Become a Creator today!Start creating today - Share your story with the world!
Start for free
00:00:00
00:00:01
India's Semiconductor Bet: Hareesh Chandrasekar on AGNIT's GaN Technology Stack
0 Plays2 seconds ago
When China banned gallium exports, it didn't hurt AGNIT Semiconductors, it made them essential. Hareesh Chandrasekar reveals how geopolitical supply chain wars created a $13 million opportunity and why India's first GaN chip company is competing with billion-dollar rivals on just $5 million.
In this episode, Hareesh Chandrasekar, Co-Founder and CEO of AGNIT Semiconductors, shares the unconventional journey of commercializing 18 years of IISc research into India's first indigenous GaN chip company. From leveraging ₹300 crores in government-funded R&D infrastructure to competing with billion-dollar global players on a $4.87 million budget, Hareesh breaks down the capital-efficient playbook for deep tech startups.
He reveals how China's gallium export restrictions created sovereign demand for AGNIT's chips, why defense contracts came before consumer markets, and the brutal reality of scaling from lab prototypes to 100,000 chips in 12 months. With three chips currently in field trials for defense applications and expansion planned into electric two-wheelers, AGNIT is at the forefront of India's semiconductor manufacturing revolution.
He shared this candid journey with host Akshay Datt, exploring the intersection of geopolitics, deep tech commercialization, and the India Semiconductor Mission 2.0. This conversation is essential for founders tackling hardware, investors evaluating deep tech, and anyone interested in India's strategic technology ambitions.
In this episode, you'll discover:
👉How Hareesh Chandrasekar spent 18 years building GaN expertise at IISc before raising a single VC dollar, using institutional R&D as non-dilutive capital to de-risk AGNIT Semiconductors
👉Why semiconductor startups take 2-4 years and $2 million just to reach VC-fundable stage, and how the deep tech timeline differs radically from software
👉The military-to-commercial strategy: starting with defense jammers, radars, and drone communication chips before pivoting to high-volume electric vehicle markets
👉How China's control of 87-90% of global gallium reserves and export restrictions created guaranteed sovereign demand for indigenous semiconductor supply chains
👉AGNIT's fab-lite model: controlling IP and critical manufacturing steps while outsourcing volume production, competing with $300M+ funded rivals on $5M
👉The make-or-break challenge: scaling from hundreds to 100,000 chips in 12 months to validate foundry partnerships and achieve commercial viability
👉India's semiconductor ecosystem reality: zero domestic wafer production, complete import dependence, and why $500M GaN foundries are more achievable than $20B silicon fabs
👉Why pitch decks work for investors but defense customers demand working prototypes, data sheets, and field trial results before taking startups seriously
#Indiasemiconductor #GalliumNitride #GaNchips #semiconductorstartupIndia #IndiaSemiconductorMission #ISM2.0 #defensetech #compoundsemiconductors #semiconductormanufacturing #chinaexportban
Disclaimer: The views expressed are those of the speaker, not necessarily the channel
Recommended
Transcript
Introduction to Semiconductors
00:00:00
Speaker
A chip is nothing but a piece of semiconductor real estate that implements a certain functionality, which is clearly defined. In the semiconductor space, it could take you you know anywhere between two and a three years to get to that proof of concept stage and $2 million, dollars let's say. How did you get that initial $2 million dollars that you needed to reach PMF? In Silicon, if you want to set up a cutting edge boundary, I think there are a lot of reports that will tell you this will cost anywhere close to $20 million dollars
Emergence of Agnit Semiconductors
00:00:20
Speaker
today.
00:00:20
Speaker
Harish Chandrasekhar is the founder of Agnit Semiconductors, India's first gallium nitrate chip company. Agnit is building high-power semiconductor technology used in defense, telecom, and next-gen power electronics. In 2014, the Nobel Prize in Physics was given for the invention of efficient white light emitting diodes. That's the genesis of gallium nitrate as a material. China has about, I think, about 87 or 90% of the world's gallium reserves. Technology doesn't automatically become products, and products don't automatically fly off the shelves as well. I think our biggest problem is going to be something like manpower shortage.
Agnit's Focus on Gallium Nitride
00:00:59
Speaker
Why are you, I mean, fairly simple, silly sounding question, but why are you running a semiconductor company? Like what makes you qualified to run one? Let me put it that way. So, you know, um We started Acne. Acne today is India's first gallium nitride semiconductor technology and product company.
00:01:19
Speaker
and So all big words, so we'll have to explain what what exactly they are. So ah just to give you an idea of where the journey is, you know, gallium nitride research started in the Indian Institute of Science, Bangalore, in the year 2006-2007.
00:01:33
Speaker
And Acne is a spin-off and we started operations in January 2021. So, you know the team here had about 15 years of experience in developing technologies all the way from making the gallium-natured material.
00:01:45
Speaker
And then, you know once you've made the material, you need to to run through hundreds of semiconductor manufacturing process steps in order to implement, let's say, a chip. And then the chip needs to be designed. So there was significant expertise which was built over many, many years, right over a decade and a half of doing all of this.
00:02:01
Speaker
And that's when we you know decided that we had technology which could be manufacturable. And the time also seemed to be right in order to do this you know kind of commercialized technology that we all had.
00:02:12
Speaker
A slightly broader way of thinking about it is that, you know, if you India has, in the semiconductor space, gone over from being a service-oriented ah sort of a development to being at a stage where Indian companies are now owning IP and creating IP and then creating products, right?
00:02:28
Speaker
And that's just journey and acted as part of the journey. I think what makes us unique, as I said, is probably we the first startup working on this new material called gallium nitride. And, you know, we also have manufacturing technology. So historically, it's again, you know, design forward in the Indian ecosystem.
00:02:43
Speaker
We not only do design, but we also do manufacturing. And, you know, happy to explain why for ah for a variety of reasons. I think this makes sense for gallium nitride. Right. So happy to talk about why that is.
From Academia to Entrepreneurship
00:02:52
Speaker
ah So, you know, ah that's in ah in a brief about the effort itself.
00:02:57
Speaker
So I myself have worked on gallium nitride since the year 2009. I graduated in 2016 from ISE with a master's and PhD in gallium nitride. I then spent two years as a postdoc in the UK. Then I spent two years in the US as a postdoc.
00:03:11
Speaker
And of course, you know, having done two postdocs, I thought I was going to have a distinguished academic career. But you know, life had other plans, right? so So here we are. So this is why we did Acnit and what it is that Acnit is today doing, right? So I'll pause there and, you know, over to you. Both your postdocs were on this only like GAN? said Yeah, yeah, it was all in Gallium Nitrate. I'm a one trick pony, so that's all I know. So you're essentially saying that IAC built up both the talent and the know-how of GAN as a material and using GAN for semiconductors. And what you're doing is commercialization. So your co-founding team are essentially academics who are now commercializing the technology that they built.
