts, carried by the Wikipedia entry on the laboratory and by secondary Indian coverage that follows it, places the Union Cabinet's decision to create the enterprise in 1976, with production beginning in 1984 on technology from American Microsystems Inc, starting at five-micron CMOS. Those are secondary sources rather than primary instruments, and this record labels them as such. The Observer Research Foundation's 2023 account instead describes the government establishing the company in 1984, made possible through licensing agreements with Hitachi, AMI and Rockwell. The Tribune, in its retrospective of 8 October 2022, gives 1984 as the establishment year.

These accounts are reconcilable in the ordinary way that Indian public-sector projects are: a Cabinet sanction and a commissioning are different events separated by most of a decade, and different writers pick different ones as the founding. This record therefore uses both dates with their meanings attached, sanction in 1976 and production from 1984, rather than asserting a single founding year. The licensing roster is the second point of disagreement, since the accounts differ over whether one partner or three supplied the technology, and no source examined here settles it. Both are stated as reported.

The technology position at that point was not marginal. The Observer Research Foundation's assessment is blunt about it: at the end of the 1980s India stood about two years behind the latest semiconductor manufacturing technologies. A two-year gap is a recoverable gap. It is the closest India has ever been to the frontier, and it closed rather than widened only because of what happened next.

On 7 February 1989 a fire tore through the Mohali plant. This is the single most mythologised event in Indian semiconductor history, and it is worth separating what is documented from what is atmosphere.

What the sources agree on: the date, and that the damage fell heavily on imported equipment, with the loss reported at about Rs 60 crore. The Tribune's account of 8 October 2022 records that the plant restarted in 1997, eight years after the fire. The Observer Research Foundation's account says that the revival that followed produced only a limited number of chips for the space programme, which is consistent with the plant's later trajectory as a strategic rather than commercial line.

What the sources do not agree on is the cause, and this is where retelling has done the most damage. Popular coverage, including headlines that describe the fire as mysterious, keeps a sabotage reading alive. The Tribune's reporting records something more prosaic and better evidenced: the probe ruled out external sabotage and pointed to an accident or a short circuit, and it found serious lapses in the plant's firefighting arrangements. Reporting also records that Intelligence Bureau officials visited the plant to assess the causes and that nothing conclusive emerged publicly.

The honest statement is therefore narrow. A fire of undetermined proximate cause, on a date nobody disputes, destroyed a large share of an irreplaceable imported toolset, and the investigation as publicly reported did not support the sabotage story that has attached itself to the event. This desk prints no cause because the record supports none.

The more useful observation is about consequence rather than cause. What made the fire decisive was not the flame. It was that a plant dependent on imported tools, in a foreign-exchange constrained economy, under an export-control regime that limited what could be bought and how fast, had no route to a rapid retool. Eight years of downtime in an industry that halves feature sizes on a schedule is not a setback. It is an exit. The two-year gap of 1989 was structurally unrecoverable by 1997.

The 2007 round: a policy, a licence, and no plant

India's next serious attempt came two decades later. The Observer Research Foundation's account records the 2007 semiconductor policy as targeting Rs 24,000 crore of investment over three years alongside the establishment of three fabrication units. The best-known proposal of that round was SemIndia, a consortium of Indian technologists that signed a licensing agreement with AMD to use AMD's technology for a wafer fab, with Hyderabad's Fab City as the intended site.

No fab was built. The causes named in that account are worth listing precisely, because each one recurs: delay in the passage of the policy itself, minimum investment requirements strict enough to deter entrants, AMD's separation of its manufacturing arm into what became GlobalFoundries, and fundraising and production delays on the proposer's side. Every one of those is a financing or partner-stability failure rather than an engineering one.

The structural read is that 2007 offered a policy without an instrument. A stated intention to support investment is not the same as a defined share of capital expenditure, disbursed against milestones, under a signed agreement. Without that, a proposer must raise the full cost against an incentive whose value cannot be modelled, which is a financing problem no amount of ministerial enthusiasm solves.

The 2013 and 2014 approvals: two consortia, two exits

The third attempt is the one most often cited as evidence that Indian fab approvals mean nothing, and it deserves careful handling because its dates are genuinely muddled in circulation.

Two consortia received letters of intent. The first was Hindustan Semiconductor Manufacturing Corporation, partnered with STMicroelectronics and Malaysia's Silterra. The second was Jaiprakash Associates, partnered with IBM and Tower Semiconductor. The Observer Research Foundation's June 2023 account carries both pairings. Reporting at the time placed the first consortium's proposed plant in Gujarat and the second's in Uttar Pradesh, along the Yamuna Expressway near Delhi REPORTED.

The date is where accounts split. The round is universally referred to as the 2013 fab approvals. Later reporting on the collapse of the Jaiprakash project describes that plant as one of the two approved by the Cabinet in February 2014. The likeliest reconciliation is that a Cabinet clearance and the subsequent issue of letters of intent were separate steps in different years, which is how most Indian project approvals work. This desk has not examined the underlying Cabinet decision, so it prints both dates and declines to pick the tidier one REPORTED.

The outcomes are firmer than the start dates. Per the same Observer Research Foundation account, Jaiprakash Associates withdrew its proposal in 2016, citing capital outlay, and the government cancelled the permit held by Hindustan Semiconductor Manufacturing Corporation in 2019 after years of paperwork delays on the company's side. Measured from the letters of intent, that is two to three years to the first withdrawal and five to six years to the second cancellation, during which neither plant was built.

