Pranay Kotasthane Unravels the Global Semiconductor Industry

Shruti and Pranay Kotasthane discuss the evolution of the global semiconductor industry, India’s semiconductor industry ambitions, and the interplay of geopolitics, economics, and technology in the semiconductor industry.

SHRUTI RAJAGOPALAN: Welcome to Ideas of India, where we examine the academic ideas that can propel India forward. My name is Shruti Rajagopalan, and I am a senior research fellow at the Mercatus Center at George Mason University. 

Today my guest is Pranay Kotasthane who is the deputy director of the Takshashila Institution and chairs the High Tech Geopolitics Programme

Pranay co-writes Anticipating the Unintended,a newsletter on public policy ideas and frameworks, and co-hosts Puliyabaazi, a popular Hindi-Urdu podcast on politics, policy, and technology. He is the co-author of Missing in Action: Why Should You Care About Public Policy, and the graphic nonfiction narrative We, the Citizens. He has co-edited India’s Marathon: Reshaping the Post-Pandemic World Order.

Today we are discussing his most recent bookWhen the Chips Are Down, coauthored with Abhiram Manchi. We spoke the evolution of the semiconductor industry, industrial targeting, Moore’s law, Rock’s law, Taiwan’s Comparative advantage, whether India can lead in semiconductor chips and more.

For a full transcript of this conversation, including helpful links of all the references mentioned, click the link in the show notes or visit mercatus.org/podcasts

Hi, Pranay, welcome to the show.

PRANAY KOTASTHANE: Hi, Shruti, glad to be back here.

RAJAGOPALAN: I’m so thrilled to have you, and this is because you have a new book out. This one is “When the Chips Are Down” with your co-author, Abhiram Manchi. You have another book out. Because you write books faster than I can read them. This is a graphic narrative, graphic novel about all things related to our republican constitution.

KOTHASTANE: “We, the Citizens.”

RAJAGOPALAN: We, the Citizens,” which is all things related to our republic, our rights, our constitution and more generally how we must live together. Now that you have three books out, and I like them all, I have decided that I should start publicizing to the world that you are actually my student, one of my very first. As you know, success has lots of parents and teachers. Now, that you are a three-times published author, well done, one of my earliest students.

KOTHASTANE: Yes. Absolutely. Actually you taught a course on constitutional history, and that was one of my first introductions ever into this subject, and I still remember some of the concepts and ideas, so thanks.

RAJAGOPALAN: I think you are the master now, and you have written more about the topic than I have, but I’m so thrilled. Today, I want to talk with you primarily about “When the Chips Are Down.” Before you took my class, which is of course the event that led to all your future successes—I can’t even say that with a straight face—in your past life you were actually one of the engineers working on semiconductor design, starting with Texas Instruments and other places. Can you walk me through very briefly that journey and that part of your life before we jump into the geopolitics and geoeconomics and technological aspects of the book?

KOTHASTANE: Sure. Yes. I did engineering from NIT Surathkal, one of the top 10 colleges back then in India. Once I did that, and I did my electronics engineering there, I always had an interest—I was always fascinated with semiconductors and the fact that you have all these brilliant and really wonderful magical things like computers, et cetera, and the hardware behind that software always fascinated me. I always wanted to learn that better, so went to electronics by choice. Then after doing engineering, I got a job opportunity to work at Texas Instruments.

Again, it was my dream company back then to work at that place because Texas Instruments was one place where all these things really started. Even the Nobel Prize for the integrated circuit was shared by Jack Kilby, who was an engineer at Texas Instruments. That was fascinating. That’s when I started a career as a design engineer. My title was IC [integrated circuit] design engineer. Back then in India, in Bangalore, even though we know it as a hub of software, it is also a hub for chip design. Now, the government estimates there are 20% of the world’s design engineers in Bangalore, but roughly 300,000.

Back then, also there were a lot of things happening, and these are largely MNCs, which have set up their R&D centers in Bangalore and so on and so forth. Texas Instruments was the first MNC to set up shop in Bangalore. Technically, Bangalore is a Silicon Valley in the sense that it was the first company which does make silicon. I started in 2007, was largely doing one small part of the chip design process. My team used to make mobile processors for phones. Phones weren’t the most—not the iPhone types, but new smartphones, which were coming up back in 2008, ’09. Back then, we were making the processors.

One processor team might have thousands of engineers working on it. It’s a very modular process, so some work on a small, let’s say, an imaging aspect of it, or some are working just on the USB side of it, et cetera. All of that gets integrated into someone who’s working on the entire chip-level stuff. Then there are some other teams working on how does it integrate with the phone and the final OEM, et cetera. Those are the layers. I was working on one small component of it, which was one small part of the graphics processor. After a few years, I was working on the entire chip processor-level team.

One of the good experiences I had is, in India, for example, we know that a lot of design happens there, but you don’t get to actually see the hardware because we don’t have any commercial manufacturing facilities. The good thing at Texas Instruments was we did have low-cost tester. We wouldn’t manufacture the chips, but they used to come back in the office, and you could actually get to see the physical stuff and do some testing, et cetera, on a big room-size machine. That was some of the good aspects of it, and it was great fun. I did that work for around seven years before your class changed my journey.

RAJAGOPALAN: I wish I could take credit for that. Everything you’ve described about what you did, which is this extreme level of specialization, how the design aspect, the manufacturing aspect, the assembly aspect, all of these parts got separated. By the time you entered that business, that had already taken place, and why India has really been more on the design side and not been in the manufacturing and even the assembly, which now we’re hoping to enter. You explain these things beautifully in your book. Before I get to India, I wanted to talk with you a little bit about the historical progression of the semiconductor technology and that global industry.

In your book, you describe it in three specific ways. You talk about the geopolitical progression, the geoeconomic progression and the technological progression. These aspects are separated. When I was reading the book, to me, it felt like at the core of these three are Moore’s Law, and that’s basically the observation that the number of transistors on a microchip doubles every two years. It was actually a prediction which turned out to be so accurate that it became a law. 

The second is Rock’s Law, which was again a prediction that the cost of a semiconductor chip fabrication plant actually doubles every four years, which again has turned out to be so true that it’s a law.

The third is, of course, the standard old, established laws of absolute and comparative advantage. To me, it seems like these three are the governing laws of the history of the semiconductor industry. Can you parse this out for us bit by bit?

Moore’s Law

KOTHASTANE: Yes, sure. Let’s begin with the Moore’s Law and that idea. As we know, before semiconductors and chips, we still had electronics in a sense through something called a vacuum tube. But these were really huge machines, used to take substantial part of the electricity of a small town back then, to do the kinds of things you can now do at very low voltages and current. That was the technology back then. 

Then through the 1940s, ’50s there were a lot of things which happened with at AT&T Bell Labs, et cetera, a lot of innovation, which was also a byproduct of the Cold War era. The Cold War era led to this competition, the Sputnik moment, and a lot of investment going into new areas of technological progress including on semiconductors. AT&T is one place which becomes the hub where a lot of these things come together, and it is able to convert a lot of the promise of transistors into reality.

When we talk about semiconductors, we are broadly just referring to the material used. As the name suggests, the conductivity is somewhere between a full insulator and a conductor. Conductor, the electrons, so many electrons will move at a short time that you cannot control, whereas in insulators, the electrons won’t move at all. Again, no use, but in a semiconductor, you can control it after modifying the properties of the particular semiconductor using something called doping. Once you can do that, you have this magical thing called the transistor, which is really a fundamental building block.

Just like the cell is the building block of the human body, this is the building block of our digital lives, is how I say. Once you can control the flow of electrons in one part of the circuit using current that is flowing in another part of the circuit, you essentially have a switch. It’s really a superpower because you can do a multiple number of things. You can control current in one part of the circuit to control voltages, or to control temperatures or to control a bunch of other things. You can use these transistors to amplify current, to store value, to produce signals—really wonderful things. That was the first step, but yet transistors, there were skeptics because transistors were made now, but the wires that are used to connect these transistors kept growing in number.