Challenges in Deep Tech Commercialization
00:03:55
Speaker
That is correct. ah You know, there is a lot of engineering and science that actually goes into even commercializing technology. The one thing that we've learned in our journey so far is that technology doesn't automatically become products and products don't automatically fly off the shelves as well.
00:04:08
Speaker
right So there is a lot of engineering that needs to be done in order to put in, that needs to be put in just to get the technology out into a usable product by any customer. And then you still have to figure out how is it that this becomes a viable business.
00:04:20
Speaker
right So all of those is are something that you know that keeps us very interested. And you know that's yeah that's kind of reason for existence for us. In the Indian academia, how is how does this happen in terms of, you know, there would be, for example, who has the IP for it? Is it IISC because you built those things in IISC or is it with Agnet and how hard or how easy is it for a bunch of academicians to raise an initial seed check and stuff like that? like Like, how does that happen, that commercialization of ah unique technology that is built up?
00:04:59
Speaker
The IP question is very important. So, you know, we when we started out, we had about 10 odd patents. And of course, it was IAC which had those 10 odd patents, right? So they very kindly licensed it to us when we were doing, when we started Acnit.
00:05:14
Speaker
So that is really helpful because that's the baseline technology on which products get built. And over the last four years, we ourselves have filed for another you know six odd patents, mainly resulting out of you know all the other original work that we had to do in order to translate those technology into products in the first place.
00:05:29
Speaker
So not everything will work, some things will not work, you necessarily have to come up with newer solutions. So that's where we are on the technology front. In terms of what it takes for a few academics you know to to raise a seed check and then get started, this is a very important question. And it's primarily to do with, in any deep tech space, it takes time ah for you know for you to even make a proof of concept or making ah make an MVP that you can actually put before customers.
00:05:56
Speaker
Normally, you know the vc ecosystem is very good once you establish some level of product market fit to come in and help you scale up. But in the semiconductor space, you could, you know even to get to the first and MVP, it takes you anywhere between three years to four years, and it could take you a couple of million dollars.
Funding and Financial Ecosystem
00:06:12
Speaker
So that is a fairly significant difference. And I think that that's exactly where you know we also took the time that it it took the time it took for us to do that. But, you know, that is primarily what, you know, that's that's where we are today.
00:06:27
Speaker
So not easy, but, you know, we pulled on. And at the same time, this is also sentiment driven to ah to a large extent. So today, I think every every large economy sees the value of having homegrown semiconductor technology, right? And that, I think, has been an important part of what we're doing.
00:06:42
Speaker
How did you raise that or how did you get that initial $2 million dollars that you needed to reach PMF? Like who who funded you pre-PMF? Yeah, we were we were very lucky. We were able to raise angel funding just about the time we started the company. And our angel investor was Mr. Lakshmin Dharan, who used to be the CEO of Cognizant.
00:07:00
Speaker
So he was a real agent. So I think ah he invested us into us based off a pitch deck and the kind of vision that he was able to see and the team. So which was really helpful. and that's what kept us going. And then we got to a stage where we know we had the first products. right So it took us about two and a half years to to get there.
00:07:18
Speaker
At that point, we could go out and raise a seed round. And as I said, it's not also just about what we were doing internally. It's also the external environment also needs to be receptive to efforts like this. Now there's a lot of talk about funding deep tech innovation in the country. right So this wasn't the case in 2021 when we started out.
00:07:34
Speaker
So even the ecosystem has matured to a certain point.
Chip Design and Manufacturing
00:07:37
Speaker
Okay. um What is the journey of, ah ah like, you know, idea to millions of chips?
00:07:48
Speaker
And I'm assuming a chip only makes economic sense if you're selling millions of them, right? like Like selling thousands of chips wouldn't be economically viable. Like that wouldn't be commercialization. So what's the journey from an idea of a chip to millions of chips?
00:08:04
Speaker
Actually, it's very interesting. So selling thousands of chips, I mean, we should put a pin in that and revisit, right? okay But I'll just walk you through the journey of what it what it means to it is to go from, you know, let's from design, let's say to volume volume production, right?
00:08:17
Speaker
So typically, when you start out, you have some, like i said, a chip is nothing but a piece of semiconductor real estate that implements a certain functionality, which is clearly defined. So firstly, you know how is it that what is the kind of design that's going to come in?
00:08:31
Speaker
How is it that you're going to design this particular chip? What is the functionality? So you need to have that specs ready. And then you know the design process begins. And the design process again is different, whether it's logic memory or whether it's a discrete sort of an application.
00:08:44
Speaker
Why I say that is because typically in logic or in memory, you're you're dealing with millions and billions of transistors. right So that's ah that's a reality. So this is VLSI, very large scale integration.
00:08:54
Speaker
but Although you know we should probably come up with another term for it now that we're at the billions of transistors scale. VLSI was a term that was coined, but we were still at ah you know maybe a billions of transistors out of the scale. But anyway, so so now that we have, you know ah so it's very different on the other hand in case of discrete.
00:09:10
Speaker
So a light emitting diode is one chip. So the power switch that, the Calium Nitrate power switch that you use in your in your phone is typically one, it could be one transistor, two transistors, right? But it's a handful of transistors at the end of the day.
00:09:23
Speaker
So very different kind of scale. So the complexity is different as well. So the design complexity, just talking about design now. So even the design complexity is different. you're You're trying to handle extremely high power. So you also need to factor in other things. right So it's not about just integrating.
00:09:37
Speaker
It's not about integrating it and then seeing that, oh, do I get the functionality here? And you know I have an IP block that I can drag and drop and then you know sort of use that as a functionality, et cetera. So in this case, when you're using, let's say, a smaller number of transistors,
00:09:51
Speaker
then you better make sure that, you know, the design for manufacturability piece is also addressed right at the beginning. And then you also need to make sure that they are the best possible. So if you're using a handful of transistors, you better make sure that two, three, five, whatever you use, they are the best possible transistors for that job. So that is a decent amount of customization that can also happen, right?
00:10:09
Speaker
So because every use case is different, every design is different, the user is different, the geography is different. So it comes with different standards, right? So that's the whole design piece. So once you've got the design, whether it's in in a discrete, in a logic, in memory, whatever it is, right?
