China is both a model and a warning for India's chip ambitions: US professor on what New Delhi should do

India has entered the semiconductor race with an ambitious plan to build a domestic chip ecosystem, but China's experience offers both a roadmap and a warning, says Professor Rakesh Kumar, author of The Chip Age.

China has spent tens of billions of dollars building capabilities across the semiconductor supply chain, yet it still cannot make the most advanced chips.

"China serves both as a good role model as well as a cautionary tale," Kumar said in an exclusive interview with Business Today. "They've spent so much money, but in spite of this, all it takes is for the US or the Netherlands to stop shipment of some chip or some chipmaking equipment, and then you're stuck."

For India, Kumar, a Professor at the University of Illinois, argues that the lesson is to be more strategic about where it puts its money - strengthening chip design, building expertise in packaging and testing, and developing capacity in legacy chips rather than chasing advanced manufacturing too soon.

EDITED EXCERPTS

You call the book Chip Age. Why the Chip Age? And why is it that only a few countries dominate this whole chip-making process?

I call the book Chip Age because, since chips were invented in the 1950s, they have had an enormous impact on society. These are tiny brains that undergird even basic systems - computers, telephones, and so on. The birth of the internet and the revolution in computing depended critically on chips.

Similarly, chips have had an enormous impact on healthcare, GPS navigation, aircraft, spacecraft, and missiles - all of them have chips in them. Chips are increasingly at the centre of politics, geopolitics, the economy, and planetary health.

Some of the most strategic technologies in the world, such as artificial intelligence, quantum computing, renewable energy, and biotechnology, also depend critically on chips. So it really is the Chip Age.

Countries dominate different parts of this supply chain for very different reasons. Sometimes it is history and the first mover advantage. For example, transistors were invented in the US, and because chip progress depends critically on transistor scaling, the US came to dominate, at least initially, the chip industry.

That led to the creation of the right research, talent, and education ecosystem. The US continues to have some of the world's best universities in this area. So chip design is dominated by the US.

On the other hand, chip manufacturing is highly capital-intensive and requires state support and subsidies. It also requires low input costs - cheaper materials, cheaper labour, and loose regulations. It is easier to do those things in Asia.

But then there are other reasons related to where something is found naturally. For example, you need neon for chip-making, and a lot of neon in the world comes from Ukraine and Russia. You also need a lot of copper. China now dominates processed copper, rare earths, and certain other elements.

When South Koreans wanted to get into the memory industry, the Japanese were dominating it. Japan was so formidable in the memory industry that the US and Japan started having exactly the kind of tussle that the US and China are having right now.

The US started imposing different kinds of trade sanctions on Japan, and the US directly and indirectly started propping up the South Korean memory chip industry as a counterweight to the Japanese industry. So, there are a multitude of factors because of which different countries end-up being dominant.

Consider Taiwan. While many other countries have low costs, Taiwan made a sustained effort in supporting manufacturing. And manufacturing is one of those things where the more you do, the better you get at it. And once you get better at it, you can become more efficient and lower the cost even more.

Take ASML. Why does this company in the Netherlands dominate the advanced chipmaking equipment market? Their foundational technology was developed in the US, but nobody was ready to use it. ASML was part of a US-based consortium. The US government and some American semiconductor companies propped up ASML. They licensed key technologies to ASML and asked it to develop them further. US companies also invested heavily into ASML. And now here we are with ASML.

The US seems to be a common thread in the rise of every major Asian semiconductor power. Japan initially built its chip ecosystem on American technology; Taiwan's chip industry began with technology transfer from the US. Is it fair to say that no major chip ecosystem has emerged without significant help from the US?

This has been more or less true for almost every industry for more than a century now. The US is a highly innovative country with great universities and a strong entrepreneurial ecosystem. And those things feed each other. It creates dominant and recurring advantages. This is true even in the chip industry.

