What is different about India’s chip push this time around?

There were actually no trials in the past—there were isolated attempts by private individuals, mostly from abroad, who would work with states, never with the federal government, to launch a fab. It did not materialize for the obvious reason that there are lots of factors involved in building a fab—not just building a fab, but you need to build an entire ecosystem. I know it because I was invited to be a figurehead in those activities many times.

This is the first time that the government of India has come to a recognition that this technology is vital to India’s strategic interest, for it to gain some self-reliance, and to perhaps capitalize on the fact that if the whole world will take advantage of this market doubling in 10 years, why can’t India take advantage of it?

Do you think the targets are achievable?

I see much more viability, probability and possibility of it happening now than ever before. If it’s all smoke and mirrors, then you can come, and you will look good for a while, but it will eventually fizzle out. Let’s look at three-four different levels of it. If we sign up with an existing integrated device manufacturer (who has its own fabs), they can start tomorrow. It will take them 18 months or so to build a fab. Then, let’s say, within two years, they build their products, have their own machines, and bring their own people initially.

What we have given them is land, electricity, water etc, and an incentive to the tune of 70%. They can start producing within a year or so after that, which makes it two-and-a-half, a maximum of three years.

The longer period is where we do a joint venture between an Indian partner and a foreign partner, where they bring in all the technology and equipment, and we bring in capital, management and skills. The challenge for both of them would be what product they should make in the fab.

Could you elaborate?

Ideally, the partner coming in should have a product in mind. For that, the Indian market needs to be analysed to find our top 10 needs because we’re importing $180 billion worth of electronics. The rule of thumb is 10-20% of that is semiconductor value. If you round up, that’s $10-20 billion. These days, it’s more like 20 than 10, because the number of chips going into a device is larger, which gives you a $20 billion margin.

We need to analyse where we have an edge and where we can try to make it for the Indian market. Otherwise, the issue will become that the Indian market will buy from outside if they can buy at lower prices there. We may have to incentivize them a little bit initially to kick-start this whole thing.

We’re trying to do so many things—train people, create demand, create the fab to make the products. Unlike the rest of the world, where it has been done for over 50 years, we’re trying to do it over a five-year period, which makes it so complex.

So, how will we do it?

What they (other countries) did in 50 years is that they gradually learned things. We don’t necessarily have to learn all that. For example, an 8nm node is an advanced sweet spot and is like a piece of cake (for the industry); our engineers can learn that. Remember that the manufacturing side of this is also highly automated.

If the biggest semiconductor firms run R&D in India, why do we not have the skills available?

Our skill set is in design. We have enormous skills in what is called very large scale integrated circuits (VLSI) design. There are around 25,000 design engineers in Bangalore designing chips for companies (like Samsung, Intel, Qualcomm, etc.).

But what we’re talking about is domestic fabrication of the chips. We want to have some engineers step out into startups, which is why design-linked incentives are being given. We’ll give them free tools and money to design chips.

We saw the PLI being revised to provide a 50% incentive across technology nodes. What kind of fabs should India build?

We’re not yet there where we give the highest incentive to the most advanced nodes. We ourselves aren’t pushing for that. It’s not the most optimal way to start, so we kept it flat.

Have Indian companies shown interest?

We have three local proposals for just semiconductors. We have a couple of proposals for displays, and dozens for the design-linked PLI. Our job as advisors is to evaluate and advise on how to make it work if there’s viability. There’s also a whole group of people who look at the financial viability. We’re going through that process right now.

Do you have a timeline in mind, like when somebody actually starts work here?

We should see that within a year.

In the context of semiconductors, how long will Moore’s Law hold?

The original Moore’s Law is based on Gordon Moore’s (Intel’s founder) observation in 1965 that the number of transistors packed on a chip will double every 18 months. He extrapolated that, based on the first 3-4 year’s data, and said that if it continues that way, we will have a device in our hand, which we’re today calling a smartphone. Turns out that observation didn’t have to come to fruition. Because along the way, it got modified to two years over 18 months. But every two years, packing twice as many transistors, it’s exponential growth and increasingly challenging.

People have been calling for Moore’s Law to come to an end for a long time, but the ingenuity of people has kept on extending it. Having said that, you will notice that in the last decade or so, the rate of doubling has slowed down, which is fine. We are now manufacturing chips at 5nm, and as we get to 3nm, 2nm and 1nm, we’re fundamentally reaching the limits of atomic physics.

So, what’s the way ahead?

From a physics point of view, it will get more and more challenging. Even if people are able to solve physics, the second challenge is that the money required to come up with new requirements for that physics has gone up even faster exponentially than the law. The latest machine—I was reading—can go from $10-100 million for one. Not many countries, not even companies, can afford that, which explains why this industry got consolidated from dozens of companies to a handful now. Even in that handful, only a few have their own fabs for fabricating wafers, especially for the leading edge (chips).

There have been talks about experimenting with new materials to replace silicon in chip making.

There has been lots of talk about carbon nanotubes, graphite, etc. I’m sure there are people trying every possible way, but none of them has yet reached a point of commercial viability, which is the challenge.