00:10:23
Speaker
So you will typically, you know, interface with a foundry in order to get your production done, right? So that's the whole manufacturing piece of it. And today, foundries have evolved to a point where, you know, you can always use something called a design kit.
00:10:36
Speaker
So foundry will give you a design kit, which is basically a bunch of rules that the foundry tells you, this is the way that you can interface with me. So these are my rules. I cannot manufacture anything which is you know more than this, less than this, whatever it is.
00:10:48
Speaker
And the whole manufacturing process is sort of abstracted away. So for for a designer, a transistor could be a black box.
From Wafer to Chip Packaging
00:10:54
Speaker
In fact, typically it's an IP block which has a few, you know, 10th of thousands, if not billions of millions of transistors, which is a which is a building block.
00:11:02
Speaker
And then once you do that and then you you design your functionality, then you can then send out your chip for tape out. So this is another term that's normally used in the industry. So tape out literally means that, you know, ah and in the earlier days of silicon.
00:11:17
Speaker
So how is it that you're going to pattern? So all of this is a two dimensional process. So think about it like printing. It's like screen printing. Right. So it's a two dimensional process. So you're putting layers, you're adding layers, you're removing layers. All of it is done in a two dimensional fashion.
00:11:30
Speaker
Right. So you you could use masking tape in order to actually realize your designs on the chip. That was a very, very early days of chip. So it was called a tape out. So, you know, so that term is kind of stuck. So a tape out is when you send your design to a foundry and the foundry does its own bunch of checks and then accepts it.
00:11:46
Speaker
and then says that okay now I can manufacture this because you have designed this according to a design kit I gave you and followed certain rules that I enforced right apart from that the functionality is yours right so now I can accept this for tape out and then I'll go and produce your chips and this process could take anywhere between let's say three months to a to a year depending on how how packed the foundry is how difficult ah you know your your design is to manufacture All of those things. right So once you have that, then essentially think about it as the foundry shipping you back your chips.
00:12:16
Speaker
Now they're all in physical form. Earlier it was in design. right It was a bunch of zeros and ones again, so design files. And then you've got your physical chip back. And now you need to start the process of you know assembling it, packaging it, testing it.
00:12:31
Speaker
right And so that process begins. So this is typically called you know ATMP, Assembly Test Marking Packaging. So this is there are a lot of waste ATMP plans. ATMP and OSAT are also in some cases used interchangeably.
00:12:44
Speaker
outsourced assembly and testing. right So we have a lot of such plants, for example, coming up in India. So there have been a lot of, you know for example, the very first one that Micron project, which was announced in Gujarat, is an ATMP plant.
00:12:55
Speaker
There are quite a few like that. right So Keynes is putting up one, CG Power, Tata has one in Assam. So these are all you know examples of plants which do OSAT or ATMP. And once you have that, then you as the, let's say the chip designer are getting back your own chips, which is now packaged.
00:13:11
Speaker
And then that looks like something that you're probably used to. you know It'll have it look like an IC with legs or you know with leads. And then you can put it on your test board and then test that the functionality works and then start shipping sample products to your customers.
00:13:24
Speaker
And once your customers have, you know, you've consulted them, put down the design for their use cases. So once they test it in their particular application, first, typically, you can know, you can assume that it goes through some some kind of a lab bench test or a lab test.
00:13:36
Speaker
And then it's deployed out in the field. And then people see, is this really doing what it's supposed to be doing? And then they come back to you and say, okay, this is great. Now, you know, can we go for a larger volume? And that puts you at volume production. So your design is okay. If it's not okay, you have to go back and iterate.
00:13:51
Speaker
So if this sounds like an exhausting process, it's because that's the reason that it takes about a semiconductor startup anywhere between, let's say, two and a half years to four years in order to put a chip out, the very first one out, right? Because you're literally starting from scratch.
00:14:04
Speaker
And it's also a very expensive proposition. ah very first chip after testing or even the very first test chip no starting from from beginning from beginning from let's say defining your specs to the time that you actually get your silicon back right so that itself could be you know um a year year and a half sort of process and then you test okay and then you and then you assemble test it and then bring it put it out to customers make sure that you know it is indeed doing what it's doing corner cases are covered right um And then you know if there is an iteration, then of course the cycle will repeat.
00:14:34
Speaker
It won't take as long, hopefully, because you you can be quicker with the design. There might be modifications that you will do. but you know So that's the typical time scale of the process as well.
Economics of Semiconductor Manufacturing
00:14:43
Speaker
right And that's what it takes to put. And of course, you're also going to look at things like, is this reliable? you know Is this chip really, if I put it out in the field, is going to work for, let's say, 10 years at least?
00:14:52
Speaker
Which means that you'll also kind of stress test it internally. And that's also part of the reliability and qualification cycle that a chip will undergo. So that in a nutshell is what happens designed to let's say something like volume manufacturing.
00:15:06
Speaker
And of course, you have to provide the after sales and support as well. So let's say users are having trouble with the chip so you know um or after they buy it or in certain cases it doesn't work. Then you have to figure out why that is and then offer them support as well.
00:15:19
Speaker
right So this is the life cycle, so to speak. What is the ah assembly? which What do you get back from the foundry that you need to assemble further? Like you said, there's assembly, there's testing. Doesn't the foundry test and give you chips? like like The foundry will test chips, but they will also test things which are on the wafer. So just to put it very briefly, so you you start by, let's say, um having a wafer, which is like a silicon wafer.
00:15:41
Speaker
So you've seen CEOs of Intel's et etc. you know kind of posing with a big chip, right? So a big wafer. So that is literally why it's called a wafer. It's extremely thin circle of silicon. So think about it that way, right? It's a disk.
00:15:52
Speaker
but It looks like that is old record player record. Absolutely. it looks like the old record player record. Yeah. And then it has, so that thing could have anywhere between, let's say, you know, a few thousand to a million chips, but which is this integrated functionality on it.
00:16:06
Speaker
And then you have to cut it up. And then you know that piece of silicon needs to be removed and then put into, let's say, a package. You don't see that piece of silicon on on if you open up any, you know like say, PCB board.
00:16:17
Speaker
You won't see that. You'll normally see, let's say, something that has a black casing and with legs sticking out. right so So if you open up that package, though, you'll find that the small piece of silicon is actually sitting inside.
00:16:29
Speaker
And then that needs to communicate with the real world. So let's say we are talking about a package with with the el leads. right So it looked like a centipede right so with legs. So you put your piece of silicon inside and then you've done some amount of how do you bring that electrical connection? So I may have a pad, a metal pad on top of the chip.