TSMC was famously founded by someone who had been an engineer and an executive at a US company for decades. Similarly, China was able to get some of the world's best chip engineers and professors working and living in the US to move to China, and then ensured that all their needs are met.

In other cases, it is the flow of US technology. For example, the US owns the technology for some of the world's most advanced chip design tools. Similarly, when it comes to equipment, even though ASML dominates equipment, the technologies that it uses for manufacturing have entanglements in the US.

Third, the US is obviously the world's largest market for chips. The world's largest market will always have a big say in how things get done and what gets done. The US is also home to the world's largest chip companies. Again, that gives it an influence that no one can match.

Advanced chip manufacturing is one thing where the US is in a bind because advanced manufacturing has gotten so expensive that one company, TSMC, in Taiwan, dominates it. Same with legacy chip manufacturing. Because the US is a high-input-cost society, it will be impossible for the US to be competitive in legacy chip manufacturing.

But, again, because some of the foundational technologies across the chip supply chain were developed here, it gives the US this advantage that keeps recurring.

You mention that China is inviting professors and building its chip ecosystem. That reminds me of criticism here in India that if you want your talent back, you can't have a government salary structure. Some say you can't be paying Rs 2.5 lakh if you want to invite professors from universities abroad. What are the changes you think India needs to make, the way China did?

China has an enormous and enormously successful diaspora. India also has that. Look at the top universities - a lot of professors are from India. Top company executives are from India.

The Chinese have made a sustained effort. Announcing a scheme is easy. But executing it over long, sustained periods, even when it's painful and slow-going, that's the hard part. China has done this admirably.

China has this 'Thousand Talents Plan', for example. If you are an exceptional overseas expert who is Chinese or of Chinese origin - actually, you don't even have to be Chinese, you could be American - they'll bring you in and give you salaries higher than the salaries one will make in dollars in the US.

More importantly, they give freedom. Many of these people make enough money in the US or Europe that money is not what will make them do one thing or the other. But here's what will make them move: if the government says, "Look, I'll give you a lab, and I'll make sure that you can do whatever you want in this particular area. I'll make sure that there is no political interference." That's what's attractive to a lot of people.

China has thought through how to attract people from outside. And then it also puts in money, too. And they have been at it for over fifteen years.

Where does India fit into the chip-making ecosystem? India has started in legacy chip-making. Is it the right way to go about it, because we have seen criticism that India is putting so much money for basic chips?

Currently, India is a minor player at best. India has always had some strengths in chip design. A lot of global semiconductor firms have used chip design engineers in India for design or at least design services. While this has been mostly focused on the low-end design, it has allowed India to participate in the chip supply chain, at least indirectly.

The question for India is, can you retain your strength on the design side, especially after all the upcoming disruptions due to artificial intelligence? Second, can you make progress or capture a slice of the pie in other stages of the chip supply chain?

Both of these are going to be challenging. I hope that India is thinking very hard about how it is going to retain its strength it has in chip design since AI might greatly reduce the number of human chip designers you may need for a single chip. Now, to get into other parts of the supply chain, there are a bunch of challenges like technology, input capital, and human capital.

If only money lets you dominate a chip supply chain, then all of the chips in the world will be made in the richest countries in the world. That's not how it works.

It's hard to get into this industry. It requires an enormous amount of patient capital and a state-of-the-art technology portfolio. Developing and applying human capital requires carefully identifying and leveraging the inherent advantages you may have.

I think it's particularly hard for India to get into advanced chip manufacturing, at least in the medium term. In terms of manufacturing, India should focus on legacy chip manufacturing if it focuses on chip manufacturing. Even there, if it focuses on legacy chip manufacturing, then it must use a combination of tariffs, preferential access policies, and good tie-ups with the local market, especially the automobile market, to support and sustain this. This needs to be a 10-, 15-, 20-year game.

