India's Semiconductor Mission: What ₹1.64 Lakh Crore Has Actually Built

06 Jun 2026 · 12 min read

In December 2021, the Union Cabinet approved the India Semiconductor Mission (ISM) with a budget of ₹76,000 crore. By 2024, that number had grown to ₹1.64 lakh crore across three approved fabrication plants, two assembly-and-test facilities, and a dozen design-linked incentive applications. Every Finance Minister since has called it a strategic imperative. Every Economic Survey has cited it as evidence of India's manufacturing pivot.

Four and a half years later, something has been built. The question is whether it is the thing the Mission promised.

One facility is operational. One is under construction with timelines that keep slipping. One has broken ground but has no commercial production date. The talent pipeline that was supposed to fill these fabs remains a decimal point of what is needed. And the real cost — in water, power, and the opportunity cost of subsidising a single industry — is only now becoming visible.

This is the ground truth of India's chip bet.

The one that works

The Micron semiconductor assembly and test facility in Sanand, Gujarat, is the only ISM project that has delivered on its promise. PM Modi inaugurated it on 28 February 2026 — approximately three to four months behind the original Q4 2025 target, but operational nonetheless.

The facility is an ATMP (Assembly, Testing, Marking, and Packaging) plant. It does not fabricate chips from silicon wafers. It takes finished wafers — cut, tested, and shipped from fabrication plants in Taiwan, Korea, and Japan — and assembles them into the packages that go onto circuit boards. The value addition is real but modest: packaging accounts for roughly 10-15% of a chip's total cost.

Micron's investment is $2.75 billion (₹22,940 crore), of which the Indian government is subsidising 50% under the ISM. The facility produces memory chips — NAND and DRAM — for global markets, not for Indian consumption specifically. Its output goes back into Micron's global supply chain.

This is not a failure. It is the most realistic outcome of India's semiconductor strategy: a global memory company built a packaging plant in Gujarat because the government paid half the cost and the labour was cheaper than China. The plant employs approximately 5,000 people, mostly at the technician and operator level.

The gap between this and what the Mission promised is the gap between packaging and fabrication.

The one that keeps slipping

The Tata-Powerchip Semiconductor Manufacturing Corporation (PSMC) joint venture in Dholera, Gujarat, is the project that carries the mission's ambition. It is a true fab — a wafer fabrication plant that will take raw silicon and turn it into finished chips. The investment is ₹91,526 crore, making it the single largest private-sector industrial investment in Gujarat's history.

The original timeline called for construction to begin in 2024, equipment installation in 2025, and first silicon (the industry term for the first functional chip produced) by mid-2026. As of June 2026, the factory shell is rising but the clean room is not complete. Industry sources tracking the project now expect first silicon in the first half of 2027 — a slippage of six to twelve months.

The reasons are structural, not anecdotal:

• Equipment import delays. The lithography machines, etch tools, and deposition chambers that populate a fab are manufactured by a handful of suppliers — ASML (Netherlands), Applied Materials (US), Tokyo Electron (Japan). Each machine has a lead time of 9-18 months. India placed its orders late in the cycle, competing with fabs under construction in the US, Japan, and Germany.
• Water. A 28nm fab consumes 10-15 million litres of ultrapure water per day. Dholera is in a semi-arid region of Gujarat. The water is being sourced from the Narmada canal system, which already serves the Ahmedabad metro area during summer months. A long-term water supply agreement is in place. Its enforceability during a drought year is untested.
• Power. Gujarat's industrial power demand grew 12% in FY 2025-26. Night-time power shortages have already been reported in the Sanand-Bavla industrial belt. A fab cannot tolerate power fluctuations — a single microsecond interruption can ruin a batch of wafers worth ₹2-3 crore. Dholera has a dedicated substation. The grid feeding it is the same grid that services every other factory in the zone.
• Visa delays. Taiwanese engineers from PSMC were required on-site for equipment installation and calibration. Industry sources report that visa processing for this category was slower than anticipated, though no project official has confirmed a specific delay directly attributable to this.

The land for the Dholera fab was valued at zero by lenders. Banks financing the project assessed the collateral value of the land at nil — because a fab built on it has no alternative use. If the project fails, the building cannot become a warehouse or an office park. The clean room specifications, vibration controls, and power infrastructure make it unusable for any other industrial purpose. Banks lent on Tata's credit rating, not on the project's asset value. The terms were, by multiple accounts, unconventional.