00:16:45
Speaker
So that's what's giving me my signal. So it's come to the end of the chip. Now, how do I get it out of the chip? you know, into that lead. So I have to connect it. So the simplest way of doing it is to use some kind of, let's say, wire. So I use a very small wire and then I bond it. There are many, many techniques of doing this. And of course, at very very large scale integration, you probably won't even use wires. You just, you know, you can do things like flip chip. And these are all industry terms.
00:17:07
Speaker
ah Bottom line is you're trying to bring bring your connection, which was on the on the silicon real estate out into the real world through a lead. And this is all packaged because you know you don't want this silicon to actually experience the real world.
00:17:19
Speaker
So you don't want it to experience the temperature, humidity, all of which could cause the chip to fail a lot faster. So you put a casing around it, just like how we put a casing around our phones. right So you put a case around this. right So that, in a nutshell, is your package.
00:17:31
Speaker
And once that's done, you also need to test it. Once the packaging is done, because you want to make sure that the packaging, the process of packaging didn't affect the functionality that you designed it into in the first place.
00:17:42
Speaker
Right. So once that's done, then, you know, um you can always, its it makes a lot of things very easy. You can, for example, let's say you're assembling the the iPhone and then you have, ah you know, you have to do this, let's say a million, you're shipping out a million pieces a month.
00:17:56
Speaker
So you have millions of such chips that you've made. So it's very easy for you if it's in a package format for an automated assembler to come in. pick this component, put it on a PCB board, and then you know kind of keep keep assembling those and then keep pushing them out.
00:18:09
Speaker
So that's primarily what that's what packaging is. So you know it's like exactly like putting a case around your phone by making sure that the functionality is not affected. But all the electrical signals still need to go through.
00:18:20
Speaker
So you have make sure that that happens. The foundry doesn't do any of this. It'll just give you that one big wafer. Even the cutting you do yourself? Yeah, typically foundries can also cut, right? So they could do, that's called dicing. So you you kind of, you know, create dice. So die is another term that you'll again normally hear.
00:18:37
Speaker
So it's that piece. So when you kind of, you know, cut up the chips, the the small, each discrete chip that you get there is a die. Okay. So that is a die. And foundries will typically also can also cut and you know and typically they don't do the packaging. Some foundries will do everything. So while they can also assemble and package.
00:18:55
Speaker
It depends exactly on what what you want to do. But this is a specialized function. And normally, you know the technology for cutting edge chips does not really lie in the same place at which it's more labor efficient for you to do packaging.
00:19:08
Speaker
right So today, East Asia is extremely popular for doing packaging functions. And a place like TSMC in Taiwan is very good at making chips. So there is also a certain amount of where the where the kind of capability as well as the economics work out.
00:19:22
Speaker
So that's typically the reason that the two functions are split. right So the foundry was very good at manufacturing silicon and will manufacture silicon. And it's probably more cost effective for it to be packaged, let's say, in somewhere like Malaysia.
00:19:34
Speaker
So the packaging might happen there. So that's that's where it is. yeah Okay. And India is now starting to get these packaging companies which are doing the testing and the soldering with wires and creating that final package. Okay.
00:19:51
Speaker
Okay. ah So you said that you manufacture in-house. You said that even thousands of chips could be viable to sell. Help me understand that. Yeah.
00:20:02
Speaker
it It depends entirely on firstly, though you know the scale of operations. right So in semiconductors, what happens is that you know it takes you a certain amount of, you put in a certain amount of investment in order to set up manufacturing capacity.
00:20:14
Speaker
And like any manufacturing capacity, you want to pack it as much as possible. right So that's the reason that economy of scale typically works. And if you think about it, you know let's say we have a few hundred manufacturing steps for you to manufacture a chip.
00:20:26
Speaker
And so one logical thing is of course, okay, so let's say I can manufacture a chip on a two inch wafer. So it's about this small. I'm still going to go through 200 steps. Let's say it takes me roughly the same amount of time if I were to expand.
00:20:38
Speaker
my chip size to four inch, I have to go through the same 200 steps. And let's say I do eight inch, right? I have to go through the same 200 steps. Let's say I do 12 inches. I still have to go through the same 200 steps. It'll take me roughly the same amount of time, right? You know, it might increase things like testing time, et cetera, but that's okay, right? So, you know, that's the reason, firstly, that the wafer sizes kept increasing. So today's silicon is on 12 inches, for example, right?
00:21:01
Speaker
So when it all started out, it started out in sort of like a one in one inch scale and in way back in Fairchild, et cetera. So if you go back to the history of silicon, And so that's the reason it's grown. So that's one kind of economy of scale.
00:21:12
Speaker
The other kind of economy of scale is once you put up a certain amount of fixed capacity, you want to make sure that you can pack it as much as possible. So you know it's not 10 times more expensive to put up a semiconductor fabrication facility that can process 10,000 wafers a month when compared to something that can process 1,000 wafers a month.
00:21:31
Speaker
So it may only be like you know two times as expensive. So, you know, so why why you not try to do that? So you do get economy of scale that way, but that's true for most manufacturing operations, right? So what I said when i when I, what I meant when I said, you know, it doesn't necessarily have to be in millions, millions is because we are we are traditionally used to, you know, a lot of the use cases that we encountered, right? So a million chips is like small volume.
00:21:54
Speaker
a year Let's say you're chipping shipping a million chips a year. It's a small volume for a certain functionality. But in certain other functionality, it could actually be a pretty huge volume. right So 10,000 could be a big number in certain applications. 100,000 would be a very good application, good the use cases. right ah For example, a classical example is, let's say you want to build something ah for the strategic sector, let's say defense.
00:22:15
Speaker
So today you could put, depending on depending on a radar, you need maybe like 3,000 to 10,000 chips for a radar. let's say right is that small volume or is that large volume right so let's say that you're able to install 10 radars a year you will need 30 000 chips to 100 000 chips right so in those cases use cases it it makes perfect sense and the functionality is much higher right but is it is it not critical for the strategic guys to have that sort of access if we are going to move up the value chain and build systems in defense not just for internal consumption but also be a net defense exporter etc
00:22:48
Speaker
So you also need that. So it's not just about economy of scale and depending on your operation, it's also more about economy of
Foundry Process and Requirements
00:22:54
Speaker
scope. I have a certain manufacturing capacity. How am I going to bring the most out of it? What is high value addition for me?
00:23:00
Speaker
It not necessarily be high value addition for somebody who is who has a different sized manufacturing, etc. right And given that we are coming off a manufacturing semiconductor manufacturing base, that's almost zero. right so it makes more probably makes more sense to also think about the economy of scope that you can actually get out of plants that are being put up, not just about the economy of scale.
