Made-in-India Chips Are Real Now: What the Sanand OSAT Launch Means for India's AI Race
On 4 July 2026, India crossed a threshold it has chased for decades: commercial semiconductor production began at the CG Semi OSAT facility in Sanand, Gujarat. Beyond the ribbon-cutting, this is a story about whether India can supply the hardware its AI ambitions depend on.
What actually opened in Sanand
The facility inaugurated by Prime Minister Narendra Modi is an OSAT plant — Outsourced Semiconductor Assembly and Test. It doesn't fabricate silicon wafers; it takes fabricated chips, then packages, assembles, and tests them into the finished components that go into cars, telecom gear, and industrial electronics. That distinction matters, and we'll come back to it.
The numbers are still significant. The plant was built with an investment of over ₹7,500 crore as one of the first four projects approved under the India Semiconductor Mission. It has already reached an annual capacity of 20 crore (200 million) units, with a target of 500 crore (5 billion) units a year once fully operational — serving automotive, telecommunications, industrial electronics, 5G, and IoT customers.
The speed is the headline: foundation stone in 2024, chip testing by August 2025, commercial production by mid-2026. For a country whose previous semiconductor attempts stalled for decades, a two-year build-to-production cycle is a real signal — to investors as much as to citizens.
Why chips are the bottleneck for India's AI story
Every AI system — from a multilingual chatbot serving farmers in Tamil to a hospital's diagnostic model — ultimately runs on silicon. Today, nearly all of the advanced chips powering AI, in India and everywhere else, are fabricated in Taiwan and South Korea, designed in the US, and packaged across East Asia. India consumes; it does not yet produce.
That dependence has three costs. It makes India's AI infrastructure hostage to global supply shocks. It drains foreign exchange as AI demand grows. And it leaves strategic capabilities — defence AI, sovereign data centres — reliant on hardware India doesn't control. The government's answer is the India Semiconductor Mission, and Sanand is its first commercially operating proof point.
The honest caveat: packaging is not fabrication
An OSAT plant is the final third of the chip pipeline, not the whole thing. The wafers arriving in Sanand are still fabricated abroad. India's announced fabs — including the Tata-PSMC project in Dholera — are still under construction, and the leading-edge nodes that train large AI models (3nm, 5nm) remain a Taiwanese near-monopoly.
So no, the GPUs training India's next foundation model will not be Made-in-India any time soon. What Sanand does is different but still valuable: it builds the industrial muscle — supply chains, trained workers, quality systems, customer trust — that every mature chip economy started with. Taiwan itself began with assembly and testing in the 1970s before climbing to fabrication dominance.
A cluster, not just a factory
The more interesting development is what's happening around the plant. Micron's assembly and test facility, Kaynes Semicon, and CG Semi have all begun operations in the Sanand region within a short window, with suppliers of specialised chemicals, testing labs, and design services following. The Prime Minister explicitly framed the ambition in terms of Silicon Valley, Hsinchu Science Park, and Tsukuba — places where chipmakers, suppliers, universities, and talent compound each other.
India has run this playbook before. Mobile phone manufacturing went from import-dependent to the world's second-largest producer in a decade; electronics production is up roughly seven-fold since 2014 and exports eleven-fold. The bet is that semiconductors follow the same curve: start with assembly, localise components, then move up the value chain.
The workforce angle everyone should notice
One under-reported detail from the inauguration: a large share of the facility's workforce is young women, many from remote and tribal regions, trained in semiconductor processes including overseas programmes in Malaysia. That is exactly the kind of skills pipeline India's AI economy needs more broadly. India has over two million AI-skilled professionals on paper, but industry estimates put those with advanced skills at only one to two lakh — the gap between basic familiarity and deployable expertise is India's real talent problem, in chips and in AI alike.
What to watch next
Three markers will tell us whether Sanand is a turning point or a photo opportunity. First, whether the plant actually scales from 200 million to 5 billion units and holds international quality standards. Second, whether the Dholera fab and its siblings hit their production timelines, giving India wafers of its own to package. Third, whether design-stage work — where most of the value in a chip actually lives — starts migrating to Indian firms, not just Indian offices of foreign firms.
For India's AI ambitions, the equation is simple: models need compute, compute needs chips, and chips need an ecosystem. As of this month, that ecosystem has its first commercially producing node. It's a small piece of a very large puzzle — but for the first time, the puzzle has a corner piece in place.
Comments