The one that's invisible

The Tata-Assam ATMP facility in Jagiroad, 60 kilometres from Guwahati, is the third approved project. ₹27,120 crore. Capacity: 48 million chips per day. It will package chips for automotive, electric vehicle, and consumer electronics applications.

No confirmed commercial production date has been publicly announced. The project broke ground in 2024. Construction is underway. But among the three projects, it has the least visibility, the fewest progress updates, and the longest supply chain — Assam has no existing semiconductor ecosystem, no chemical suppliers, no specialty gas distributors, and no trained fab workforce within 1,000 kilometres.

The Assam facility illustrates a problem that none of the press releases address: a fab is not a building. It is the centre of an industrial ecosystem that includes chemical plants, gas refiners, precision engineering workshops, and logistics providers that understand contamination-controlled transport. India is building the centre without the ecosystem.

The talent gap

The ISM's official talent target is to create 85,000 direct jobs across the approved projects. The actual number of chip design engineers in India today is estimated at 20,000-30,000 — concentrated in Bengaluru's EDA (Electronic Design Automation) and VLSI design houses that serve as cost centres for global chip companies.

The gap: India needs approximately 300,000 trained semiconductor professionals over the next five to seven years to staff the fabs, ATMPs, and design centres the Mission envisions. That is a ten-fold increase from current capacity.

The government's FY 2026-27 budget allocates for approximately 200 chip design hires and 3,000 ATMP-level technician roles across all ISM-linked training programmes. At that rate, filling the 300,000 gap would take 75 years.

The 62,000 trainees enrolled in NIELIT's semiconductor courses add a layer of qualification that is difficult to assess. NIELIT's curriculum covers foundational electronics. Whether it produces graduates ready to work in a Class 1 clean room — where the air filtration standard requires fewer than 10 particles per cubic foot — is a different question. The industry standard for fab technician training is 6-12 months of hands-on apprenticeship. No Indian institute currently offers this.

The investment math

The government's official committed figure is ₹1.64 lakh crore across all approved ISM projects. The amount actually spent — money that has left government accounts and reached project sites — is estimated at less than ₹35,000 crore.

The structure of the subsidy explains the gap. Under the ISM, the government reimburses capital expenditure on a pari-passu basis — it matches the company's spending. Until the company spends its own money, the government does not release its share. The ₹1.64 lakh crore figure represents the total project cost, of which the government's share is roughly 50% in most cases. The actual disbursement is tied to construction milestones.

This means the headline number and the ground reality diverge by a factor of five. ₹1.64 lakh crore is the aspiration. ₹35,000 crore is the commitment that has materialised. The difference is the measure of how much construction is actually happening.

The geopolitics

The entire India semiconductor strategy rests on a single assumption: that Taiwan's chip industry will continue to function as the world's foundry. Taiwan produces over 90% of the world's advanced chips (7nm and below). TSMC alone accounts for more than half of global foundry revenue. The PLA's 2027 deadline for "comprehensive preparation" for Taiwan unification is the clock that every semiconductor-consuming nation is watching.

India is not competing for the advanced node race. The Dholera fab is targeting 28nm — a node that was state-of-the-art in 2011. It is not building 3nm or 5nm chips. It is building the chips that go into power management ICs, automotive microcontrollers, and industrial sensors.

The argument for this is pragmatic: the world needs more mature-node capacity because automotive, defence, and industrial customers do not need the latest process node. They need reliable, long-lifecycle chips on proven nodes. India is positioning itself as the second source for 28nm-90nm production — the "if Taiwan is inaccessible" hedge.

The counter-argument is that by the time Dholera reaches volume production (2028 at the earliest), the mature-node market may have shifted. China is already building 28nm capacity at scale across SMIC and Hua Hong. ASEAN countries — Vietnam, Malaysia, Thailand — are expanding their ATMP capacity faster than India. The window is real. It is also narrow.

What India actually consumes

India's total semiconductor consumption was $52 billion in FY 2024-25, according to the India Electronics and Semiconductor Association (IESA). Of this, roughly 90-95% was imported, at a cost of $30.3 billion for the fiscal year. The cumulative import bill from FY 2017 to FY 2025 stands at $150 billion — a sum larger than India's entire defence budget for any single year in that period.