00:23:20
Speaker
ah Typically, what's the minimum order quantity from a foundry? So a foundry is typically, you know, it depends on if you, if you, let's say do a pilot, you're you're typically doing something like one wafer. So the foundry says that, look, I'm now running on an eight inch wafer. I'm running this particular process node.
00:23:36
Speaker
So let's say I'm running three nanometer or 10 nanometer, 180 nanometer, right? So you should book at least one wafer. so So that's typically the one ah minimum. But you know ah many nowadays, you know things also get processed in a cassette. So you could have a cassette of 25 papers and a foundry could well come back and tell you that, look, you have to book a cassette.
00:23:56
Speaker
So it depends again on the kind of foundry, what is their own internal demand, etc. But you'll book at least one wafer. And you you also, you know, for for academic purposes, people also do something called a a shuttle run. It's called a multi-process wafer.
00:24:09
Speaker
So, you know, you could pack your design. They could aggregate designs from, let's say, 10 different players and then put it all on silicon and then say that, okay, but I'll do this once a year. or I'll do it once in six months. So, you know, which is okay. You know, if you're trying to start out, you're on a budget, you want to economize, that's fine.
00:24:25
Speaker
But of course, you're compromising on the time and the speed at which you can take products to the market. But once you have your design sort of fixed, right, and your customers are coming back to you with volumes, you should, you know, try to book the capacity that you need for the founder.
00:24:37
Speaker
And one wafer gives you how many chips? It depends entirely on the kind of chip, right? So, for example, let's say you're you're doing something extremely complex, right? so And so it depends on how much is the real estate. So optimizing for chip area is really a big deal, right? Because, you know, if I could pack, I'm booking one wafer with a foundry, right? If I can get, let's say, you know, 200,000 as opposed to hundred thousand I've got... In fact, my effective cost has become half.
00:25:02
Speaker
and So, you know, kind of optimizing for size is very, very important. It depends entirely on the functionality. There's no one answer that that I can give you, actually. So, you're saying a wafer, which you said could be, like, say, 8 inch by 8 inch or 12 inch by 12 inch. That sized wafer can give you more than 100,000 chips? Depends entirely on what you're putting in. Like like you're saying, each chip is, like, maybe, and like... the the size of the nail which I cut off when I'm cutting my nails. Is that the... like it it It could be. ah But you know if you take a look at some of the complex... micro so Look at the microprocessor that you have you know on your... If you ever open up your cell phone or your laptop for that matter.
00:25:43
Speaker
You'll see something that that could be something like 4cm by 4cm. So that's how big it is. Now, not all of it has, you know, it's not the size of the chip inside. There's there's like maybe your co-package, right? That's the casing and the cooling. Yeah, that's the casing and all of that. So let's say that you have a 3mm by, but anyway, let's say that you have a, or in that 4cm by 4cm piece, let's say you have 2cm by 2cm worth of silicon.
00:26:07
Speaker
Now, how many 2 centimeter by 2 centimeter can you pack in an 8-inch wafer? There are online calculators that will actually help us do this. so But if I'm not doing a microprocessor, let's say that you know I'm doing something relatively less complex.
00:26:21
Speaker
I just want to build, let's say, I want for whatever reason to build that USB readout interface. right And that's pretty much all I wanted. I am not looking for a microprocessor. right So let's say I just want to do one readout circuit of some kind.
00:26:34
Speaker
okay So that may not be the same size. I could do something in, let's say, a few millimeter, five millimeter by five millimeter. So that's, you know, it's a different kind of conversation. right Then you will get the area is now, you know, two centimeter, two centimeter, five millimeter, five millimeter. So 16 times lower.
00:26:50
Speaker
So you will get 16 times more chips out of that. OK. Right. So there's no one size fits all. Fascinating. Okay. And ah you spoke of the process node at three millimeter, five millimeter or nanometer. Sorry. what what what is the What is this process node? Three nanometer, five nanometer? What are these terms?
00:27:11
Speaker
Yeah. So what this means is that this is the smallest feature size that can be printed. So if you think about it, like we said, so the process of making chips is like, you know, you think about it like screen printing just to just as an analogy, right?
00:27:25
Speaker
So what is the smallest dimension you can print? So that would be this 3 nanometer. When it's 180 nanometer, it used to be 180 nanometer, right? So that was where it is. So 3 nanometer, 2 nanometer is extremely small.
00:27:38
Speaker
So, you know ah today the the atomic the atom is is about you know has about 0.5 nanometer, 0.3 to 0.5 nanometer is a good estimate for you know how big atoms are.
00:27:52
Speaker
So, when you're talking about two nanometers, you're talking about four you know four itemtoms four to seven atoms. So this is really getting close to what is humanly possible here. and or Not just humanly possible, what is physically possible. you know So that's the kind of you know control that people now have over silicon manufacturing. like so Is it fair to understand that a 3 nanometer known means that the transistor is 3 nanometers?
00:28:18
Speaker
Yeah, it it used to be the case that the the place across which we we spoke about how current is going from point A to point B. So, normally this distance that the current has to traverse, you can think about it as 3 nanometer. It used to be the case, but of course, nowadays it it has a lot of other ancillary meaning. This used to be smaller, ah you easier to manage than when we were, let's say, 180 nanometer literally did mean that the electron will go across a 180 nanometer.
00:28:42
Speaker
That used to be called what is the gate, the dimension of the gate. Okay, so gate length. So which is not the case anymore. So now it means something a little bit more complicated. But yeah, so put it to put it very simply, yeah, it's it's roughly that.
00:28:56
Speaker
Okay. And so essentially it all boils down to how much processing that two inch into two centimeter into two centimeter chip can do. So the smaller the node, the more processing that same real estate can handle basically. Correct. Because you can now pack more transistors in and they will all in turn, you know, consume lower amounts of power. So this is Moore's law, right? So it was not just about packing functionality, but they also, you know, consume a lower amount of power because, you know, you don't need the same amount of If you think about it, you don't need the same effort to conduct charge across a distance that's this big versus that big. right So you can you can drop all your voltages that you're using.
00:29:35
Speaker
right So all of that helps. So you can literally pack more functionality and get more power efficiency out of chips. So that's that's the way that, you know that was Moore's Law of scaling.
00:29:48
Speaker
and So roughly, you know you could double the number of transistors in a given real estate, so let's say over about 18 months or so so. that used to be how it went. And I think that thanks to many, many, many technological tricks and visitory, this process has been kept going for you know much more than people thought it could be kept going for.