As the complexity grew, the number of wires would grow. Even the transistors were small, the number of wires going from one transistor to the other were huge. Then the next step of innovation came when people thought that, “Why can’t we put all these transistors with different functionalities into a same package?” That is what is essentially the integrated circuit. I’m simplifying it, but that was essentially one way it started that, “Can we pack all these things together?” That was a next step jump in this journey. Once ICs came into being, the next process was something called the planar process.

Until these ICs were made, but they were made manually, transistor by transistor. You can do it for 30 transistors, 40 transistors, but today, we have around 150 million transistors per millimeter square. Very difficult to do that in any meaningful way manually. The idea, it started there. Then the next step was planar process. What it did was you don’t need to do all these things step by step, but you do it in the form of layers. What they used to do is there’s a wafer on which you carve out these transistors. Layer by layer, you build this city across many, many chips on one single wafer. That was the next step jump.

Once planar process came into being, that is when by 1965, this observation came by Gordon Moore, who was back then at Fairchild Semiconductor. This observation, he made only for the chips which were being made at Fairchild Semiconductor. What he said, the number of transistors will roughly double every 18 months or so. By the way, he also said back in 1965 that ICs will lead to wonders such as home computers and automatic controls for automobiles, personal portable communications equipment, which we call a mobile phone. He made this observation when there were just 64 transistors in ICs.

Today, as I said, if you must be having an iPhone, Apple A14 chip has 134 million transistors per millimeter square. I am just baffled continuously by how it was a prediction. It was an observation which became a prediction, and that prediction became a law. It became a vision which people chased, and there was a lot of innovation, comparative advantages, et cetera. Not just the technical revolution, but also an economic revolution, which went in parallel. It became a vision which the world chased together, and it has been kept alive through all these various decades.

Rock’s Law

RAJAGOPALAN: I want to now move from the Moore’s Law, which is this doubling of transistors on a microchip every 18 months to two years, to what that leads to, which is a cost problem, which is what led us to Rock’s Law. This is the idea that, yes, people were going to use all these wonderful chips, semiconductors in every aspect of their life, you listed out a whole number of uses, even in 1965 people had a vision. But to actually get those many transistors and then keep doubling them, you needed production processes, fabrication processes that were really sophisticated, which needed very, very sophisticated and highly specific machines, which then required a lot of capital investment and so on.

That changes the way the semiconductor industry operates fundamentally. In your book, you initially described that it all starts with innovations at Bell Labs, it goes to Fairchild, it goes to Texas Instruments. There are a few American players in the business, but these American players are doing everything start to finish. They’re imagining the chip, their engineers are designing it. They have an in-house place where they actually fabricate, manufacture, assemble. I don’t even know if assembly was a thing at that point, but everything is in-house. Then that changes within a couple of decades. What happened there?

KOTHASTANE: Absolutely. What you refer to is what’s called the IDM, the integrated device manufacturer model. Even today, some memory chips do follow that process. For example, Samsung for its memories might do all of these things together by itself. When we are talking about chips today, we largely also talk about logic chips—your processors, et cetera, which are not memory chips, but they are complex logic doing a particular function. Once this idea began that you need to pack in more and more transistors—why do you want to pack in more transistors? Because you want to do more complex things.

Once you had that idea, then you started thinking, “Can I do the same thing at a higher speed, at a lower power?” Et cetera. For that, you don’t need to just increase the number of transistors, you also need to reduce the size of the transistor. Now, you started making things smaller and smaller. Once you try to do that, the cost of manufacturing increased significantly. Now, instead of 64 transistors, if you have to pack 6,400 transistors, first you will need a software to do that. You can’t do that by hand. Some designer will have to come and just focus on doing this really, really well-

RAJAGOPALAN: Like you.

KOTHASTANE: -and do all the verification testing. One of the many. You’ll start doing that. 

Once you do that, you need the manufacturing processes which can really carve out fine features on this silicon to be able to achieve that functionality. Today’s smallest feature size, we use five nanometers. It’s just a few atoms wide. You can imagine the complexity which is required. The machine which is used today, for example, for the five nanometer chips in the manufacturing, one of the machines used is for the lithography process, is by a company called ASML in Netherlands. It’s the size of a bus. It takes two aircrafts to transport it, cost $200 million a pop.

That complexity, it kept building on. That’s where we’ve reached today. Once these costs started rising of a company, let’s say, it had to pack in more transistors and make them smaller, it had to invest in manufacturing facility, and that manufacturing cost was not just incremental. You need to have an entire new process, new way of designing smaller transistors, new way of manufacturing them, new machines to get this functionality. 

The cost kept rising and that is what, as you refer to, is called the Rock’s Law. By the 1980s, this became very cumbersome for a new startup to enter. Just imagine there is a startup in Silicon Valley, wants to do this, but they are not going to make the billions of dollars of investment to be able to do that. That’s when an economic revolution took place. A former engineer from Texas Instruments who moved to Taiwan—TSMC.

RAJAGOPALAN: This is Morris Chang?

KOTHASTANE: Morris Chang. He and a few others got together, and they said to all the companies that, “We will promise you two things. One, we will be your contract manufacturer.” Whether you are AMDNVIDIA wasn’t there back then, but companies like NVIDIA, whoever are the companies which probably want to make chips, they told them, “You don’t need to worry about making your own chips. You make the design, give us those files and we will be your contract manufacturer.” The second promise they made is, we “Will never compete with you. We will not enter into design.” This is classic comparative advantage. The Taiwan story started then.

There were other factors also. This process is environmentally polluting back then. In the U.S., they were okay to let it go to another country, and Taiwan was ready to absorb that cost. That was another reason. Back then, even the value addition, which used to happen in the manufacturing segment, wasn’t very high. Again, the U.S. companies thought, “Why should I waste my effort, time, energy in doing something?” Classic how this comparative advantage-based specialization works. Taiwan started doing this from very low end. Value addition wasn’t much.

The only way you could make money was make billions and billions of chips for many, many customers, and you won’t make much per chip, but when you make billions of chips, you will make a profitable business. That’s how they started. Over 30, 40 years, they have built so much expertise and in that recipe of making chips, that actually today the value edition, which happens in manufacturing is more than the value edition which happens in design, slightly more. That’s a very counterintuitive result. Generally, we think manufacturing, not the big deal. It’s the design where there’s a lot of expertise, but that has slightly changed in the semiconductor space.

RAJAGOPALAN: Yes, because here also, the further innovation I realized is that manufacturing and assembly also used to be integrated at one point, and that got separated. Manufacturing became such a high-value proposition on its own requiring such specialized environmental requirements. When I say environmental, I don’t mean sustainability-wise. You needed an ability to take very large amounts of credit. You needed the appropriate capital environment to make that investment. And, of course, the ability to keep changing around your machinery, which will have something as specific as the rules that govern accounting and how you depreciate assets in the semiconductor manufacturing industry is quite different from virtually any other industry. There are such specific things that, past a point, Taiwan wasn’t even the cheapest assembly market. They got so rich that the assembly started happening in other places. What you’ve given us is a great summary. 

Geopolitics of the Semiconductor Industry

Now, I want to just drag us back to the present moment for a minute. After the pandemic, a couple of things have happened. One is over the last decade we’ve seen first a lukewarm war, and then an out-and-out beginnings of a cold war between U.S. and China. This has semiconductors at the core of it because semiconductors are not just used in our fridges and our mobile phones and our cars, but they’re also used in drones, in all your fighter jets, in missiles.

Anything that needs to be targeted with laser precision actually requires semiconductors, which we now know is how the U.S. won the Gulf War and started a new kind of arms race, which was a semiconductors race. This wasn’t the old arms race, which was a nuclear race. That’s one part of it. But the second part of it were the pandemic-related disruptions. This is the supply chain disruptions. Wuhan shuts down because the city shut down or is under lockdown because of COVID, you have a lot of things that are being assembled in Wuhan in terms of various electronics, the low-end chips and so on. Now, suddenly all of that comes to a standstill.