The consumption is not evenly distributed across chip types. According to IESA's market report, mobile handsets, IT, and industrial applications collectively contribute nearly 70% of India's semiconductor revenue. Smartphones alone drive roughly 40% of consumption. Consumer electronics (laptops, tablets) account for about 18%. Automotive, industrial power electronics, and telecom infrastructure make up the remainder.

On process nodes, the available data is less precise than the headlines suggest. No single authoritative source — IESA, NITI Aayog, or Counterpoint Research — publishes a verified percentage breakdown of India's demand by node. Secondary compilations estimate that roughly 55% of demand sits at 28nm and below — the node range that goes into smartphone processors, advanced SoCs, and memory. The remaining 45% spans mature nodes from 45nm to 180nm and above — the domain of automotive microcontrollers, power management ICs, industrial sensors, defence electronics, and solar inverters.

This matters because India's ISM strategy explicitly targets the mature-node segment. The Dholera fab is designed for 28nm to 110nm. It will not produce a single chip for India's smartphone market — those require 7nm, 5nm, and now 3nm processors that no ISM project addresses. India will remain 100% import-dependent for every chip that goes into an iPhone, a Pixel, or a flagship Qualcomm-powered Android device for the foreseeable future.

What Dholera can produce are the chips India's automotive and industrial sectors actually need: microcontrollers for ABS and engine control units, power management ICs for EV battery management systems, and sensor controllers for industrial automation. These are not glamorous chips. They are the ones India imports in high volume and low margin.

How this scales

If the ISM executes exactly as planned — Dholera reaches full capacity, the Assam ATMP hits 48 million chips per day, and Micron's Sanand plant operates at steady state — what does that actually cover?

The Dholera fab at full capacity (50,000 wafers per month at 28nm-110nm) would generate approximately $3-4 billion in annual revenue at benchmark utilisation rates. Against India's $52 billion market, that is roughly 6-8% of total consumption. Against the $28.6 billion segment estimated for 28nm-and-below chips, Dholera would cover approximately 10-14%. This is an inference from publicly available capacity data — no official percentage has been published by Tata or the government.

The two ATMPs add packaging capacity, not fabrication capacity. Micron Sanand packages NAND and DRAM memory for Micron's global supply chain — its output serves global markets, not specifically Indian consumption. The Assam ATMP can package 48 million chips per day for automotive and consumer electronics, but packaging adds 10-15% of a chip's total value by cost. The silicon inside those packages will still be imported wafers.

Taken together, all three ISM projects at full capacity likely cover less than 15% of India's total semiconductor consumption. The remaining 85%+ continues to be imported.

To change that equation materially, India would need 6-10 additional fabs at Dholera's scale — each requiring $11 billion and 50,000 wafers per month of capacity — plus supporting ATMP and OSAT infrastructure. NITI Aayog's May 2026 roadmap recommends a total ecosystem investment of $120-180 billion over the next decade, with the government funding roughly one-third. The current ISM allocation of approximately $18 billion across all approved projects is about 10-15% of what the roadmap says is needed.

The MeitY target of 50% domestic self-sufficiency by FY 2035 assumes this investment materialises. At current construction pace, India will achieve 50% self-sufficiency somewhere between 2045 and never.

What the numbers actually say

The India Semiconductor Mission is not a failure. It has produced one operational facility, one construction site, and one cleared plot. It has attracted the world's largest memory company to set up packaging operations in Gujarat. It has put semiconductor manufacturing on the national agenda.

But the gap between the announced number and the delivered number — between ₹1.64 lakh crore and ₹35,000 crore, between 300,000 jobs and 200 design hires, between "mid-2026 first silicon" and the creeping target of H1 2027 — is the story the press releases do not tell. And it is the only number that will determine whether the Mission succeeds.

The arithmetic is not complicated. India's chip consumption is growing at 19% CAGR and will hit $90 billion by FY 2030. Even if every ISM project delivers on time and at full capacity, the mission will cover less than a sixth of that demand. The remaining five-sixths requires a second mission announcement. Whether that announcement comes before or after the first one exhausts its budget is not a technology question. It is a political one.

Sources: IESA India Semiconductor Market Report 2030 (Jan 2025), NITI Aayog "Future of India's Semiconductor Industry" (May 2026), PIB press releases, MeitY statements (Jun 2026), Tata Electronics official disclosures, Micron Technology filings, Takshashila Institution semiconductor tracker (fabs.pranaykotas.com), McKinsey & Company (Jul 2025), Economic Times, Business Standard, Mint. Compiled June 2026.