00:30:07
Speaker
Are we reaching the end of Moore's law? You know, I'm very cautious about saying any such thing because I've been hearing this statement for the last 10 years, right? So, I'm sure. Yeah, absolutely. There's always some amazing piece of engineering that comes up, which kind of extends it. Now, I think people are trying to do 3D integration so that you can actually stack transistors. Okay. But it's no doubt that it's becoming a lot harder to do.
00:30:28
Speaker
right So, there's no doubt about that. But yeah, I mean, we may be closer to tapping out, but you know, you never know. Fundamentally, is a CPU chip different from a GPU chip?
00:30:40
Speaker
Is a CPU different from a GPU? A GPU is actually, you know, ah it is a CPU meant for certain functions, so let's put it that way. So it does a few things extremely well, right? So I think that all the buzz now is because of AI, right? there's a lot of talk about how GPUs are.
00:30:56
Speaker
So GPUs implement a certain kind of mathematics much better than than CPU. So that's that's where GPUs are very useful. It's of course, you know um now given the sheer number of use cases, I think that people are designing also more AI-specific chips, so which are which are you know capable of doing things like matrix multiplication a lot easier.
00:31:17
Speaker
So those are things that you know those are mathematical treatments that are more important for AI, and therefore it has a physical manifestation which looks like a GPU for a variety of reasons. right So, for a variety of reasons that are above my pay grade, so to be very frank with you. But physically, they look the same. like Physically, they look the same. It's the same transistor. If you dial down, it's the same transistor. It's just implemented implementing a different kind of functionality. it's more It's optimized for a certain kind of computation.
00:31:43
Speaker
So that's the main difference. A CPU is a little bit more of an all-purpose, it's more of an all-rounder.
Applications of Gallium Nitride
00:31:49
Speaker
right ah GPU is a little less of an all-rounder and then it does you know certain, like I said, so it can implement certain mathematical functions a lot more accurately and fast when compared to CPUs, which makes it useful for AI as well.
00:32:01
Speaker
Okay, got it. got it okay So, ah when did you place your order for the first ah that that chip that we were speaking about, the the full chip, that 8-inch into 8-inch chip?
00:32:16
Speaker
ah Sure. So, today gallium nitride is two things, right? So, gallium nitride today is primarily used in two major use cases in electronics, right? So, before electronics, gallium nitride was actually used in LEDs. So, you see the white light LEDs. So, if you combine red, green and blue, you get white.
00:32:34
Speaker
So, the green and blue LEDs were solved problems thanks to gallium nitride as a material. So, you had to move away from, you could do a red LEDs on gallium arsenide, which used to be previous generation technologies. So if you wanted to solve for the green and the blue, then gallium nitride and alloys were what were used to solve that problem.
00:32:50
Speaker
So that's the optoelectronic use case for gallium nitride. So that's not something that we work on in Agnet. But you know in 2014, the Nobel Prize in physics was given for the efficient for the invention of efficient white light emitting diodes, two, three Japanese scientists were doing this.
00:33:05
Speaker
So that's the genesis of gallium nitride as a material. So all of this exploration started in late 80s, early 90s. right So the electronic use cases for gallium nitride today are primarily in two broad areas. one is One of them is radio frequency and the other is in power electronics.
00:33:21
Speaker
So in radio frequency, gallium nitride is typically done on silicon carbide as a substrate. So you take gallium nitride, you put a few micrometers of gallium nitride material on top of a silicon carbide wafer.
00:33:32
Speaker
right And that wafer globally is on 4 inch moving to 6 inch. So that's the scale. So it's not yet at 8 inch, 12 inch, etc. So, and then you have for the power electronics applications, you primarily do gallium nitride on silicon for a variety of cost reasons, right? Because silicon is probably, you know, as a pure material, silicon is probably the cheapest pure material you can buy today, if I can put it that way, right? Thanks to 60 years of, you know, technology being there in order to refine and make larger and larger sizes.
00:34:02
Speaker
So today, that gallium nitride on silicon is worldwide on, you know, um it's either on 8-inch primarily, and now it's slowly also moving towards 12 inches, right? So, but you also have people on older legacy nodes doing 6-inch, right? So, which is still okay for gallium nitride.
00:34:19
Speaker
So that's the wafer size, actually. So for gallium nitride on radio frequency as well as on power. Now, when is it that we did our first chips, right? So as I said, we we tend to but we try to manufacture our own designs wherever possible, right? so So we made our very first prototype using our internal foundry in late 2022.
00:34:40
Speaker
So it took us a solid two years in order to get there. So where we were actually having the first proof of concept saying that, okay, look, all this technology can actually result in meaningful products for it for a user. right So that's the first chip that we actually could hand out to somebody and say, please test this.
00:34:55
Speaker
And once they did, I think that they could see that, okay, at least things are working here. And then, you know, then you have a much deeper conversation with customers as to what exactly their chosen Calibre-Masuride chip is, right?
00:35:07
Speaker
You could have given somebody something, a chip with 10 watts of power, but maybe they want 100 watts, right? So they tell you that when you actually put something on the table and then you can talk about it. Otherwise, it's a very theoretical discussion. So you also have to prove that the technology that you have, the baseline technology that you have can result in useful products.
00:35:23
Speaker
So that we could do towards the end of 2022. Isn't a foundry like a billion dollar investment? Yeah, that's ah again a great question. So in Silicon, if you want to set up a cutting edge foundry, I think there are a lot of reports that will tell you this will cost anywhere close to $20 billion dollars today.
00:35:38
Speaker
right So, do a meaningfully sized silicon foundry. right So, ah firstly, you know in case of gallium nitride, that's not necessarily true. So, for example, the gallium nitride in the radio frequency space is a lot more aggressive in terms of dimensions.
00:35:53
Speaker
So, there, if you could do even 50 nanometers, if you could do 500 nanometers to 50 nanometers, that would be great. It covers you know ah practically the entire usable a range of gallium nitrates functionality.
00:36:06
Speaker
So 50 nanometer technology, you know so my my first job out of ah out of after I graduated from VTech in engineering, so it was with IBM, and I was a silicon chip designer.
00:36:16
Speaker
right so And I was working on 45 nanometer node technology. And this was back in 2008. So that's ah kind of, and we are talking about how that's ah probably, you know, if that's the smallest dimension and I could print today on gallium nitride, we'd be golden in the radio frequency world, right? So it's very, very different use cases, right? So you need very, very different requirements in terms of what is that critical dimension I need to print, right? So it's not as demanding as silicon. So you don't really need $20 billion dollars to do that, right? So you could set up a reasonably sized gallium nitride foundry today for close to half a billion dollars.