We experienced a brief glimpse of that when there was a huge earthquake followed by tsunami in Japan, but everyone was so focused on the nuclear reactor that we didn’t quite pay attention to this. The glimpse of this started almost 2010, 2011. Can you separate this geopolitics aspect of it, which is U.S. and China at the center of it, and the standard garden-variety globalization leads to too much division of labor and supply chain complexity, which now people are trying to figure out how to derisk or rollback?

KOTHASTANE: We just talked about Taiwan and U.S., but again, as you said, then China started picking up pace in assembly and then Malaysia, Singapore does a bit of it. Of course, Korea, Japan—and the Japan story is also very interesting. All those players started entering and then India, through the design space. Israel has started doing things, and Europe was also doing well on design. These were the main characters involved in the storage. What happened beyond that in recent times is, one, the U.S.-China rivalry, which is a structural rivalry, as we say in the geopolitics literature, that had to be fought on some domain.

Now, can it be done on as a hot war? Not possible, nuclear weapons, et cetera. Can it be done in the economic realm? That also is very difficult because, you know better than me, if there are tariffs, et cetera, the losses to America will also be significant. You can’t do a full decoupling at an economic level as well. You can do it for some things.

Then the idea is, which is the domain where this contestation can happen? Technology has come up as one new realm where all nation-states are thinking about this as one decisive advantage that they can gain over the other. They’re trying to build strength in that. Within the technology stack, if you look at China’s technology stack, one of the weaknesses is the semiconductors domain. They have done quite well in many of the other ends, but on the semiconductor side, they are still largely dependent on the U.S. and other countries for specialized gases from Japan, for the machines, you still depend on ASML in Netherlands, or Lam Research in U.S., et cetera. There are a lot of dependencies, and that’s why I thought U.S. chose this as the area, even within the broad technology space. Now, that is one.

The other reason, of course, is Taiwan. Because China-Taiwan relations are worsening, and even some of the most advanced chips that U.S. must be using for its military, let’s say, F-35 or others, would be finally manufactured in Taiwan. There was this other risk that what happens if China were to take over Taiwan for some other reason, not necessarily semiconductors, what happens then to these facilities?

That was another reason why, one, these export controls started, and second, a huge round of industrial policy measures have started where all nation-states are trying to localize a bigger part of this industry into their home borders. That unleashed a new thing. This is the geopolitics side. Also, in COVID-19, this became something that everyone faced. It was not just at the level of— 

RAJAGOPALAN: Not just U.S. and China.

KOTHASTANE: Everyone faced these shortages that were in the news. There were many cars shipped, which didn’t have some of the features because of these shortages. In fact, in India, there were lots of delays for car productions and many electronic productions as well. This reached the common-person level as well. That was the second reason. The third is also because semiconductors underlie a lot of future advances in technology. That also became a factor. For example, you are now seeing the news that OpenAI wants to build its own chips as well, and NVIDIA’s stocks have risen a lot, et cetera.

RAJAGOPALAN: It’s overtaken Google now. It’s the third-most valuable company in the world, which is extraordinary to me.

KOTHASTANE: Exactly. All these things came together at one point of time. That’s why I thought of conceptualizing this as a metacritical technology. There are three kinds of criticalities which come together at the same time. That’s why there is a whole lot of things happening in this space, including industrial policy and other things.

Metacritical Technologies

RAJAGOPALAN: You’ve touched on so many things, but I want to start by parsing out this idea of the metacritical technology. For instance, we were looking to buy a car because we had moved out of New York City and moved to the suburbs of the Washington, D.C. area. That’s not a very pedestrian-friendly place. There were lots of delays. We eventually ended up buying a used car because the order delay on the new cars was so long. When I started digging into it, I thought, “We’re not buying a Tesla. We don’t literally have a computer running the car, what’s going on?”

I realized that even that little button that raises and lowers the window of a car has some very relatively sophisticated semiconductor chips and transistors that are making that tiny action work. Which to me is magical because the first car that I learned to drive on didn’t have power steering, didn’t have power windows. You literally had to roll and unroll the windows. The first part I learned was the ubiquitous nature. It is in everything, not just our phones. A rice cooker has a chip, which helps us time when to start it and when to stop it, as opposed to standing on top of the stove waiting for your rice or dal to boil. That part, of course, makes it metacritical.

The second part is how much can we separate the everyday garden-variety chips from the very high-end sophisticated chips? Because that’s what the CHIPS Act has tried to do. This is the continuation of the China policy from Trump to Biden. Certain kinds of chips, which are very, very sophisticated, either technologically or can be used in military terms, are excluded. 

First question is, is that feasible? The second question is, doesn’t that go against the historical progression, which you partly described in the book? Which is that semiconductors became what they became because there were so many commercial everyday users.

That’s the reason you saw Moore’s Law. That’s the reason this industry took off, and this technology transformed the world in less than a century. One, can they be separated? Two, is that bad actually, as opposed to ideal when it comes to national interest?

KOTHASTANE: Yes. Like you rightly said, initially when this industry started, so the biggest customer or the only customer was the Apollo program. It absorbed most of the chips that were produced, and then it entered the missiles of the U.S. Armed Forces, et cetera. Slowly, as we discussed, the costs kept rising. The only way you can make that productive is when you’re selling it to large number of customers and which are not your militaries. Then a lot of advancements thereafter started happening on the commercial side. Today, the most advanced chips in terms of power and in terms of the speed, your iPhone will have the far more advanced chip.

Then it’ll be there in an army platform as well or a naval platform in the U.S. because their considerations will be about reliability. You should be able to operate at different temperatures. They are not so concerned about the fastest chip or the most advanced, but in the commercial space, we are. Even OpenAI or NVIDIA’s chips are being used for commercial purposes, not for military. This is what in the military literature we call spin-ons. Earlier you had spin-offs, but now you have spin-ons. Commercial technologies lead your advancements, and then they get carried over into military. Whereas it used to be exactly the opposite earlier.

That’s why commercial things are important, and all the things that happened in the semiconductor industry earlier were very useful for this progress. You essentially could do comparative advantage with specialization, and the things just worked so fine. You mentioned the ubiquitousness of this. The reason why we didn’t care about it is it was because the system just works so perfectly well. Why do you need to know the gory details of what’s there? Because it just works absolutely brilliantly, perfectly. Because of the three reasons that we discussed earlier, there has been this idea that now we need to think more about this resilience versus efficiency tradeoff.

The supply chain works efficiently very well. If there were these three conditions that are to come together about the geoeconomic, geopolitical and the technological imperatives, then there might be some problems. Hence, we need to build some more resilience. That’s the whole idea. Now, I have a lot of problems with that. Even I think, how will that happen?

RAJAGOPALAN: Can it be done?

KOTHASTANE: Now, I think it can be done, but it’ll come at a cost. All of us will have to pay the cost of higher prices. Maybe the speed of innovation will decline, et cetera. Those are the costs. Even if you see in the U.S.-China side, I don’t call it a war, U.S.-China chip war, because it’s not a war in my assessment. The decoupling between U.S. and China on chips is not at all the levels. To come to your question about is there a difference? There are no restrictions on U.S. companies to sell to China below, say, 28 nanometers or 12 nanometers, whatever. If your chips are less complex than what comes to below 12-nanometer chips, then they are free to sell.

If they are more complex, then there are restrictions which are coming in with an eye to, “Can they be used in military?” et cetera. Again, it’s very difficult to assess. Some of those things can also be used for commercial processes, but still they are trying to be restrictive to a particular level. Beyond that, the trade between U.S. and China on chips, it will go in other spheres because it is, again, very difficult to decouple. Chinese players need U.S. chips, and U.S. needs Chinese customers, companies. The idea is to dissociate at very specific levels. The State Department has kept moving the goalposts, trying to sharpen these export controls, et cetera.

It’s very difficult. I think because of the amount of money countries are putting in, probably you will have some even redundant capacity building in. When the semiconductor industry—as it is prone to its ups and downs, there are cycles—once you have a lot of manufacturing capacity built in and the demand doesn’t go at the same level, all these fabs which are going to come up in the next three, four years, they are going to face some problems. Then they will, again, demand another round of subsidies from their governments to keep them sustaining. We have entered that world actually.