For India, assembly, testing, and packaging are relatively easy. It's still going to be hard to be cost-competitive and to find patient capital, but it's going to be easier to get into this because India still has sufficiently low-cost inputs. And I'm glad the government is making the right decisions.

India is now focusing on assembly, testing, and packaging. Those are the right things to do.

Raghuram Rajan has questioned the level of subsidy India is providing to build the ecosystem. His argument is, why spend so much money producing low-margin legacy chips? But no chip ecosystem elsewhere has been built without state support. You have also written about cases where dependence on foreign chips can create the risk of sabotage or even a 'kill switch'. So, does India need some domestic chip capability for its weapons systems?

Yes. India is an ambitious country. It wants to sit at the geopolitics high table in the near to medium future. India has to seriously think about access to chips for both economic and national security.

Similarly, India needs to grow its economy, and because the chip industry is highly lucrative, it also makes sense for India to think hard about how and where it can participate in this industry.

Chip imports are also extremely expensive for India. So, it needs to reduce its import bill.

So these are the motivations, and they make sense. For a country like India, you cannot not think about securing chip supplies. The real question is: How do you secure your chip supplies?

This is a question that other big countries like America, Japan, and China are asking as well. That's where Professor Rajan's comments and all these nuances come in.

You can either secure chip supplies by onshoring or by chip sovereignty. Chip sovereignty is where you say, "I need chips; therefore, I must build chips."

But that's not the only way. You can also do friend-shoring.

Friend-shoring is, maybe I won't build chips, but I will support chip building by friends. So, let's say Japan is my friend. India can put money into Japanese companies, just as the government puts money into HAL and HPCL. Why can't the government put money into TSMC in exchange for some chip guarantees?

Somebody who's good at it, knows how to do it, and has done it before, is taking most of the risk and doing all that. You're minimising your own risk. So, friend-shoring is another thing you could do.

Second, form the right alliance that guarantees chip security. You can say to your allies, I do the design, you do the manufacturing, and I do the assembly and test, etc.

If you ask me for my opinion on what India should do there, at least for national security reasons, India did the right thing with SCL ( Semi-Conductor Laboratory) - the Chandigarh facility that existed to supply defence chips before it burned down. Even in the US, they've always had some foundries that are dedicated to building trusted chips for defence.

Every ambitious country in the world thinks the same way: for their core defence needs, they don't want to get chips from adversaries or somewhere else that they don't trust.

So then what do you do?

Either you invest in or create some foundries, especially for defence. The problem with that is that chip technology grows so quickly. The difference between what you can have - let's say the government creates another SCL - and what is available elsewhere can be enormous.

So, if tomorrow the government says it needs defence chips and puts in $2 billion to set up a foundry, that foundry will come online in four years. By then, the technology would have advanced, and the chips in the rest of the world could already have 20 times the performance.

The second problem is, you put in that $2 billion or $20 billion once, but are you now going to consistently put that money in? India can't afford to do it, at least right now. In fact, most countries can't afford to do it.

So the problem is a real one, but the solution is unclear because of the rate at which chips progress.

For certain things, some avionics, radars, and missile systems - they need chips that don't need to be advanced. So you can set up a defence chip industry, and this industry produces all the chips for your defence needs. But how many future wars are going to be won by these? They're likely going to be won by the most advanced jets, fancy laser machines, and hypersonic missiles, and maybe all of them need the most advanced chips.

Then you get back to the same question: How are you going to build advanced chips? Where are you going to get them? And that's why, coming back full circle, chips are the most critical resource because everybody wants them. Everybody needs them. And everybody is trying to figure out how they can get access to them.

But that alliance can also have its limitations. Some say the US is an unreliable partner. In chip manufacturing, TSMC is closely linked to the US ecosystem. The US dominates design, and ASML leads in lithography. Given India's strategic concerns with China and its dependence on these key players, does it create a vulnerability for India?

Just because things are hard doesn't mean they are impossible. When I was growing up in India, I used to be told how, when India became independent, it couldn't manufacture sewing needles. Clearly, India is a long way from that.