00:36:49
Speaker
so it's big But ah you don't have half a billion dollars, right? like we We don't have half a billion dollars. no how do you What internal foundry are you talking about? I didn't get that. Great. So, you know, so the Ministry of Electronics and IT has actually funded IAC's incubation arm, Foundation for Science, Innovation and Development.
00:37:07
Speaker
So they gave them both 300 crores in order to set up a gallium nitride foundry. pilot production factory here okay so that about 50 50 million dollars when when it was uh sanctioned way back when right uh this is small is a pilot production line as i said you could do you know prototypes and small volume production here and this is actually located on the iac campus right and uh this is mainly because you know if you think about it even in the kind of fab projects that we have announced in the country today whether it's OSAT or whether it's, you know, the FAB that's coming up, right? The technology partner is always somebody elsewhere, and somebody with technology, proven technology at the end of the day, right, which which has never been Indian, right? if you so if you if you try to take one step back and try to answer the question, what will it take to set up foundries and semiconductor manufacturing using indigenous technology, even for the manufacturing, not just the design. Design, we've we've always been as strong strong as a country, right? So even if you want to do that manufacturing piece of it, what will it take?
00:38:04
Speaker
So what will it take for you to set up a foundry? Let's say this $500 million dollars foundry I'm talking about, where you need a technology provider who's actually willing to give you that manufacturable gallium nitrate technology, right? So that is what Agnet is going to be.
00:38:18
Speaker
So so that's that's exactly what our vision is. So we don't have 50 million, so but we need a much smaller thing in order to prove that the technology is even manufacturable in the first place, that useful products can come out. right So those are the questions that we are trying to address. Interesting. So IAC has this gallium nitride foundry on which you made your first tip chip wafer. um What does that foundry look like? What is it like like what is the input? What are the different processes that actually are happening inside it? Just describe to me in a physical sense.
00:38:53
Speaker
Sure. So we'll we'll take radio frequency products as ah as a gallium nitride radio frequency products as a good good example. If you think about the inputs, the You know, I said gallium nitride is done on silicon carbide wafers for radio frequency. So that's the base material.
00:39:10
Speaker
What comes in is actually a silicon carbide wafer. And we have the technology to actually first, you know, make that silicon carbide wafer into a gallium nitride on silicon carbide wafer.
00:39:20
Speaker
You have to put a few micrometers of gallium nitride material on top. that' started very know for How is that done? It's actually done using means of something called chemical vapor deposition.
00:39:31
Speaker
So, you know there is that is an atomic arrangement layering. So that actually happens. So it's not just sprayed on top. So it's actually kind of a, think about it as an atom to atom alignment. So silicon carbide has a certain atomic structure. So that means that the atoms are sitting in certain, like a very nice crystalline configuration.
00:39:49
Speaker
And you have to make sure that your gallium nitride atoms also orient in a certain way so that they are all kind of sitting exactly matched. Otherwise, it gives rise to defects. And every time you have defects, you have worry about, okay, um am I going to get all my current out of here?
00:40:02
Speaker
What happens to heat? Is this going to be reliable? right So you want to make sure that you have that kind of ah like ah atomic orientation also to be perfect. right So this process is called epitaxy. right so and which Is it like dipped into some solution where this...
00:40:16
Speaker
happens No, actually what happens is that it goes into a furnace, right so which is extremely high temperatures. It could be like a thousand degrees. And then you you're passing, let's say, a source of gallium and a source of nitrogen.
00:40:28
Speaker
And that at that high, extremely high temperatures, they combine and you have gallium nitride, which then sits in a certain way on top of the silicon carbide wafer that you have in here. And you're doing this, you know, multiple wafers at a time and they're all sitting in a certain way and then it takes, you know, let's say about four hours for you to do this entire process and then you get like anywhere between one to eight wafers with gallium nitride on top.
00:40:51
Speaker
right And typically, as I said, this is four inch four inches now for GAN and silicon carbide. So that's one input. So of course, the other input is I i told you we have a source of gallium, a source of nitrogen. So all of that needs to be put into place. Where do you get the ah silicon carbide wafer from?
00:41:07
Speaker
Is that domestic? It's not domestic. Unfortunately, nobody's yet making silicon carbide. But you know recently there was this project which has been announced in Norisa in order to manufacture silicon carbide as well.
00:41:19
Speaker
So hopefully they will be making wafers, in which case you know if they make the kind of wafers that we are interested in, that supply chain could also
Global Semiconductor Supply Chain
00:41:25
Speaker
be indigenous. And what about Silicon Bifots? Are they made domestically? Silicon wafers are also not made domestically. So they're all, you know, right now we don't have much of a presence in the whole, India as a country does not have much of a presence in the whole manufacturing value chain at all. Literally starting from zero probably, right? So, you know. Do these come from like TSMC, Taiwan and like Samsung? TSMC also does not make their silicon wafers. So they are very good at taking the wafers and then, you know, have they have all the intellectual property and the e equipment in order to convert that into, let's say, process chips.
00:41:56
Speaker
So there are very specialized manufacturers of silicon wafers worldwide. Which countries? That's pretty much all they do. um It could be anywhere. so in Primarily the expertise is concentrated in the US, is in Europe, it's in you know ah places like Japan and in East Asia now, quite a few East Asian places as well.
00:42:14
Speaker
right And that's been the historical norm. So these are all countries which have been extremely strong in semiconductor technology, US, Japan in particular, Europe to a certain extent. And then, you know of course, you have the whole East Asian ecosystem that's now coming up.
00:42:26
Speaker
So coming up in terms of also doing the the semiconductor manufacturing, I mean, they've historically also been very strong in the whole packaging and assembly and all of that. right And of course, you have Chinese players also now making silicon wafers, right? so And all of that is also rapidly growing.
00:42:40
Speaker
and is it Is this easy to make or hard to make silicon wafers? A silicon wafer? um yeah It depends, right? So if you know how to make it, all everything is easy to make. But today, it's it's mainly, um it's not a raw material problem, right? Because if you think about it, a silicon wafer is probably the purest material that can be made today, right? So it it could have, you'll you'll have less than, ah you know, ah practically if you take a silicon wafer, you'll have less than 100 defects, atomic defects of any kind across a centimeter squared of area, right? That is extremely small.
00:43:12
Speaker
Okay, so, and and this is like, not only is it chemically pure, it's also atomically pure. So that is extremely difficult to to make if you just think about it, right? So that's that's like extremely difficult to make.
