Compound semiconductors are low-hanging fruits. India must focus on silicon fabs

It is now two years since the Union Cabinet approved an incentive scheme for semiconductor manufacturing in India with an initial overlay of Rs 76,000 crore — and not much has been achieved to showcase.

As per a recent report published by the Ministry of Electronics and Information Technology (MeitY), only one out of 14 applications received till October for silicon fabs, compound semiconductor fabs, and chip packaging, has been approved. The one approved is the chip packaging unit of Micron, and despite claims from the minister that the “first Made-in-India chip” will be rolled out from this facility in 2024, it is likely to bring only about 5 per cent in value addition. Moreover, companies like Sahasra Semiconductors have reportedly started chip packaging in India.

The minister, in a recent interview, said that the “industry believes that compound semiconductors will be around half the market. That market currently has no dominant player”. This statement too warrants fact-checking.

Compound semiconductor-based products are used for certain segments in the industry, but they are not used for high-margin large-volume products such as processors and memory. Even in those select segments where compound semiconductors are used, silicon continues to play a major role though; some of its share is likely to be conceded to compound semiconductors in the years to come. 

As per a report by the US Department of Energy, in power electronics, key compound semiconductors Gallium Nitride (GaN) and Silicon Carbide (SiC) are expected to reach only about 20 per cent share by 2026; the rest will continue to be silicon. In itself, power electronics is a very small segment compared to much bigger segments such as communication and computing.

In a recent report, semiconductor market analysis group TrendForce predicted a $5.3 billion market for SiC devices by 2026 including automotive, renewable energy, industrial, charging pile, rail, and other segments. A Yole Group report from 2022 predicted a $6.3 billion market by 2027 including automotive, energy, industrial, transportation, telecom, infrastructure, consumer electronics, and others.

Another Yole Group report predicted a $2 billion market for GaN by 2027 including consumer electronics, industrial automotive, telecom-datacom, mobility, energy, defence, aerospace, and other segments. Even adding other compound semiconductors, the combined market is likely to be less than $10 billion — nowhere close to half of the overall semiconductor market that is expected to cross $700 billion by 2025 itself. 

The minister’s second claim that the market currently has no dominant players in the compound semiconductor segment also needs scrutiny. China has already progressed a lot and European and other companies are closely behind.

The word ‘devices’ is used in the compound semiconductor market report because the ‘chips’ made using those do not fall into the Very Large Scale Integration (VLSI) category. Silicon-based VLSI has surpassed 100 billion transistors on an integrated circuit or chip, however, compound semiconductor-based devices are mostly ‘discretes’ like high-power or high-mobility transistors, thyristors, etc., or at best small scale integration (SSI) with a few tens of transistors. 

The minimum feature sizes of the compound semiconductor devices are often in micrometres or even millimetres as opposed to the nanometre scale transistors fabricated on silicon for processors, microcontrollers, and memory. Hence, opening compound semiconductor fabs in India is not going to help us to later extend it to silicon fabs — the former is much smaller in scale and complexity when compared to the latter.

Anshuman Tripathi of the National Security Advisory Board (NSAB) debunks the claims of compound semiconductors like GaN replacing silicon in the near future. He also adds that since the compound semiconductor-based chips have very few transistors, the intellectual property content in them is much less, and hence the profit margins are quite less as well. 

One wonders why the right advice and numbers are not reaching the minister. Having many chip packaging units does not automatically prompt companies to show interest in opening silicon fabs in India. A case in point is Vietnam, which has had chip packaging units for over one-and-a-half decades, but no silicon fabs yet.

While the argument that ‘something is better than nothing’ holds good for Micron’s chip packaging unit and may be repeated if and when commercial compound semiconductor fabs are approved in India, those cannot be excuses for the continued delay to get the most important piece of the ecosystem — commercial silicon-based fabs.

Arun Mampazhy is a semiconductor engineer. X: @nano_arun.

(Disclaimer: The views expressed above are the author's own. They do not necessarily reflect the views of DH).