RAJAGOPALAN: With the U.S. and China, surprisingly, I’m less concerned about that. Because as you mentioned the policy, I think this government has called it a small yard with high fences or something like that. The idea is that it’s a very small list of goods, which will absolutely be controlled. On everything else, we will keep things relatively open. 

The reason why I think it’s important is I am that housewife who wants a cheaper washing machine. For better or for worse, my washing machine uses relatively sophisticated chips and semiconductors. I would like my clothes clean and dry on time. Same thing with something as basic as my rice cooker or my coffee maker. These are literally things that save me time, and I deeply care about these things. 

Now, the attitude is not quite the same in other countries either because they don’t have such a large volume of trade with China, or they may already have other historical legacy problems with their trade policy like India. In those countries, my fear is right now, there are a lot more export controls that are being placed, and they are being placed to build resilience.

Are we going to end up in a world where you neither have resilience nor do you have innovation? The person who loses the most is not just your space program. Because eventually, they are willing to do things at a higher cost, military and space. They will figure out how to make their own manufacturing or fabrication units. It’s really going to be the housewife who can’t set her rice cooker and walk away to play with her kids or walk away to go to work and has to stand over the stove. That’s what’s happening in smaller countries. That’s what I’m just much more focused on, maybe because I’m not a geopolitical person. How do you think about how that’s playing out?

KOTHASTANE: I agree that costs will rise. I think a recent article by Gita Gopinath had tried to estimate some of the costs that are coming in because of these export controls and the loose decoupling that we are having. I think it put estimates to 5% of global GDP shrinkage if we take these things to the extreme. That’s a really significant one, and it is going to hurt obviously the developing countries more than the countries which can absorb some of these losses. These are real costs which are going to happen. I think the side to blame is also in the case that the fact that China also tried to do Made in China 2025.

That was the starting trigger for this whole idea about structural rivalry and the fact that, are we too dependent on China for our own good? Once that conversation started in a lot of countries, including in India, it’s very difficult to stop that. We keep making the argument that trade is positive. It’s okay if we are buying some low-value chips from China. How does it matter? It’s not as if they can engineer some specific thing, which will be used for espionage against you, it’s very difficult to do that. Even if they can, you can figure out technological solutions to that.

Once national security becomes the predominant frame through which you start seeing these issues, then many of our arguments take a back seat. We’ve entered that phase for good or bad, and that’s why every nation state is getting into this game in one way or the other.

Geographical Concentration

RAJAGOPALAN: Now that we’re talking about every nation-state is getting into this game, I want to go back to the earlier conversation, where you talked about there are three parts to the process: one is design, the second is manufacturing and the third is assembly. Now, when it comes to design, design started as a highly centralized thing in Silicon Valley, literally in two or three labs and companies. Actually, design, counterintuitively, has become the most decentralized aspect of it. Especially because of the economic innovations of separating the fab and foundry aspect from the design aspect.

Even a startup in an apartment in Koramangala can do its own design because it can license a whole bunch of software, it can actually ride and license on others’ patents and only make very, very specific design changes. It seems India is doing pretty well on this or at least Indians are doing pretty well on this. There’s a lot of design work that involves Indians. I think you estimate it as, I think, 90% of the chips in the world have at least Indians in one out of five processes or something like that. 20% of Indians are involved in 90% of the chip design in the world. That part seems straightforward.

Now, the second part is manufacturing. Manufacturing, which one would normally think is, “Oh, there are many countries that have manufacturing abilities.” We know even when it comes to textiles or shoes, we know China dominated for a long time. There are lots of countries—Vietnam, Taiwan, Bangladesh now—which started entering that world. That has not been the case for the semiconductor industry; that is just highly concentrated, especially in East Asia, and, of course, Taiwan dominates that world. Can you walk me through what is it about chip manufacturing that it has this bizarre geographical or economic concentration, and why the rest of the world can’t quite catch up?

Because when TSMC creates—they just built out a new fab in Arizona, it’s astounding $12 or $15 billion or something like that. It’s still going to be about 40% to 50%, more expensive than the same chip that’s produced in Taiwan, which itself is crazy. I recently read that, apparently, TSMC is not even considering Arizona as a serious venue because they’re moving things to Japan because of all the DEI and environmental and labor restrictions that have been worked into the CHIPS Act, and they’re not able to hire the talent they want. They flew in a few hundred Taiwanese to the fab and apparently, the local workers got quite mad about it. Now, they’re apparently moving everything to Japan. One, is this all true and why is all this happening?

KOTHASTANE: Yes. The central part about the Taiwan thing is that given that China-Taiwan relations are worsening, even TSMC would want to diversify and have an option in other places. Of course, U.S., Japan want to have that option that TSMC comes to other places. Even Indian government was wooing TSMC a long time. When all these things happened, it was, “Why would TSMC pick a place like India when other countries are giving more incentives and have experience in making chips?” That was the starting point. I don’t think U.S. and Japan are substitutes; it will happen in both, and U.S. taxpayers will pay for the chips made in America.

That is how it will go because U.S. is trying to woo TSMC, and U.S.-Taiwan political relations are important. This is no longer an act between two commercial parties, the governments are involved. U.S. government would have insisted that TSMC build some capacity in U.S. as well, even if it is doing so in Japan and other countries. The facilities will get built. After that, whether the chips will be economical enough or not or who buys a premium for that, that is really a problem. I fear that you might have some local sourcing mandates going forward.

RAJAGOPALAN: There are local sourcing mandates.

KOTHASTANE: Yes. That’s where the Atmanirbharta [self sufficiency] narrative will come there also. That’s the only way you can sell a chip which is made by the same company and is 40% costlier.

RAJAGOPALAN: I have a secondary concern there, if I can stop for a moment. My problem is not just that it’s 40% more expensive. Maybe there is reason to say that, “Hey, when it comes to certain tech, we’ll just use it for stuff we use for NASA, and the Pentagon and the U.S. military,” that the Arizona fab will be used just for that. For the chip that’s rolling down the windows in my car, or whatever is going in, I don’t know, every single electronic good that all Americans are consuming, we can continue fabricating that in China. Because even if there’s a pandemic or a war, of course, there are going to be disruptions, but it won’t be at the military level or something like that. U.S. is not hobbled. I can understand that.

That’s not how it’s being sold. It’s being sold as a place that will also produce commercial chips. My question is, who’s going to buy chips 40% more expensive? Because at the end of the day, we’re standing outside Walmart on Thanksgiving Day trying to get the cheapest electronics at the best price. That’s the part which doesn’t add up for me. Who’s actually going to buy this? How do we actually subsidize every single washing machine? It just doesn’t make sense.

KOTHASTANE: Yes, absolutely. Actually, the two things that you said are not necessarily substitutes, because even if you total all the military requirements of the U.S. Armed Forces, it will really be insignificant compared to the commercial demand, which needs to be produced by that chip to sustain itself as a viable economic entity. Really, military, yes, you can get some “Made in America” chips for U.S. military, but still for that fab to survive, you will need to do commercial manufacturing. Yes, the problem they will hit is who’s going to buy those chips.

The only way I foresee it might work is if you have local sourcing mandates, which is another big, big problem. Believe me, no government has thought through this. The narrative about national security is so strong that many of these questions aren’t being asked enough. The idea is that, “We need to do things to reduce dependence from China. Even if we have to pay a cost, we’ll pay it,” and many of the companies are hoping that governments will bring in local sourcing mandates. Earlier industry bodies in the semiconductor domain were singing paeans globalization and how this is a poster child for everything that we learn in trade theory. The same industry bodies are now— 

RAJAGOPALAN: Now on protectionism.

KOTHASTANE: Yes. We want to do things made at home, we’ll make more secure chips. That’s really the part of it. Coming back to your question on why things happened in East Asia. East Asia, again, it is diversified, there were chips being made in U.S. as well, and still U.S. has a significant—

RAJAGOPALAN: Micron.