Combined investment figures for the round circulate widely and inconsistently. They are omitted here for the reason given at the top of this page. What the round demonstrates without any figure at all is that an approval is a beginning, that a consortium whose lead partner is a construction and infrastructure group rather than an electronics manufacturer carries a specific fragility, and that a permit unaccompanied by land, utilities and a running process can sit unexecuted for five years before anyone withdraws it.

What actually went wrong, across three decades

Read together, the failures share four features rather than one.

• No durable financing instrument. Until 2021 there was no scheme that told a builder what share of its capital expenditure the state would carry, in what form, against what milestones. Every proposal therefore had to be financed as though the incentive were zero.
• No technology partner with a running line. A licence to use a technology is not the same as a partner transferring a qualified process with its recipes, its yield history and its design kit. The 2007 and 2013 attempts largely bought the former.
• No serviced site. Fabs need very large volumes of power, ultrapure water feed, bulk gases and effluent handling before a single tool arrives. Choosing a location and acquiring the land is the easy part of that sentence.
• No back end. A fab without domestic assembly and test ships its wafers abroad to be finished, which weakens the industrial case for building it at home in the first place.

The mechanics of how the current scheme addresses the first of those are set out at how fab incentives work in India, and the programme itself at the India Semiconductor Mission explained.

Why the wave that began in 2021 is built differently

The Semicon India programme was approved by the Union Cabinet in December 2021 with an incentive framework of Rs 76,000 crore, offering fiscal support of up to 50 percent for silicon fabs, compound semiconductor facilities, assembly and testing units and chip design, per the Press Information Bureau backgrounder of 7 February 2026 DURABLE. That single design change fixes the financing defect of every earlier round, because a builder can now model the state's contribution before committing.

The second change is the partner. The Dholera project pairs Tata Electronics with Taiwan's Powerchip Semiconductor Manufacturing Corporation, a foundry with a running mature-node process rather than a licensor of documents, an arrangement examined at the Tata and PSMC partnership. The third is sequencing: India approved back-end assembly and test plants alongside the fab, so packaging capacity arrives before front-end output rather than after it.

The fourth is that the money has moved into concrete. The fabrication support agreement of 5 March 2025 cites Rs 91,526 crore for the Dholera project, with central support covering 50 percent of eligible cost, and the plant is being built on a site with trunk infrastructure already installed DURABLE. Civil work stood at about 50 percent as of mid-2026, with cleanroom fit-out under way REPORTED. Commercial production is guided to mid-2028, per the Union minister's statement of 17 July 2026 TARGET. No supported first-silicon date exists, and any December 2026 claim in circulation is unsupported. The dated build record is kept at the Dholera fab construction timeline.

This page would be dishonest if it treated the present round as failure-free. It is not. The most cited Indian fab proposal of the last five years, the Vedanta and Foxconn joint venture, is dead: Foxconn withdrew from it on 10 July 2023 DURABLE, no plant was built, and the widely quoted project value was never committed capital. Aggregate investment tallies that still carry it are out of date by three years, a habit examined at the Vedanta file.

The lesson is procedural and it applies to the current wave exactly as it applied to 2013. Announcements enter cumulative totals and are never retired from them, so a running sum records intent and failure at the same face value. This is why every entry on this record carries a tier, and why a memorandum is never counted as a plant.

Where the history stands after 15 July 2026

The current position is the strongest in the sequence, and it is worth stating with its dates attached. As of 15 July 2026 the India Semiconductor Mission's published position records twelve approved semiconductor manufacturing projects, of which three had entered commercial production DURABLE. Aggregating the dated Cabinet decisions behind those approvals puts stated investment above Rs 1.64 lakh crore; the Mission's own project pages list projects individually without printing a cumulative total, so this record treats that figure as an aggregation rather than as a published one. Micron's Sanand plant was inaugurated on 28 February 2026, Kaynes Semicon's Sanand plant was recorded in commercial operation from 31 March 2026, and the CG Power joint venture with Renesas was reported to have begun commercial production on 4 July 2026 REPORTED.

On the same day, 15 July 2026, the Union Cabinet approved the next phase of the Semicon India programme with a total budget outlay of Rs 1,27,500 crore, recorded in Press Information Bureau release 2284784 and corroborated on the Prime Minister's Office site DURABLE. That is the largest commitment in the history described on this page, and it arrives at a moment when the programme has already converted approvals into three operating plants and one fab in construction rather than into another round of lapsed letters of intent.

Set against 1989, 2007 and 2013, the difference is not the size of the number. Every previous round had a large number. The difference is that this one is attached to a signed agreement, a partner with a running process, a serviced site, and plants that customers can already buy from. Which of India's approved plants are genuinely fabs, as opposed to packaging units, is sorted at semiconductor fabs in India, and the one line that has been fabricating wafers throughout this entire history is documented at the SCL Mohali record.

What would falsify the claim that this time is different

A history page earns its keep by naming the evidence that would overturn its own reading. Four things would do it here. A slip in the mid-2028 commercial guidance without a company-issued replacement schedule would move the Dholera project into the pattern this page describes. A withdrawal by the technology partner would remove the single largest structural difference from 2013. A stall in disbursement against the fabrication support agreement would show that the financing instrument is weaker in practice than on paper. And an approved back-end plant failing to reach commercial operation on its stated schedule would suggest that the sequencing advantage is smaller than claimed.

None of those has happened as of 25 August 2026. Three have not even been credibly reported. That is the honest state of a forty-year record whose previous chapters each ended with a permit and no plant, and it is why this desk scores the current round on dated deliverables rather than on the size of the announcement.