Setting up a fab in India
The recent curbs on Huawei by US shows how technology has become a critical element in geopolitics. As the world becomes more dependent on tech , countries need to think strategically to safeguard their semiconductor supply chains, the building blocks of technology.
Setting the context
Worldwide, semiconductor manufacturing has consolidated over the years. Today, Taiwan’s TSMC is the only pure-play foundry in business. Samsung is the other contract manufacturer, but it is an Integrated Device Manufacturer (IDM) as well. Currently, only TSMC and Samsung have moved to the below 10nm nodes commercially.
The materials and chemicals for the fab are concentrated with Japanese players (like Tokyo Electron) while for lithography there is only one company worldwide which makes EUV systems costing around $200Mn — Dutch owned ASML.
The process to move the process to lower nodes has become so specialized that very few companies are able to make it. A lower node fab (less than 10nm) takes upwards of $12Bn of investment. Still, even with massive investments, making semiconductors at lower nodes is not everybody’s cup of tea —
India has the ISRO-owned Semi Conductor Lab (SCL) in Chandigarh with a 200mm size silicon wafer handling capability and 180nm technology node. This fab is still now outdated for analog RF chips.
If India has to play a leading role in 21st century, she has to take control of her semiconductor ecosystem and needs to focus on ways to build manufacturing competence.
Indian market has a huge demand for electronics. Over the last 5 years (till 2019), India has seen growth in production from $29Bn to $70Bn in electronics manufacturing. Our imports have also growth at a CAGR of 11% to $57Bn. A high import growth shows that electronics components & chipsets are still being imported. To kickstart the local electronics manufacturing along with setting up of a fab for chipsets, the government has come up with National Policy on Electronics in June 2020. The policy has 3 key elements:
— Production Linked Incentive(PLI): To create large scale manufacturing units primarily for mobile phones and electronics components.
— Scheme for promotion of manufacturing of electronic components and semiconductors (SPECS): This scheme focuses on helping companies Assembly, Testing, Marking or Packaging (ATMP) units to set up here
— Modified electronics manufacturing clusters (EMC 2.0): The modified scheme aims to create clusters where a few big players come together to work on a common technology. This is possible in the world of semiconductors.
Strategy moving forward
Start with analog chipsets
From the graph above, one can clearly see that there will be significant demand of higher nodes well into 2025. Lower nodes of 3 digits (130nm/180nm) are not technically high end and yet demand for these ICs will be more than $25Bn worldwide by 2025. Analog/RF chipsets typically require these nodes — with demand for 5G, one would require more analog and RF chipsets both in handsets and base stations.
India should start small and focus on these nodes (130nm/180nm going upto 65nm at the maximum). These require lower investments and would be great stepping stones to move to advanced nodes.
Get foreign players for advanced nodes. Give preferential market access (PMA) to Samsung
The above chart shows the top global foundry players worldwide. Indian government enjoys good relations with majority of them — TSMC and UMC are Taiwanese, Samsung is Korean, Globalfoundry is UAE based with fabs in US, TowerJazz is Israeli. Only with the exception of SMIC which is Chinese, all other players are from countries where Indian government enjoys good rapport. We should use these relations to encourage these players to setup a fab in India. The PLI scheme and EMC2.0 schemes should be fully exploited to encourage these players to setup a shop in India.
TSMC and Samsung are too invested in developing nodes lower than 7nm. We can talk to other players to set up nodes till 28nm. For this, we should be open to buying old used equipment used by foundries who have moved to higher nodes.
Use India’s demand to get fabs in India
India has a burgeoning mobile market. It is the largest segment in electronics as well. Looking at the top mobile players, we see clearly that majority are Chinese players. We should have a policy tied in to PLI scheme where government gives sops to mobile sets which have local manufactured chipsets in them. Local fab may need 3–4 years to come up. To get around this Arun Mampazhy explains in his Swarajya article, the idea of ‘virtual fabs’. Basically, start with an MoU with a foundry player with commitment to open a fab, and start routing chips through the Indian subsidiary.
III-V Fab and R&D Fabs for prototyping
A III-V fab which caters to defence and space sectors at an investment of upto $1Bn. A III-V fab uses Gallium Arsenide (GaN) or Indium Arsenide (InAs) or combinations from the III and V groups of the periodic table instead of silicon wafers as starting point. Since these materials can withstand extremely low temperatures, they are great for space applications and high power defence applications.
Along with a III-V fab, India needs a common R&D fab which can come up with an investment of $0.3Bn. We can pool in our research institutes, IISc and IITs for this initiative.
Startups which are making cost effective solutions will need services of these fabs wherein the existing material will be changed.
India has an established design ecosystem in the country. The physical design of these design houses are then sent to foundries outside for the actual product. These design houses can send their designs to the R&D fabs for the final chip fabrication. A good product can also lead to commercial production in the country as well.
These fabs can be used for academic level tape-outs as well. It is here that India can also explore a MOSIS type of service as well. This service can support Multi Project Wafer (MPW) technology, access to foundries, which is great of low volume production.
India should embark on an ambitious plan to set up a running fab by next elections i.e. 2024. That will be a real feather in government’s cap and will be the real Atmanirbhar program for India.
Taken many inputs from Arun Mampazhy’s writing in various media.