KOTHASTANE: —manufacturing capacity. Just that it happens at the lower end, Europe did have—but again, costs kept rising. Just imagine, manufacturing and assembly are both labor intensive, and manufacturing is capital intensive. I said it had environmental costs earlier. All those led to geographies, which had two advantages where, one, assembly costs were lower, so labor costs were lower, and second, they were very globalized and connected with the U.S. ecosystem. That is East Asia. Again, Taiwan starting 1950s, ’60s, trade liberalization happened.

A lot of U.S. companies are very comfortable doing the transfer of technology to Taiwanese firms even back then. So is the case with Japan as well. Without FDI, they were able to get a lot of technology to these places, which was not the case with China, but for Taiwan, Japan, because of them being allies and in good relations with the U.S., that happened.

RAJAGOPALAN: China also drove its experts away. It was making its physicists turn into gardeners and farmers in rural areas, and the Cultural Revolution really just killed engineering and physics expertise in a way that it just decimated any chance of the semiconductor industry showing up there.

KOTHASTANE: Absolutely. No, the first IC in China was made before Taiwan got there. Again, it’s a very typical case of how government-run programs of this kind fail. In fact, we’ve tried to draw comparisons between USSR, China and India experience. The lessons are same in all of them. Taiwan, because of the global integration, the East Asian market really caught on to that. Other countries have tried, and to some limited success, like Israel, et cetera, have developed some expertise in manufacturing. This confluence of factors, good relationships with the U.S., integrated with the global market, low import tariffs, so on and so forth was there in East Asia and not in many other countries.

Zelenograd

RAJAGOPALAN: This is a good point to actually talk about what you just mentioned which is what happened in China, what happened in the Soviet Union, what happened in India, these are all high socialism. Varying levels of command and control where actually India comes out looking good in this comparison.

I want to start with the Soviet Union which is the most extraordinary example because at some point, they actually built a city from scratch which was supposed to mimic Silicon Valley and that environment of Bell Labs, Fairchild’s, Texas Instruments, and the innovation that took place. Let’s start with what happened with the Soviet Union. Then we’ll move on to India. Did this work? We know it didn’t work, but why didn’t it work?

KOTHASTANE: In classic USSR style, the idea was, they’ll have one city to make one kind of thing. There was one city, Zelenograd, which was supposed to make chips and it was supposed to be the chip capital of the USSR, but today USSR has no commercial manufacturing facilities, much like India. Obviously, that failed. A lot of things they tried to do, but from a very different angle compared to the U.S.

Again, in the U.S. the idea was it happened far away from Washington, D.C., it happened in California. A lot of factors going into that. Whereas in the USSR, it was a government-directed program, largely looking at, what can we do for the military side? Where can we go ahead? All these projects have a common feature because they are government run. They do have a quick beginning, the government is able to put in money, things start working.

As you go on to the next step, capital costs will keep rising because of Rock’s Law. You need to be able to make a chip which will go to commercial players to recoup the costs that you’re putting in on the capital. Why would a government firm do that?

RAJAGOPALAN: They couldn’t even get the amount of soap, light bulbs and toothpaste they needed because of the calculation problem. The idea that they will improve chips because it’s going in microwaves and washing machine is ludicrous, right?

KOTHASTANE: Absolutely. All those things didn’t happen. The cutting edge of design also requires a large number of engineers from across the world coming in and trying to do this kind of innovation and new ideas. That wasn’t the case in USSR as well. A lot of their initial projects were trying to replicate what were the things that were happening in the U.S. It largely became a copycat model and trying to build things. Then there were a lot of ideas about espionage, et cetera. It went into a very different realm which was not conducive to quick innovation and cycles of repeated new things.

RAJAGOPALAN: The worst thing is trade policy because you have to import so much globally. Not just the raw materials and the materials with which you dope and rare minerals and so on, but also simple things like the nuts and bolts for a very expensive fabrication machine which are only available in certain places. If you have across-the-board crazy tariffs, but you only liberalize them for military use, suddenly none of it works. You need a very flat open trade and tariff regime for even one chip to be built efficiently is what I’ve learned really from the Soviet Union example.

KOTHASTANE: Exactly. That is the same lesson for India and China as well because we had such—in India’s case, because of limited foreign exchange. There was again the idea that foreign exchange is so precious that anything imported is big evil sin done. The idea was you preserve that, and throughout ’70s, ’80s, even when companies were trying to do a few things, when they had to import manufacturing equipment, they couldn’t. By the time they got their licenses and they made that chip that— 

RAJAGOPALAN: Moore’s Law had kicked in.

KOTHASTANE: You could just buy—

RAJAGOPALAN: fifteen years too old.

KOTHASTANE: The government companies themselves bought chips from outside. Why would you buy 50 times costlier chip which comes 10 years later?

RAJAGOPALAN: I want to come to India. India is a slightly different case because it seems like, unlike China, which literally killed sometimes its own engineering talent, India’s engineering science talent was pretty good. From what I understand from your book, the early stages, even the tech transfers were workable, right? Philips actually did collaborate with India, as did the Americans.

The same kinds of collaborations that Taiwan took advantage of, they were offered in some shape or form to India. India had the right talent, but India also had all the other things that you mentioned. It had the cheap labor that you need for a large assembly. It had spaces where you could think about what do we do about environmental pollution or look the other way because happens to be much larger country than Taiwan.

You can seclude pockets where such industrial activity takes place, but it never takes off in India. Can you tell me what are the reasons for that before, say liberalization, the very autarchy-related examples that you were talking about a minute ago? Then we can talk through what happened post liberalization and why it still didn’t take off.

KOTHASTANE: Initially, the difference between Taiwan and the Indian idea was a lot of the experiments from the Indian side were done by government companies for government products. We call that first version by the government and for the government. The idea was there will be two government companies which will make chips. Obviously, back then the demand was also largely in the government. It was never thought of that you need private investment. Maybe the capacity also wasn’t there, but the idea was, only government companies will do this.

BEL, Bharat Electronics Limited, started making transistors in the 1950s. It wasn’t that bad. Was doing reasonably well. You said there was technology transfer also coming in; Philips and Radio Corporation of America, RCA, which did tech transfer in India as well. That was the same company which did tech transfers to Taiwan as well.

Tech transfers did happen initially, but you, again, entered that cycle where companies were able to make these chips initially, but then you had this problem because these are government-run firms, the government then started saying—two firms were doing, Semiconductor Complex Limited, SCL Chandigarh and BEL. Then the government said, “Hey, why are two firms doing the same thing? Competition is—”

RAJAGOPALAN: Wastage.

KOTHASTANE: “—not good.” Yes, it is wastage. “Why is this being done?” Then they said, “Okay, SCL, you do chips, and BEL, you only do the integration.” You kill competition. There’s no incentive for SCL to do anything new. That was the first reason.

The second reason was on the trade side that we talked about already. The third reason was, of course, that capital cost kept increasing and the government said, “I can’t pump in two times the amount every four years. I’m not going to do that.” Hence, these companies started actually falling behind the cutting edge slowly and really couldn’t do much. SCL still exists today, but largely in 2004 or ’05 it was taken over by ISRO. 

RAJAGOPALAN: Only for ISRO, right, basically?

KOTHASTANE: —by Department of Space, and it largely supplies to the space. Recently, this last year, it has been moved to MeitY, the Ministry of Electronics and Information Technology. The aim is that they want to convert it into a private commercial fab probably down the line, but that is the phase. Again, these three figures—the same thing happened in China as well. Pretty much the same things.

RAJAGOPALAN: Plus the Cultural Revolution. They did one more terrible thing in addition to India.

KOTHASTANE: Many people would’ve gone to Taiwan because of it as well. 

RAJAGOPALAN: And Hong Kong, right? There was an exodus to Hong Kong and Taiwan and so on. Coming back to India, the way I thought about it when I was trying to read the parallels between what happened in the U.S. and Taiwan and what happened in India, it seems like, one, of course, like you said, the government socialist command and control hates wastage because they don’t understand quite the point of it.

There were two other things. One, because of the autarkic nature, India was not an export country. If your domestic market is limited, when Taiwan started on this journey, its domestic market wasn’t rich. People were poor, they were in the countryside. They weren’t actually demanding the microwaves and the washing machines that the chips were being made for, but they were a huge export economy, which means they were supplying to the rest of the world.