So, the way I would phrase it is that getting into advanced manufacturing and being cost-competitive with incumbents is challenging. But challenging and impossible are two different things. Every country needs to do a combination of things to make progress.

For example, I would have alliances to ensure economic security in the short to medium term. At the same time, I would invest in capabilities for the long term, knowing that I will not win at least for a long time at the global stage. That's where you decide where you want to have sustained investments.

There are two big worries I have. In some other article, I used the phrase "chip nationalism". It is when you start putting in money without strategising for the long term.

For example, let's say this government puts $20 billion into chips, and there's no national consensus that we should be competitive in chips, that we should be putting money into chips. Let's say if the next government comes in and Professor Raghuram Rajan is the finance minister, and he's like, "I'm not going to put money in chips." That's $20 billion lost.

This is the kind of thing that's a long game.

Look at other countries in this space. South Korea was nowhere in chips. Malaysia, Singapore, and Vietnam were nowhere. Chips started in the US, but how did these other countries develop strengths?

China is a great example. When China started truly rising, policymakers in China were having exactly the kind of conversation we are having, since all their chips were coming from abroad. In some years, China imported more chips than oil in terms of money value. So, imagine China having this discussion 15 or 20 years ago. They decided as a country that, look, semiconductors will be the key to us being competitive, and therefore we will invest enormous amounts of money. Since 2015, look at the amount of money they've invested.

And the glass-half-full way of looking at what they've done is that now, they have become extremely competitive in almost every part of the chip supply chain. They have choke points that they can control in the legacy chip world. They have indigenous, world-class programmes for really every level of the chip supply chain.

The glass-half-empty way of looking at it is that, in spite of putting in tens of billions of dollars, they still can't manufacture the advanced chips.

That's what you decide as a country. Chips are one of those things that are highly capital-intensive, and you have to make sustained investment and have some kind of national consensus if and when you decide that you're going to get into the space. And my worry is: do we have that consensus?

India should be investing in the chip industry, but do it prudently. It should invest in strengthening the chip design ecosystem to move from low-end services to cutting-edge design - Israel serves as a great role model, and start with putting in an effort to be competitive in packaging and assembly. It should start doing capacity-building in legacy chip manufacturing, knowing that they may not be competitive in the short term.

But in order to still sustain it, you have to have a mixture of tariffs and mandate the government bodies and local automobile industry to buy local chips, within reason, to keep the investments into the industry viable.

You have written in the book that if AI progress stalls, it would be bad news for the chip industry. That is exactly what Sam Altman and Anthropic CEO Dario Amodei are now talking about. What is your take on whether AI progress should slow down?

Developing frontier AI models critically needs access to a large number of the most advanced chips. These chips are extremely expensive.

The cost essentially - when you look at the capex for these AI companies that are running different kinds of AI data centres - is extremely high. A lot of these companies have high valuations, but their capex-to-revenue ratio is extremely high.

At some point, if AI progress stalls, then how do they pay back all this money that they have spent?

A lot of these companies are funding this investment through debt.  And the collateral for this debt is often the chips that they already have.

The whole industry is running under the assumption that, at some point, advancements in AI will be able to pay off all this capital investment they're making. What if it doesn't work out for whatever reason? What if we don't see the AI companies generating enough revenue, and they start defaulting on the debts that they have?

That's the primary problem. So if AI does not pay at some point, then any default that may happen can lead to catastrophic economic consequences, especially if the creditors have exposure to public banks.

Where do you stand on what the tech bosses are saying, that advancement should slow down? And some tech CEOs have called it a bubble, while a former HCL CEO has said we don't see use cases as yet. What are your views?

Current calls for slowing down AI have nothing to do with chips. That discussion is about the potential security and ethical consequences of unfettered AI development.

As to whether AI is a bubble, I think AI is a transformative technology. And what more evidence do you need than the fact that every day it discovers a new drug or a new enzyme or solves a new previously unsolved maths problem?

