Risk of over-indexing Chris Miller's views on India's chip fab ambitions

Having the OSAT or the ATMP models does not automatically trigger interest in setting up fabs, a case in point is Vietnam which has had the OSATs for over one-and-a-half decades but no commercial fabs.

On the sidelines of the Tamil Nadu Global Investors’ Meet (GIM) in Chennai, the views expressed by Chris Miller, an Associate Professor of International History, Tufts University, have gained traction. Miller is also the author of a book which focuses on the geopolitics around semiconductor integrated circuits (ICs), commonly called silicon chips.

 In a media report, Miller is quoted as saying that the government ‘should be wary of spending too much in wooing semiconductor fabs’. Given that so far neither the Indian government nor any of the states have given even a single rupee as incentives to a commercial silicon fab, such statements could be wrongly interpreted and even influence decision-makers. The greater danger here is that it could kill the Ministry of Electronics and Information Technology (MeitY)’s policy towards semiconductors which looks sound on paper.

 Based on a Cabinet approval in December 2021, MeitY released policies for setting up silicon fabs, compound semiconductor fabs, and chip packaging units (ATMP or OSAT) in India for which the incentives offered by the Union government initially varied between 30 per cent and 50 per cent, but were made a uniform 50 per cent a few months later. Rs 76,000 crore was declared as the initial overlay for this incentive scheme whose mode of support is pari-passu and not production-linked as is often wrongly called.

 However, as per a recent MeitY report, only one out of 16 applications received till November for the categories mentioned above has been approved. Out of these applications, six are mentioned as ‘semiconductor fabs’, which are likely to be the most important silicon-based fabs as compound semiconductor-based fabs are mentioned separately. Only time and MeitY can tell if even one of those six applications will be approved, and, in turn, get the incentives.

 Miller is quoted as saying, “While it is good for countries to have a domestic fabrication base, they shouldn't over-index on fabs”. With zero commercial chip fabs of any kind and especially none of the Silicon-based high-volume fabs, India is far from the situation of doing any ‘over-index’ on fabs. 

 The statement that “there's more money made in chip design every year than fabs” needs context, lest its misunderstood. Fabless design companies are said to have had a 33 per cent share of the overall chip market in 2020; a majority of the remaining portion is cornered by Integrate Device Manufactures (IDMs) who design and fabricate chips as well as pure-play foundries that fabricate chips on a contract basis for fabless companies (as well as for some of the IDMs).

 In other words, even though in theory fabless companies could exists with majority ownership in a certain country even if there are no commercial silicon fabs, practically such cases seem to be rare, or none.

 It is estimated that Indian companies — defined as those in which a resident Indian citizen beneficially owns at least 50 per cent of the capital — and the homegrown semiconductor design sector, collectively, has a revenue of only $30-40 million at the moment compared to the $300 billion or more global annual revenue for chip design — i.e. about 0.01 per cent.

 Miller's interview to Deccan Herald highlights this: “A lot of design happens in India, but they are mostly for foreign companies. I think the challenge is taking the expertise in design that already exists and building new products and new (Indian) brands around that expertise. I believe it is not a technical challenge but a business expertise challenge”.

 However, the statement that “usually, the first step on the ladder is device assembly, followed by testing and packaging, and then fabrication” too needs context. Historically, fabs themselves did the testing and packaging of the chips. Over the last couple of decades, the lower-value-addition part has started getting outsourced either to a third party (OSAT model) or to a facility of its own in a different country (ATMP model) like the case of US-based Micron opening a unit in India. 

 Having the OSAT or the ATMP does not automatically trigger interest in setting up fabs, a case in point is Vietnam which has had the former for over one-and-a-half decades but no commercial fabs.

 To summarise, as Miller said focusing on branding the design expertise is a way forward, but the real difference in the semiconductor space, from commercial and geopolitical angles, is in setting up high-end fabs. That will make the difference!

(Arun Mampazhy is a semiconductor engineer.)

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