As their people got richer, they were also supplying to domestic firms. In India, we were never export-led because we were autarchic, and because we were autarchic, we had foreign exchange control problems so we couldn’t import. The second aspect of it was the Licence-Permit Raj, which also is—trade and tariffs, they are a tax on the foreign producer and the domestic consumer. That’s how I think about it.

Whereas Licence-Permit Raj is a tax on the domestic producer and the domestic consumer. If you have Licence-Permit Raj and you won’t give Bajaj scooters or your Ambassador cars and so on an expansion permit to serve the domestic market, and you kill all competition in the domestic market, then you also won’t have innovation in the domestic market. You’re not going to have people—we could have had Indian-made cars that needed advanced semiconductor chips to roll the window up and down, or everyone in India cooks rice, so you could have had an electric rice cooker so early in India, but the Licence-Permit Raj kills that side of it.

It’s so many own goals, it’s not just one thing. My sense is the government-led firm for the government and by the government is just the starting of the story. Even when you liberalize that, even if they had sold and when they were willing to sell, SCL and BEL, there are no buyers for it because no one wants to deal with any of this even by the early ’90s. 

Now with liberalization, we have some liberalization on the trade side. We kill the Licence-Permit Raj. A domestic market starts opening up. Why doesn’t the semiconductor industry take off at that point in India? Because now it looks like this could happen.

Unease of doing Business in India

KOTHASTANE: The second phase, we call it the unease of doing business in India. That was the phase between 2005—2005, there were a few attempts done. There were a few attempts done in 2013, ’14 and then in 2015. All these we’ve classified together as a second set of attempts.

The idea again, there was, one, we should start from some assembly and then we will get to fabs. A few projects were tried. By that time, a couple of things happened. One, the idea was that the fab market itself was consolidating. Because of Rock’s Law, costs were increasing. Companies like AMD, et cetera, which were interested in trying to do things, they themselves became fab-less. They said, “We can’t run this foundry business; it’s too costly for us.”

The entire consolidation phase was happening. Globally, there were headwinds. Second, in India itself, there were things like retrospective taxation that we were doing in 2012, et cetera. Just imagine, for a manufacturing facility to come up, even if you are going to put money today, the first chip that will come out is five years down the line, and maybe it’ll come into a product six years down the line.

RAJAGOPALAN: The  regime uncertainty is a killer.

KOTHASTANE: Yes. How can you put billions of dollars of investment for something which is so uncertain? That policy uncertainty was something which really hurt. That was the major reason in our assessment. The third reason, of course, was that the domestic demand was not that great. GDP per capita was not that high. India doesn’t naturally think of exports like you were saying. The idea still was that can we serve the domestic market? Domestic market wasn’t huge. Why should you do this in India and not Malaysia, which already is well integrated into the international market already?

That’s why India lost out on these factors. None of the investors which showed interest went ahead with this. All of these were at the MOU stage, memorandum of understanding, never made it to anything more significant than that. All the multiple projects just died down because largely of the policy uncertainty, tax uncertainty, high tax environment and the trade policy because we still did have import tariffs which made it difficult for companies to do business here. All these.

RAJAGOPALAN: This is an odd thing. We get this insight later from Krugman, but our idea is in our heads. When we study textbook economics, that if you are producing semiconductor chips, it’s the law of comparative advantage, like how Ricardo wrote about port and cloth. You are producing one thing and you’re deeply specializing in one thing. You’ll sell that one thing and you’ll buy everything else.

What we now realize is that Taiwan is both the biggest exporter of semiconductor chips, but also the biggest importer of semiconductor chips. This tends to happen because of diversification. Like I said, this was the big insight from Krugman that the kinds of things we end up importing largely mirror the kinds of things we are exporting because of specialization and diversification, which happens across the board in the same economy.

The big problem with this kind of government involvement, especially in modern-day post-liberalization India, is it comes with protectionism. That’s the guarantee. We will protect you from foreign competition and chips and the chip manufacturers and everyone else who’s associated with them, like, say, Apple or any big electronics, they’re like, “No, you don’t understand.” We need lots of different kinds of chips. You can’t say we will be a chip exporter and also have an electronics industry that’s thriving or a car industry that’s thriving because they’re going to import chips.

This I think is like such a fundamental Econ 101 insight that no one in government seems to have understood or cracked. It’s just like at home, the homes of people which have fantastic cooks, they’re also people who are foodies. They end up getting a lot of great food diversified from all over the place, but they also end up cooking up a storm at home and having lots of different kinds of cuisines.

If you are a foodie, that’s how you think about the world. If you are a foodie equivalent, if you are a chippy, you specialize in semiconductor chips, you want to buy lots of chips, you’re also going to sell lots of chips. This doesn’t seem to get processed. Why is this insight so difficult to crack?

KOTHASTANE: You don’t have to tell me that. I am myself completely flummoxed by this, but just one minor thing. Taiwan is not the biggest importer of chips, but their biggest import is also chips.

RAJAGOPALAN: Sorry.

KOTHASTANE: It’s like their biggest export.

RAJAGOPALAN: Exactly. Sorry, yes, that’s a very good correction. Their biggest import and their biggest export are chips.

KOTHASTANE: Oh, yes. That’s one, but I think in the Indian side, with the trade, we have this relationship. We always want to do both industrial policy and tariffs and trade restrictions simultaneously. That’s what we are doing even in this current phase. The whole idea is we will have the phased regime of where we’ll keep increasing the import tariffs on different components such that over time all of these things will happen in India. That’s what we did in electronics as well.

I do see that now because a lot of the people who are investing are the MNCs, which do rely on a lot of these transfers, they have been pushing for the governments to change this idea. In fact, the ministry itself released a report a year ago, which said that, we should first globalize and then try to localize instead of trying to do both simultaneously, which has been our approach since last 30, 40 years. Both we need to do simultaneously. We will globalize by giving some money to companies to invest here. We will localize by just putting tariffs and some restrictions.

RAJAGOPALAN: Has this report managed to cross the street and enter the Commerce Ministry at all? Because the news doesn’t seem to have reached them.

KOTHASTANE: No. At least last month or so for example, the tariff rates were reduced.

RAJAGOPALAN: After a big U-turn before that.

KOTHASTANE: We also did things like the laptop “monitoring” regime. Then they went back on that.

RAJAGOPALAN: That’s the regime uncertainty part. It’s cuckoo, right?

KOTHASTANE: Yes. Absolutely. The fact that what we say about trade and this industry itself has operated in a way that you need to import a lot of things to export is something which is really central to it, and it has escaped the government thinking. Now, I guess that point has been made by enough players and the government is changing policy.

At least I can see tangible changes in terms of reducing tariffs, acknowledgment that we can’t become independent of Chinese chips all of a sudden. In fact, 64% of the chips consumed in India come from China, Hong Kong, Macau combined. There is a realization now that you need some of this to happen and you don’t want to shut it down all of a sudden. We might see the repeat of this two, three years down the line. I don’t have that confidence.

RAJAGOPALAN: It’s one of those—it just keeps coming back. When I wear my economics educator hat, I’m like, “All of us should just die in shame because we have not managed to communicate this basic insight.” On the other hand, I’m like, “It seems like enough people have talked about it since the late ’80s, enough government reports seem to have these ideas, but there’s something about this command-and-control nature of bureaucrats and how they like to do things by diktat and notification.

I think all the three issues that you mentioned are in some sense integrated. The government likes pet projects because that’s what politicians like. Bureaucrats like to be able to sign off on things and undo the signing off on things when there are protests or when circumstances change. That general idea that you need rule of law, you shouldn’t have highly centralized control, you should not have lots of exceptions when it comes to your tariff or tax policy, like GST should be one or two rates, it should not be eight rates.

I think this is the Indian DNA, which is why anytime I become slightly hopeful when I read good news, I’m like, “No, this good news came because we have a chip engineer who is also the minister at the moment.” The moment that goes away, it’s going to be ruled by the typical command-and-control bureaucrats, and then everything will go to hell. That’s my reason for pessimism. You’re laughing.