But even when you have bubbles, they lead to massive investments and, when done right, these investments do lots of great things for society.

In the 1840s, there was a bubble called railway mania in Britain. Everybody was speculating on the value of railway infrastructure, which was nascent back then, and when that bubble burst, a lot of people lost money, but you still ended up with a great amount of railway infrastructure in Britain. In fact, more than half of the railway infrastructure that exists in Britain today was developed during that bubble.

Similarly, a number of companies lost money during the Internet bubble. But even though a lot of people lost money, the result was that we ended up building all these underwater cables and other fiber-optic telecom infrastructure that support the internet today.

Even if AI is a bubble, it is leading to one of the largest buildouts in recent history. We are upgrading grid infrastructure to support AI. In the future, we'll have much more access to power than today. We are building new chip companies and developing new energy technologies.

So even if you assume that AI won't transform the world in fundamental ways, it's creating this build-out that will be good for society as a whole.

Most companies that existed during the dot-com boom have gone bankrupt, but that's okay because companies such as Amazon and Google survived and changed the world.

That's what will likely happen even now. Most AI companies will die. Some will survive. We will do a lot of build-out. But the jury is still out on what kinds of companies will make the most money.

What is the lesson that we can take from China when it comes to chip-making? And second, can we also at least try to attract foundries like Samsung and TSMC, the way we did for Apple phones? Is it doable? Should we look at that option as well?

At some level, China serves both as a good role model and a cautionary tale. It's a good role model in the sense that they have been very strategic, very methodical, and very persistent in how they have pursued the chip industry.

Once they decided to get into it for economic and national security reasons, they made sustained investments. They used a combination of tools. The persistence, the strategic manoeuvres, the consistent state support, the ambition - all of that India can learn from China.

At the same time, China also serves as a cautionary tale in the sense that it tells you how hard it is to actually become a leader in this industry.

They've spent so much money, but in spite of this, all it takes is for the US or the Netherlands to stop shipment of some chip or some chipmaking equipment, and then you're stuck.

Chip controls still bite China in spite of it having spent tens of billions of dollars over the last 10-plus years.

So it's a cautionary tale in the sense that it also tells you whether it's a good idea for a country like India to invest money into chip manufacturing, especially advanced chip manufacturing.

Ultimately, the government has to decide what the right thing to do is. Money does not grow on trees. If you had infinite money, it's a good idea to invest in the chip industry. But the way economists like Dr Rajan think about the question is, if you have only $100, where should the money go?

We can learn from the Chinese roadmap if India chooses to go into the chip industry. But it also serves as a cautionary tale that tells you the limitations of what you can do, even if you invest a lot of money and make your best moves.

On the question of whether India should attract TSMC or Samsung: In the Apple case, when you look at exports only, of course, it looks like a success story. But you should also look at the imports. We should see how much value is being added. Value addition is what captures money.

What you're doing right now in India with iPhones is primarily assembly. Apple is not building any chips that have high margins. Apple is not doing any design in India that has high margins. The components are coming from China and elsewhere; the design is coming from the US.

You want to have something in the country, even if it is a foreign-owned, either because you want to get a significant fraction of the revenue in a highly lucrative industry - mostly through taxes or local employment, or because of economic security, meaning you want to make sure that chips for your planes, trains, and automobiles are always available, that nobody can stop access to them.

And third is for national security - that your fighter jets need some chips, then you always have access. That's the right way to think about it.

So now let's say if Samsung shows up in India. If Samsung runs its factory in India and gets most of the revenue, it's a bad idea for India, especially if the subsidy that needs to be provided is large versus the tax revenue.

It may be a better idea if Samsung employs a lot of people. But foundries don't need a lot of people to run. It's not like the textile industry in that sense. If I were an economist, I might rather put money into the textile industry in India than a foundry because textiles are human resource-intensive. And a country like India needs jobs.

