Semiconductors

This industry exists because a transistor, a tiny electronic switch, is the fundamental building block of every piece of modern computing, and packing billions of them onto a single chip is what turns a lump of purified sand into a smartphone processor. No other manufactured product compresses this much precision engineering, physics and capital into something small enough to fit on a fingertip.

The second reason is that chips have become the substrate almost every other industry now depends on. A car, a washing machine, a fighter jet and a UPI payment terminal all rely on chips, which is exactly why chip supply has become treated as a national security and economic sovereignty issue, not merely a component sourcing question, by governments including India's.

Almost nothing electronic works without a chip, and almost nothing India buys electronically is chip-independent, from a car's engine control unit to a UPI payment terminal. This industry designs, manufactures, packages and tests those chips, a chain so specialised that even the world's richest technology companies depend on a tiny handful of suppliers for the most critical steps.

India's position in this industry is a deliberate, staged bet: rather than trying to leap straight to leading-edge chip fabrication, the country has built its first real capability in design, testing and packaging, the less capital-intensive layers, while its first high-volume fab targets a mature, well-understood process node rather than the cutting edge.

Every other number in this industry, node size, fab cost, packaging sophistication, ultimately funnels into one governing metric: yield rate, the share of chips on a wafer that actually work. A fab pays roughly the same cost to process a wafer regardless of whether 95% or 60% of the resulting chips are sellable, so yield is not a quality footnote, it is the direct multiplier on whether a fab makes money at all.

This is exactly why new fabs and new process nodes are treated cautiously in the early years. Yield typically starts low on any new line and climbs over months or years as engineers eliminate defect sources, which is why India's own semiconductor ambitions are consistently described as a multi-year project rather than something that shows commercial results overnight.

India currently imports nearly all the chips it consumes, alongside importing essentially all of the fab equipment, wafers and specialty chemicals now going into its first domestic fab and ATMP facilities. This is a rare case where India's declared long-term strategy is to convert a large existing import bill into domestic production capacity, rather than the more familiar story of an entrenched, structural import dependency the country cannot realistically close.

The investment scale involved is substantial for a first-generation effort. The Tata-PSMC fab alone represents roughly Rs 91,000 crore (~$10.9 billion) of investment, targeting first silicon in late 2026, while Micron's Sanand ATMP facility, a $2.75 billion investment, was inaugurated in February 2026 and began shipping memory modules to Dell and HP within weeks. Combined PLI and DLI government incentives of roughly $9.1 billion are structured to cut qualifying projects' capital costs by up to 50%, directly changing the investment calculus for global chip companies choosing where to build next.

Foundries like TSMC and India's own Tata-PSMC fab earn a manufacturing margin per wafer processed, with profitability driven overwhelmingly by yield and capacity utilisation rather than chip design cleverness, since they are not selling their own chip designs.

Fabless design companies earn on the margin between what a foundry charges to manufacture their chip and what they can sell the finished, branded chip for, meaning their profitability depends on design differentiation and brand strength rather than manufacturing scale.

ATMP and OSAT companies like Micron's Sanand facility earn a service fee for assembly, testing and packaging, a lower-margin but also lower-capital-intensity business than fabrication, which is exactly why it has been India's practical entry point into the industry.

This is one of the few industries on this site where the upfront-cost story genuinely holds without much qualification. A leading-edge fab costs tens of billions of dollars and years to build before it produces a single sellable chip, and once built, its ongoing running costs are dwarfed by the sheer scale of that initial capital commitment.

What makes this cost structure especially unforgiving is that the payoff is entirely contingent on yield ramping up fast enough, before the process node itself becomes commercially obsolete. A fab that is a generation behind the leading edge by the time it reaches good yield has effectively already lost the race it was built to run.

Over the next five years, expect India's first fab and ATMP facilities to move through their yield ramps, a second wave of PLI/DLI-backed projects to add capacity, and India's fabless design ecosystem to keep growing off a genuinely strong existing engineering talent base.

The realistic five-year outcome is not India competing with TSMC at the leading edge, it is India securing a meaningful, defensible position in mature-node fabrication, packaging and chip design, categories where the barriers to entry, while still high, are not the near-impossible ones guarding the industry's absolute cutting edge.