India set up its first chip making unit in 1983 in Mohali but then made little progress in this segment, until the government launched a Rs 76,000 productivity linked incentive scheme for chip making last year.
Geopolitically, India’s improving relations with the US have garnered it global cheer in its ambitions. But a long, complex road lies ahead before it becomes a chip making hub
Chris Millerauthor of the corporate bestseller Chip War spoke to ET about the fast evolving, powerful business of semiconductors and India’s prospects in it.
India is a late starter in the highly complex world of semiconductors. What are its prospects?
India is a small player today, but the country has large ambitions across the electronics supply chain from the production of chips all the way through their assembly into devices like smartphones or computers. The Indian industry and the Indian government are trying to play a much larger role in the industry.
How do you read India’s ambitions to become a significant player in this industry?
India wants to move up the electronics value chain, not only assembling devices, which is the lower value added portion of the electronics industry, but also making some of the higher value components inside of electric devices.
If you look at a typical phone, the assembly process only costs a handful of dollars, while the chips inside sell for up to hundreds of dollars. And so most of the money in a smartphone is made by the companies that produce the chips inside of them. So this economic rationale is a key driver of India’s interest in semiconductors.
There is an obvious geopolitical angle too…
Yes, the second facet is political because India doesn’t want to be reliant on China for the electronics it uses. And developing independence and semiconductor capabilities is seen to be an important means of establishing India’s technological sovereignty vis-a-vis China and other countries. The Indian government has been pouring substantial incentives into the chip industry, trying to attract foreign companies to invest, as well as trying to support Indian firms. This comes at a moment when the entire electronics industry outside of China is looking to rebalance away from the current position of extensive reliance on China for the production of simpler components, as well as for the assembly of electronic devices.
And many companies are looking in India as a potential location for their packaging and assembly. Companies like Foxconn and Micron want to build up their electronics production in India, making India a larger player in their supply chains.
India wants to embark upon what you have described in your book as one of the most complex and expensive pieces of manufacturing in the world. What makes chip fabrication so complex?
To design an advanced chip, you need ultra specialized software. You need highly precise machine tools capable of undertaking manufacturing processes such as laying down thin films of materials, just several atoms thick with basically perfect uniformity.
Then you need to collect chemicals and materials that are ultra purified with 99.99999% purity in many cases..so pure that hardly any atoms of the wrong material are present inside of them. And this purity is necessary because to make an advanced semiconductor, you need to produce thousands, millions, often billions of microscopic transistors, tiny electrical switches that turn on and off on the surface of your chip.
And in a typical smartphone, for example, just the primary chip is roughly the size of your fingernail, but it has carved into it, billions and billions of tiny nanometer scale transistors. Those are transistors measured with the billionths of a meter. They are among the smallest devices humans have ever produced, and we’ve produced more of them than any other manufactured goods.
We produce more transistors every year than there are cells in the human body.
And the super complicated manufacturing is just the start. Any new player will have a tough time embedding its operations in the global semiconductor supply chain, which to cite your book again is full of monopolies and oligopolies which make up what you call “chip choke” points. Can you please explain the complexities of the supply chain?
Manufacturing of advanced semiconductor requires cooperation from dozens of different companies in multiple countries. From the earliest stage of a chip’s production, there is a series of choke points throughout the supply chain in which companies often have predominant market share or in some cases a complete monopoly of the know-how and machinery needed to make a chip.
The production, for example, of the most advanced lithography tools, is done by just one company, ASML in the Netherlands, which has 100% market share.
There’s one company, Taiwan’s TSMC (Taiwan Semiconductor Manufacturing Company) that produces around 90% of the most advanced processor chips that are in smartphones and computers, and advanced data centers.
If you dig deeper into the chemicals used in the process, you just have just a couple of companies capable of purifying materials to the required level. And they’re largely based in Japan as well as a handful in the US and in Europe. The design of semiconductors is also something where there’s, in many cases, highly concentrated markets.
Your book is perhaps the best chronicle of the semiconductor, right from its inception during the Cold War to bring computing into the military to its current status as something that powers the modern world. Along the way, how did it become such a powerful geo-political tool?
The chip became a powerful geo-political tool from its earliest invention. They were used by the US in missile guidance systems during the early Cold War, during the arms race between the US and the (then) Soviet Union.
Today, almost every type of military system has dozens or hundreds, even thousands of semiconductors inside, helping military systems act semi-autonomously. The application of AI for defense uses is one of the new frontiers that militaries are exploring. The most advanced chips will be critical inputs to this effort.
How do AI power chips?
To train an advanced AI system requires access to some of the most advanced semiconductors. Systems like ChatGPT are trained by reading vast amounts of text. They recognize patterns between the different words, and that’s how they’re able to produce languages
And this process requires an extraordinary amount of data processing. And only a couple of companies can produce chips with that kind of processing power. Across the world, there’s a deficit of the most advanced AI training chips above all the types of GPU graphics processor unit chips produced by Nvidia.
How are big tech firms trying to make it big in this segment?
The past several years have seen many big tech companies begin to design their own semiconductors. They’ve realized as they build out vast data centers, if they can have chips that are perfectly designed to optimally carry out the specific workloads they need, they can get major advantages in terms of speed. So today, almost every big tech firm–Facebook, Google, Amazon–is designing and in many cases, using, its own semiconductors for its own internal data center purposes.
And some of these chips have become quite important: Google’s TPU (tensor processing unit) is a really important and central chip in the training of artificial intelligence systems.
We are going to see over the coming years, even more tech companies bet on producing or at least designing their own semiconductors that are perfectly tuned to the types of computing they need to undertake.