KOTHASTANE: Yes, I know that. That scares me. Hopefully, the point has been made by enough people and companies. Some of these are irreversible, in a sense, once the companies have invested. For example—

RAJAGOPALAN: Enough skin in the game, yes.

KOTHASTANE: —for Apple, they have already thought that—many of the Apple suppliers are Chinese, so you need to get them in. There have been exemptions made for these companies. I think that idea is dawning upon a lot more people now.

Industrial Targeting in Taiwan

RAJAGOPALAN: One of the biggest success story examples of industrial targeting is TSMC. The idea is actually, it went up to the president level, a red carpet was rolled out for Morris Chang. The government put in money and gave a bunch of subsidies, but phone calls were made by the Taiwanese government to a number of industrialists saying, “Hey, you got a lot of help from us many years ago. It’s time to pay back, and you all have to put in money into TSMC,” and so on and so forth.

When we say industrial targeting, industrial planning doesn’t work, TSMC is the poster child that it does. First question is, is it the poster child for industrial planning, or were there other factors? Second, what is the appropriate role of government which led to success in the case of TSMC? Is it at all replicable?

KOTHASTANE: Tough question if we can replicate it; it’s always multicausal. Even in this case, the fact that you had someone like a Morris Chang who did rise to the top level of Texas Instruments and then wanted to come back here, also, the fact that U.S.-Taiwan relations were great, technology transfers could happen, the fact that Taiwan could export to the richer world back then, and the fact that Taiwan didn’t do a command-and-control approach.

They were financing then, but there were multiple companies which were doing manufacturing. There is TSMC, but Taiwan also has PSMC, there is UMC and there were many others as well. The approach was not necessarily that manufacturing has to be done by government, but the finance part did come from government. A lot of backing was given. If you think of finance, produce, regulate as the three functions of the government, they were not into the production side of it. They were into the finance side or the regulate side.

RAJAGOPALAN: Even on the finance side, there were enough other private players who had to put in money and who would lose their investment if TSMC didn’t work.

KOTHASTANE: Yes, and some of them were foreign investors as well. All those things were well designed by Taiwan. Many countries tried to do industrial targeting, and even Taiwan would have tried for many companies, but the confluence of factors for TSMC worked a lot better than the other countries.

China also tried. There’s Richard Chang, who comes back again from the U.S. By that time, obviously, China doesn’t get transfer of technology from the U.S., so it is a lot through FDI and a lot of engineers from Taiwan going to China as well. That’s how they build their industry in their second phase. A lot of funding for this, again, comes from foreign investors plus the government. All of these have to fall in place. It’s not just one of them.

RAJAGOPALAN: To me, the differentiating aspect is that despite every red carpet rolled out by the government, at the end of the day, the East Asian countries demand that their producers meet the global market test. Whereas the Soviet, Indian and erstwhile Chinese, of course not the new modern-day Chinese, but they not only had industrial planning or targeting, but they also protected their—the same firms that were given the special treatment were also protected from the global test.

In a sense, the bad companies never perished and we never knew which are the good companies that actually benefited from government subsidies or certain schemes or certain credit benefits and tax exemptions, and so on. To me, that seems to be the incredible thing about South Korea and Taiwan, that we’ll give you everything, but if you don’t meet the global market test when it comes to price, innovation, efficiency, and you fail that test, it’s over for you.

Not only will these benefits be taken away, you will be completely out of the loop and the government will not give you any—especially in the case of South Korea, they were pretty brutal with failed entrepreneurs.

KOTHASTANE: Yes. Letting losers go is equally important. I somehow think, and this is not a well-founded idea, but I think, is there a curse of a big-sized economy? Once you have a bigger economy, then you have industries and domestic players who settle to this low-level equilibrium of saying that, “I can just try to service the domestic market. Why should I even think of being the world’s best?” Whereas for South Korea and the smaller countries, that was never an option. You have to target the wider world to be able to do something significant.

Maybe with India and China, there is this idea that “Let’s serve the domestic market, it’s big enough.” Also knowing fully that you can stop the foreign players from coming in through some protectionist measure. Just put some restrictions, and then serve the domestic market and be happy with it. Many times we fall into that low-level equilibrium is what I think.  

RAJAGOPALAN: Yes and no. I’ll tell you what I mean by yes. I agree with your observation that big economies do suffer from this kind of thinking. What I mean by no is they can’t get away with it. It’s not like India got rich, despite behaving like this. The fact of the matter is, India did not get rich. India was richer than South Korea in 1955 when South Korea was the second- or third-poorest country in the world. It was where Syria and Yemen are today.

It was devastated by civil war. It had no natural resources. In the early ’60s and late ’60s, it was exporting human hair as a large chunk of its exports. That’s how bad it was. When India was making semiconductors, that’s what South Korea was doing. South Korea got rich. India, by doing this, did not get rich. There are a lot of small economies across Africa, across Latin America, which are poor because they didn’t do what South Korea, Malaysia and Taiwan did.

I don’t think the difference is just one of big versus small economies. I think large economies have more arrogance. That’s, I think, the right observation. I don’t think large economies can get away with it at all. China could not get away with it until it turned into an export-oriented economy and passed the global market test, which India frankly has still not passed when it comes to manufacturing. It’s only passed it when it comes to services in some reasonable level.

KOTHASTANE: Yes, I agree that the countries can’t get over it, but some conglomerates in that country can.

RAJAGOPALAN: Yes, exactly. You can have large car manufacturers, for instance, in India, which never innovate, which will have like 200%, 180% tariffs, which will say, “Okay, we’ll put in an FDI plant somewhere in Tamil Nadu, and we just continue to manufacture the same old car over and over again.” When I come to India, I see people, I’m thrilled that people are excited about getting new cars, but those cars are like 30-year-old models in the United States, and at a higher price actually in India than what they would sell for a dollar value, not even purchasing power value.

Those things, they’re getting away with it. In a world of Instagram and YouTube where everyone can see what’s out there, how long they can keep this going, this arrogance going is also a little bit unclear to me. 

The last question I had on this Taiwanese domination is how important are individuals in this story? Because in your book, when you are writing about the geopolitics, the geoeconomics, you also weave in all the interesting characters; the people who made the early innovations, the people who won the Nobel Prize. The people who then started these companies who are—the way Morris Chang actually comes out of retirement when he thinks TSMC is going the wrong way and not investing enough of the global financial crisis. There seemed to be some very important people in the mix. How important do you think is the actual individual at the helm of these things?

KOTHASTANE: A lot of people say that initial conditions matter for institutions. One of the initial conditions that matters is leadership. Many of the people who were in the right place at the right time with the right skills do matter. Even at a metalevel, individuals, as in not a specific person, but the role of people and talent is very important in this industry.

Just the fact that TSMC’s magic is not so much in the machines that they are using, but they have perfected a recipe and they have the processes that have allowed them to do the same thing, which technically other companies can also do, but they’re able to do the higher yields with lower wastage, greater efficiency. The advantage of TSMC and their strategic advantage is in the fact that their engineers are able to do that and their leadership is able to do that, is able to work with Netherlands firm. It has been able to work with a Japanese firm and make this perfect recipe.

That’s why I often think of even if China were to take over Taiwan for some other reason, but TSMC will still not be workable. In fact, the TSMC CEO has been on record saying this because they think their strategic advantages in terms of what their people bring to the table and working with the world, which will be difficult for many other countries or companies to replicate.

RAJAGOPALAN: I was having an interesting conversation. Actually, I don’t want to say his name in case there are issues, he’s one of the long-time listeners of my podcast and works at ASML. I was asking him about this idea of the [Indian] government trying to get a semiconductor industry going from scratch and so on. You know this, ASML is one of the most specialized firms in the world.

He also reinforced the fact that nowadays when we think of tech transfers, it’s not just about patents and so on. It’s actually about the people. The ASML, the process they had that they have to run, let’s say over 24 or 48 hours, you need two or three very specific engineers to be available in the room during that time without which that process can’t be conducted.