Second, do you have economic security?

If tomorrow India and Korea get into a fight and the Korean government says, "Don't sell the Indian government the chips." How would that work? Especially, the Indian footprint is a small fraction of the overall footprint.

Just having Samsung in India doesn't guarantee you economic security or national security.

So yes, India should try to attract these companies because it allows capacity-building. That way, you learn how to build chips, get some employment, and develop ecosystems around.

The question is, what is the cost? If Samsung comes to India, I'll happily have Samsung in India. But if Samsung says it's a $3 billion foundry and I want you, the government, to give me $2.5 billion to set it up, then I don't know.

So there has to be this right balance. You have to think carefully about how much subsidy to provide versus the value you're getting out of it. It's easy to say we will have chip manufacturing. But it takes money.

Of course, building chips in India is a good idea. But if you have $100 today, what's the right way to spend it is the question.

But if you look at all countries leading the chip industry today - the US, Japan, Taiwan - no company in these places has ever been able to set up these ecosystems without government support.

Absolutely. I am not contradicting it. I'm actually saying exactly that. These are capital-intensive industries, so a decision needs to be made at the governmental level. India needs to decide whether it wants to get into chip manufacturing. If it decides to get into chip manufacturing, the only way is through massive and sustained state support over decades, even for legacy manufacturing. Otherwise, it's not going to work.

You give Japan and Taiwan as examples - I also talk about this in the book - decades after decades, they gave support for the chip industry.

If you were running India's ambitious chip programme, what are the two or three things that you think India is doing right, and what are the things that you think need correction?

India should strengthen its chip design ecosystem and make it AI-ready. It also needs to move to cutting-edge design and try to cultivate world-class fabless chip companies. Then, assembly, packaging, and testing are good beachheads. So that's a good way to enter the hardware market.

It should strengthen alliances to ensure chip security. India is an ambitious nation, and it will need access to chips, both advanced and legacy ones. It's not prudent to focus on advanced chip manufacturing.

India should focus on legacy chip manufacturing and the right combination of tariffs, policies, and connections with local markets, especially the automobile market and the government as a large customer.

Now that India has entered the chip race, is there a metric to judge its success? When would you call it a success?

Everything has to have a metric for success. So what's the definition of success here? The easy answer would be that India should have a certain significant share in the chip industry by a certain year. Currently, that share is close to zero. Now, where that will come from, that's the question. Should it come from manufacturing or design? Or something else?

Focusing on manufacturing, if you're putting in all this money, better get a return. That's absolutely the first metric of success.

Second, you should also have some clear benchmarks. For example, manufacture at least 28-nanometre chips in the country by this particular year.

The advantage of setting a goal is that then you can work backwards. You will say, okay, what does it take to manufacture 28-nanometre chips in 2030? How much money does it take? What human capital does it take? What skills does it take?

India should identify some benchmarks and meet them. And when it cannot, it should have a course correction.

You write about counterfeit chips in Boeing maritime aircraft, and also in F-16. This is scary because we rely on imported chips. You have also written that 'imagine a passenger plane that suddenly falls out of the sky because one of the counterfeit chips inside failed'. This reminded me of the Air India crash last year. What came to your mind when you wrote this line?

Yes, this is real, this is not hypothetical. The reality is much more dramatic and scary than fiction.

NVIDIA builds these advanced high-end chips called H200. The US had decided that NVIDIA is not allowed to sell these chips to China. But after some time, the US said: China can have these chips. But you know what happened? The Chinese said, we don't want these chips anymore. And one of the reasons is that there was this rumour that now there is location tracking on these chips that had been put in, so that they can track whether there's an unauthorised use of the chips.

And so then China said we don't want it because who knows what other things are on the chip that can do snooping. So, countries have good reasons to worry about chips snooping or behaving maliciously. Because you have the supply chain that goes through all these countries. And anybody can do anything.