India’s Semiconductor Race: 40 Years of Challenges and Revival

It is tempting to blame the 1989 Mohali fire for everything. It mattered: SCL lost eight years of production in the decade when Japan, Korea and Taiwan scaled up. But India had three more serious chances afterwards, in 2007, 2014 and 2022, and each failed for reasons that had nothing to do with the fire.

1. Capital that never stops. A fab is not a one-time build. Each new process generation needs new tools, and a factory that stops upgrading falls behind within a few years. SCL, a single government unit, could not fund that treadmill.

2. Customers. In the 1980s India’s electronics market was small and protected. A fab needs enough orders to run its tools around the clock. Without phones, PCs and cars made in India, there was no one to buy at scale.

3. The supply chain. Fabs depend on ultra-pure chemicals, specialty gases, photoresists, wafers and machines from a small group of suppliers in the US, Japan and the Netherlands. None had a base in India.

4. The foundry model. After 1987, TSMC showed that manufacturing could be a separate, specialised business serving the whole world. Taiwan built that; India’s government fab could not compete for global customers.

5. Partners who walked. The 2014 consortia and the 2022 Vedanta-Foxconn venture had money on the table and still did not proceed. A fab needs a technology partner that commits for decades, which is why the 2024 Tata-PSMC deal is watched so closely.

Meanwhile India built the other half of the industry. From Texas Instruments’ Bangalore centre in 1985 to the global capability centres of the 2010s, Indian engineers designed chips that were fabricated elsewhere. In 2026 the government says India has nearly 20 percent of the world’s chip-design workforce.

Semicon 1.0 paid mostly for factories. The government’s own September 2026 backgrounder says Semicon 2.0 “adds machines and materials, design, research and talent to the mission”, so that the chain runs from raw materials to finished chips, “and not merely its closing stages”.

The headline race is about 3 nm and 2 nm processors for phones and AI servers, and in 2026 Taiwan makes nearly all of them. India’s first fab starts at 28 nm to 110 nm. That sounds far behind, and in pure technology terms it is.

But most chips in the world are not leading-edge. Microcontrollers in cars, power-management chips, display drivers, sensors and telecom parts use mature nodes, and the 2020-21 shortage was mostly a shortage of exactly these cheap parts. Starting there gives a new fab real customers. Semicon 2.0’s R&D pillar then aims to push toward more advanced nodes with partners inside and outside India.

Can Dholera start on time? The government’s date for the first fab is 2028. Fabs routinely slip; a new fab in a new country is harder still.

Can it reach good yields? A fab that produces too many defective chips loses money on every wafer. Yield is learned over years, which is why partners like PSMC matter.

Will suppliers come? A fab needs hundreds of local suppliers of gases, chemicals and spare parts. That is exactly what the machines-and-materials pillar is meant to attract.

Will Indian chip IP sell? The DLI start-ups have designs; the test is whether they win global customers.

Will policy last? Every earlier attempt faded. A fab takes years to pay back, so support has to outlast political and technology cycles.