That is the transfer, which is so difficult when it comes to tech transfer as opposed to just handing over blueprints or having a licensing arrangement that’s friendly in terms of patents or designs or something like that. I think what makes me sad is there are a lot of Indian engineers, Indian-born, Indian-trained engineers who actually end up going to Eindhoven and working at places like ASML.

We do manage to produce the human capital. India is one of those bizarre countries. We’re in two, three centuries at the same time. We also have the ASML engineer capability in large enough numbers, and we also have enough people who are leaving farming and part of the structural transformation of coming to get an assembly line job in the city or in a factory and so on.

We could actually be part of the entire process, start to finish, not in the same state. You’ll probably have the ASML and the design people in Bangalore. You’ll probably have an industry somewhere in interior Karnataka or interior Andhra Pradesh, and then you’ll have an assembly line somewhere in MP or UP if they ever get their act together. We could do this. What saddens me about this is India is large enough, it has the talent at every level. It has the cheap labor supply, it has the high—it has every part of the mix, and it’s just terrible policies, regime uncertainty and poor infrastructure, which have stalled the entire thing.

KOTHASTANE: That was the trouble. Now recent policies because of that, we have some initial wins. Again, the cost numbers, et cetera, we have to see whether they are able to, again, make chips economically, which others will buy. That all remains. Now on the assembly side, there are three or four projects which are starting up, including Micron.

There’s one fab which has been announced just a couple of weeks ago which is coming up in Gujarat. Now, the focus of the government is shifting to design. I think that should have been our primary focus. Nevertheless, if we also get our act together on that, that’s the quickest way because as you said, the talent does exist in many places, but there are difficulties for a firm to prototype to get things out quickly.

The gestation periods are much longer compared to software. Also, we don’t have the VC ecosystem like we have in software for this hardware. My mental model is Indian hardware design firms are 10 years behind where Indian software firms were. We’ll get hardware products as well. It’ll take some time. For that phase to happen, a lot more things need to happen on the design side. Hopefully, if that happens, we will have some ecosystem. Again, manufacturing, you have to run the 40-year race. It’s a marathon which—

RAJAGOPALAN: You have to liberalize everything. That’s the thing about manufacturing. You can’t have special exceptions. It may work for a one-off thing, but for a manufacturing world to sustain itself, only if it’s broad-based across the board, will you get better infrastructure, will you get better transportation, will you get better usage of ports. Everything is interconnected. It’s not just about creating a special SEZ within a special corridor where you put a special plant that is treated specially by the special government.

KOTHASTANE: Even labor reforms. For example, the Taiwanese firms are used to two shifts of 12 hours each. When they wanted to set assembly for plants, one of their demands has been that you want to do the same thing because we are used to doing this. Now, we have to do three shifts. It’s a big cost for them. None of our states, as you know, has done that 12-hour shift. 

RAJAGOPALAN: This is not semiconductors. The same problem with Apple. Everyone is used to the Chinese and the Taiwanese and the South Korean model working, which is the 12-hour shift. If there were three or four things that you could say absolutely need to be done before we can even think of emerging as an emerging player in the semiconductor industry, what would that be?

KOTHASTANE: It would be different across the different segments. At a metalevel, we already discussed, we need to get the—

RAJAGOPALAN: Let’s start with assembly because I think that’s where there are actually some green shoots in India right now.

KOTHASTANE: Yes. Assembly, as I said, first the trade policy, import tariffs generally need to go down, you’ll need to import wafers to finally make them into packaged ICs. Import tariffs need to go down there. Tax policies as well. Even with GST, there has been an improvement, but further reductions will generally help across the board, not just for semiconductor industry. The third one, as I said, labor law reforms are also important. Better infrastructure is also a part of that. These industries also require water, electricity. For example, electricity is not under GST, so the cost disabilities add up.

RAJAGOPALAN: Our DISCOMs are in terrible shape and our manufacturing, there’s no stability. Each manufacturing unit actually has to generate its own power to even be functional.

KOTHASTANE: Right. The captive power plants, et cetera. Electricity, bringing under GST would be one way out. These are the kinds of things which will help, not just the semiconductor industry, but—

RAJAGOPALAN: It’ll help everything.

KOTHASTANE: —many others. That is one. On the manufacturing side, this is one segment where every country in the world, the government has put in money. Even though we would like that this become a completely private-run enterprise, but manufacturing semiconductors with the kind of capital cost required, governments have financed across, so the finance part is not a problem in India now. The government has backed at a considerable cost.

Again, the idea is, will the industry be able to make chips economically? Now, that requires, again, to be able to run the business efficiently, so again, infrastructure matters, the electricity costs matter, the water.

RAJAGOPALAN: Regime uncertainty, no demonetization, no Vodafone, no crazy—no laptop bans. We really need to work on a very certain and stable economic rules environment.

KOTHASTANE: Yes. That’s one hope. I’m seeing it this way because the government is so invested in it, and the fact that now there is a big project coming in, I think it might prevent it from doing some of the things that you mentioned because there is now skin in the game. The project is happening in—much like what Montek Singh Ahluwalia wrote on the Maruti thing, because it was a personal project of the government, a lot of things were—the pathway for it was cleared compared to for other industries. Here also, Dholera is a big project for the PM, the current PM. He wants to see this happening. Maybe a lot of—

RAJAGOPALAN: It’s exactly the same reason I’m pessimistic because what happens is, with all the government backing, you feel like you have the government so invested. Now you can start asking for protectionism when you actually don’t do well economically. Then there is a tendency of, “Oh, this is our project, it can’t fail,” because losing in the market test or failing the market test is a big sort of egg on face for the government, so you start putting in protectionism, and you say, “No, now only domestic chips can be bought. Now, all the cars produced in India will only buy the chips produced in Dholera,” and that’s how that goes. I see both reasons. I see reasons for optimism, but I’m also quite nervous about the Maruti story playing out again.

KOTHASTANE: Yes, I absolutely agree. That we’ll come to know five years later when the chips come out, but absolutely, a very possible thing which can happen. Finally, in design, I think the idea is—

RAJAGOPALAN: We have an edge.

KOTHASTANE: —you have a lot of people who can do the products, but the idea is now the government is again—the costs come in terms of the EDA tool, the software license costs, where the government is trying to help out, but the idea again there is there are very strict restrictions on ownership of the intellectual property on whether the company is “Indian” or not.

Again, our argument has been don’t bother about it as long as the firms are doing work in India. If you have to make things easier for them, just do that irrespective of whether the ownership is fully Indian or not. Because it’s quite natural that these companies will have investors in the U.S. or other places because the VC ecosystem doesn’t exist in India as of now. Those are the—

RAJAGOPALAN: That’s part of the tech transfer. That’s part of the knowledge transfer. Other companies in other countries putting in the money brings with it that kind of expertise and skin in the game. There, I think, the biggest reform will be judicial. We need better enforcement of contracts, quick enforcement of licensing arrangements, actually upholding patents and not doing the sketchy things China did.

These are the things we need to focus on. I think the big thing I took away from the book is all the reforms one needs to do to become a leader in semiconductors and protect national interest and to build resilience, you need to do for all broad-based manufacturing and export-led economy anyway, so that’s really the way to go.

KOTHASTANE: Absolutely. We tend to get focused on the specifics of that industry, but the more you see that, the more these lessons become clearer, that it’s not something one policy or one great idea, which can change things. But you need to do pro-market reforms, which help every industry.

RAJAGOPALAN: This was really amazing. I was hoping to also talk about “We, the Citizen,” but now I have the perfect excuse to have you come back, maybe with your co-authors, Khyati, I haven’t had a chance to speak with her yet. We can all have a conversation. Anupam also, I don’t think we’ve had him yet on the podcast, so it would be lovely for you to come back with your other co-authors and then we can talk about that book. But thank you so much for doing this. This was a pleasure.

KOTHASTANE: Thanks, Shruti. Thanks for having me.

About Ideas of India

Host Shruti Rajagopalan examines the academic ideas that can propel India forward. Subscribe in your favorite podcast app