There’s a fog of uncertainty over semiconductor fab costs

When it comes to India’s semiconductor mission, the Vedanta-Foxconn JV is the most talked about, but there are many unanswered questions about its capabilities

After the Government of India offered schemes granting fiscal support for semiconductor manufacturing units (fabs) in December 2021, there has been little progress on the ground in realising India’s fab dreams.

Out of the three groups that applied for silicon fab incentives before the first round deadline of February 2022, the proposal from the International Semiconductor Consortium (ISMC) has been put on hold until the acquisition of its technology provider Tower Semiconductors by Intel is complete. Meanwhile, there are reports that the main stakeholder, Next Orbit Ventures, is considering selling its stakes to the BC Jindal group as the government asked the ISMC to rope in a ‘strategic investor’.

A second applicant, IGSS Ventures, had claimed in an interview in April that ‘The technologies proposed broadly encompass 28 nm digital/logic, analogue and mixed-signal, Silicon photonics, GaN-on-Si, high voltage, etc.’ and given that all this is possible for a project cost of $1.5 billion, but later the project cost figure was mentioned as $3.5 billion.

India’s incentive policy demands production-grade technology license and the IGSSV has so far not revealed where exactly it would get that. As per a latest report, the IGSSV is backing down from its 28nm plans. Even for its 65nm fab plan the technology partner it claims to have runs only an R&D fab, which is much different from licensing from a high-volume manufacturing fab, and ideally this should not qualify for government incentives.

The third applicant, a joint venture (JV) of Vedanta and Foxconn, has been making the most noise in the media ever since they have applied for the incentives. For example, more than a year ago, Vedanta Chairman Anil Agarwal claimed that the semiconductor fab will come up within the next two years; the fact remains that the Centre is yet to even approve the project.

For long, the JV claimed that it has the necessary knowhow of 28nm technology, however Foxconn at best is known to have access to only 110nm technology — that too indirectly through partial ownership of a fab in Malaysia. Silicon fabs below 90nm technology are typically operated using wafers of 300mm size, and Foxconn does not have experience running any such.

With multiple questions raised about its capabilities, the JV has come out with an interview which unfortunately raises more questions and doubts.

The newly appointed CEO, David Reed, says ‘Foxconn has access to production-worthy high-volume manufacturing for 40nm nodes. It is signed, sealed, and accessible to us’. The question is how? Foxconn does not run or is not known to own stake in any such fabs.

Could it have licensed it from another fab? Possible, but the talks with STMicroelectronics (STM) was reported to have hit a roadblock, and it also appeared that STM was more comfortable with Vedanta and not with Foxconn. This is understandable as technology going into the hands of a group like Foxconn with ambitions to move up the value chain can be long term competition and challenge for STM itself.

Nevertheless, let us assume that there is indeed a 40nm licence; the next question is, has Vedanta reworked the project cost? As per media reports, the JV ‘has submitted a capital expenditure estimate of $10 billion to the Indian government, the people said. The government considers that figure inflated and estimates $5 billion is closer to the true cost’.

In his latest interview, Akarsh Hebbar, the MD of the group, said the investment projections “remain the same”, which does not add up. The government estimate of $5 billion for a 40nm fab of 40,000 wafer starts per month (WSPM) capacity is more realistic. Even for a 28nm fab of that capacity $7-$8 billion (including technology license fee) is reasonable—and both estimates are far from the quoted $10 billion.

Claims by Reed that ‘The 40nm is the basis, and on top of that drops, the 28nm. The interchangeability between those two is about 80-85 per cent’ should get the government all alert on such a claim. The 28nm fab is much more challenging, the kind of flavour STM offers at 40nm is not what it offers at 28nm. Vedanta will have to either buy the 28nm licence from STM or someone else to be able to qualify for incentives based on estimates for a 28nm fab, or it will have to develop the 28nm capability inhouse — which is not easy, and will be akin to ‘reinventing the wheel’. A detailed analysis of the technology differences or additional challenges at the 28nm level will need a separate article.

Fifty per cent incentives from the Centre and another 20 per cent from the Gujarat government should be for proper project cost estimates based on production-grade technology licence already possessed — not based on what ‘we can get in future’, and, in either case, exaggerated project costs must be rejected.

A lack of overall understanding, transparency, and tall claims has shrouded the progress in India’s semiconductor prowess thus far in uncertainty.

(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.