Semiconductor chips are used by millions of Indian citizens every day: from cell phones and laptops all the way to data centre servers (which power email, social media and instant communication apps, among many other things), to the GPU servers needed for AI training and inference are powered by semiconductor chips.
Semiconductor chips are used not just for carrying out the computation, but also for the storage and retrieval of information (like data, pictures, videos, etc.). The latter types of chips are used in storage devices like SSDs, which are responsible for quickly storing and retrieving data and videos.
Contemporary computer systems are assembled via a very complex supply chain that is located at multiple places all across the globe. Consider an Android phone. All such devices need multiple components: a System on a Chip (SoC), which is the brain of the computer, a display device, memory (DRAM) and storage (NAND Flash) devices, which help in permanent and temporary processing and storage of data, pictures, etc.
These components have to be procured and then assembled together to create the working phone. Once assembled, a complex software ecosystem of operating systems, compilers, runtimes and application development frameworks is required to ensure that the phone is usable by the end user. Let us focus purely on the hardware.
Assembling a phone in today’s world is no easy task. To get a sense of the complexity and the geopolitics involved in assembling the phone, let us trace the journeys of the four main components that make up a modern phone—the SoC, the display device and the memory and storage devices.
The SoC’s journey begins as intellectual property (IP), with core designs often licensed from ARM, a UK-based company. An American firm like Qualcomm or a Taiwanese one like MediaTek then creates a detailed design for the processor, which is then sent to a semiconductor foundry to be “fabricated”. These multi-billion dollar foundries, run by TSMC (Taiwan) or Samsung (South Korea), etch billions of transistors onto silicon wafers. These wafers are then shipped to Malaysia or Vietnam, to be cut, tested, and packaged into the final SoC to be sent to the phone assembly line.
The same is the case with the other components in the phone. An OLED screen for the display comes from the US or Germany, the glass cover from Corning (USA) and is all assembled together to form the display in South Korea (Samsung/LG) or China (BOE).
The phone's RAM and storage rely on chips dominated by three big companies: Samsung and SK Hynix (South Korea), and Micron (USA). These chips also need “fabrication” like the SoC and are often sent to another country for packaging. Finally, the camera is made of technology licensed from Sony (Japan). The camera also requires a large number of minuscule motors to provide autofocus and image stabilisation, which generally come from another Japanese specialist company.
Finally, all these intricate parts are meticulously assembled into one compact module by a company like LG Innotek in South Korea or Sunny Optical in China.
One thing is quite obvious: there is very little in the supply chain steps from before that India is currently capable of doing indigenously. The design of these chips, both for computation and storage, is mostly done outside India. Whatever design is being done within Indian borders is being done by multinational corporations, who hold the intellectual property for those designs.
The country, of course, doesn't have any fabrication facility for “manufacturing” semiconductor wafers, except SCL Mohali, which is capable of a much older fabrication process based on 180nm technology. This is 11 technology generations behind the technology node in which the current generation of SoCs is being fabricated (which is 2/3 nm process technology).
Catching up in the semiconductor world, especially for setting up fabrication facilities, is a capital-intensive task. For example, setting up a facility for fabricating chips in 45 nm, which is roughly 5-6 generations behind the current state of the art, but still an acceptable technology node for SoCs that do not need high performance, would require capital investments of approximately $10 billion. This number can increase by one order of magnitude for a state-of-the-art technology node facility.
But it is not all doom and gloom. Given that we started late as a country in this area, in order to catch up, we should start doing the things that are not as capital-intensive, have smaller setup timelines, would create a large number of jobs and provide incentives to the ecosystem to grow. OSATs (Outsourced Semiconductor Assembly and Test), or facilities which are responsible for assembling together discrete components (e.g. all the components of the smartphone mentioned above to create a phone) into working products, are a step in the right direction. Although many components they put together will still be imported, these are a good start for multiple reasons. Similar to what happened when car manufacturing companies started in India, the hope is that OSATs (or the in-house versions for other companies) will incentivise the local ecosystem to create, manufacture and supply semiconductor components which they need, creating a virtuous cycle of organic growth, and not just for the semiconductor industry.
The OSAT ecosystem will also provide an impetus for local designs by providing local startups access to markets that were previously non-existent within the Indian mainland. There are a large number of low complexity but high volume semiconductor chips called microcontrollers that are being used in everyday appliances like washing machines, CCTV cameras, microwave ovens, BLDC motor controllers and hobbyist development boards (like the Arduino).
Today, more than 90 per cent of these chips are being imported and are put together to be used as modules, many times by units within India. OSATs with their high-volume production capacities will only increase the demand and volume for these chips. The other benefit of microcontrollers is that they do not have very high performance requirements like server-class CPUs or GPUs, and hence can be fabricated in much “older” technology nodes indigenously, for example, in SCL Mohali, reducing the dependence on imports.
Once the OSATs are established, they can serve as customers which Indian chip startups (or even established design companies) can readily access and interact with. The presence of these facilities within the Indian shores lowers access friction for Indian fabless startups to have conversations with their potential customers (the OSATs), help better understand their requirements, as well as the market they serve. These conversations allow the existing players to gain a better understanding of where the gaps lie and help design better products which can serve as replacements to existing ones.
Successfully creating IP for low complexity and high volume devices, translating them into products, which in turn are used within the semiconductor ecosystem within India, will provide the indigenous companies the confidence and maturity to be able to create and deliver products. In addition, it will also provide the much-needed confidence in the indigenous ecosystem to be able to take on more complex designs like server-class CPUs and GPUs and in due course, on time, deliver on them successfully.
The author is Associate Professor of Computer Science, Ashoka University
The opinions expressed in this article are those of the author and do not purport to reflect the opinions or views of THE WEEK.