00:43:24
Speaker
The properties of steel changes if you add 1% of some impurity to it. Right, right, right, right. If India was to want to manufacture silicon wafers, then... ah Would it need money or would it need an IP partner who will give the IP? definitely Definitely needs both. But money, I think, is the lesser of the two constraints. You definitely need an IP partner. And that's that's very important. And then, of course, once you start making these silicon wafers, somebody is buying a wafer, which means that you'll have to grow a big ingot.
00:43:54
Speaker
worth of silicon and then you know have it all cut up in a certain orientation so that it's atomically flat, polish it. And then so those are all like, you know, very interesting add on manufacturing operations that need to happen for somebody to even get a vapor to begin with. that So this is necessarily a very complex supply chain. So then you start thinking about, OK, if I'm going to cut this, cut this, let's say big, they call it a pool.
00:44:15
Speaker
okay So it's it's very big, it's much taller than I am. okay And it's like 12 inches in diameter, so think about it. And then you know all of that needs to get sliced out. So if you're going to slice out, you need equipment to do that slicing.
00:44:28
Speaker
And you need the confused let's say you need you need blades or you need lasers in order to do that. And somebody is making that and then you want to polish it. right So even let's say I want to do let's say a very very atomically smooth polish.
00:44:39
Speaker
So there is a tool which does this. And then you have consumables for the tool. There's something that's grinding, lapping, if you think about it, right? So somebody is making that so that your thing can be atomically smooth as a slurry that's used.
00:44:51
Speaker
So the value chain is, you know, very, very disaggregated across the world. and So it's impossible to get to 100% sort of localization for anything. right So you should pick what is key and then see if you can actually get there. And that depends on you know whether you have some indigenous technology available. Do you have the, you know let's say some motor supply chain already in place?
00:45:13
Speaker
right And you have you know you can even take a step back and say, do I have certain critical minerals and resources that I need? Do I have it here? too Do we have the manpower to do that? right So it's impossible to be 100% indigenous in semiconductor technology.
00:45:24
Speaker
so But we can get close. Get close in terms of important jobs. Gallium and nitride, are these rare earth? Nitride, what is it? Nitride.
00:45:35
Speaker
Nitrogen. like nitrogen isgen Nitrogen is not. So that's like, yeah yeah there's no nitrogen is lack of it. And gallium, what is it? Is it a rare earth? Gallium or gallium is not a rare earth metal. um But, you know, it is also a fact that I think, you know, ah China has, it's a critical mineral. And then China has about, I think about 87 or 90% of the world's gallium reserves, known gallium reserves.
00:45:59
Speaker
Okay. So that's another, that's that's true for most of these, you know, kind of rare earths. Gallium is not a rare earth, but yeah, it's it's a critical element. Got it. So coming back to that foundry, what a foundry does. So it it goes through that high ah temperature furnace and then it comes out with a coating of gallium nitride, the silicon carbide.
00:46:20
Speaker
ah wafer has a coating of calcium nitride on it. Then what next? So next it goes through, you know, um for example, it go through about 200 distinct semiconductor manufacturing steps.
00:46:32
Speaker
So you have a design that you put into the foundry, right? So that's, you're saying that, look, I want this to look a certain way, right? So this is this is the kind of, you know, functionality it needs to implement. And that's what it means for me physically. I need to put layer one, layer two, layer three, whatever it is on top.
00:46:47
Speaker
And I need to remove, you know, some... some These layers, how are they put? Is it like etched through a laser? or how How is that? how So normally, this is where a process called lithography comes in. And lithography is, you know, writing with light. So that's, ah you know, that's how patterns are transferred onto the wafer physically.
00:47:06
Speaker
once you and Once you have that pattern, then essentially you need to keep that pattern there in some way. lithography will typically use certain organic chemicals in order to make sure that certain areas are open on the wafer or certain areas are closed now let's say certain areas are open therefore you can work on them so what do you do you can either add material or you can remove material put very simply right So it's like, you know, imagine having a shovel and then you can scoop wherever there's some sand.
00:47:31
Speaker
But if there's concrete on either side, you can't do that. So let's say I put a concrete lid on it. You know, I have a sand pit. I put a concrete lid on it. I have a cutout so I can scoop wherever there's sand. Right. So once that's done, right. So or you what you can do is you can add material to it.
00:47:46
Speaker
So I could add other kinds of impurities if I wanted, right? Or I could add, let's say some other metals, or I could add some you know insulators, whatever it is, right? So I could i could be adding material. So these are the two things that you can do.
00:47:58
Speaker
And then you need to make sure that this happens over a sequence of, let's say about 15 to 20 different lithography steps. So you're trying to put in 15 to 20 different kinds of patterns. They're all aligned in a certain way.
00:48:09
Speaker
right So with respect to each other. And that's again very important. And typically when you run through this process from end to end. From the time your gallium nitride coated wafer goes all the way to the end. Where you have chips on it.
00:48:21
Speaker
You could be running through you know a few hundred semiconductor manufacturing steps. All of this needs to be done on specialized equipment. and You need IP to do this. You need trained manpower. right And you're working typically in a clean room.
00:48:33
Speaker
right So that's the way that's what it takes in order to implement all of this. You know, there are these ah reels in which they show somebody spray painting on a, like there's a pattern which is cut out from a paper. He spray paints and then another with a different color and the third and a fourth and a fifth. And so are multiple times he's spraying on it. Each time the pattern is different. Then in the end, there's a reveal and it is like some beautiful picture. So it is something similar happening here. It's exactly it's exactly similar with nanometer tolerances.
00:49:05
Speaker
Okay. Okay. Okay. yeah okay Fascinating. Okay. so okay so So that gives a tip. And so this IAC has this facility to create a wafer of gallium nitride. But you said this facility is only good for pilots, not for bigger quantity. Why is that?
00:49:26
Speaker
Like for bigger quantity, you need more machines or some more automation. You need more machines, more more throughput. right okay so So that's what it it actually needs. So it's good for prototyping and doing small volume production.
Agnit's Strategic Product Focus
00:49:38
Speaker
ah But of course, what is small volume and what is large volume is very different in the silicon and calcium nitride worlds. and So that's also another thing that needs to be there. So what we have here is actually a decently sized facility for gallium nitrate. But you know it's it's very good for doing like pilot production.
00:49:52
Speaker
Now, how you will ramp this up, let's say you want to make let's say we want to make a make a product today for consumer fast charger, your cell phone chargers. So we couldn't cater to the volumes, assuming that we could win designs. right So we couldn't cater to the volumes that will come up.
00:50:05
Speaker
you will need 10 million, 20 million depending on how successful this is. You'll need about 10 million to 20 million chips a year, right so which is something that cannot be made on the facility that was housed here.
00:50:15
Speaker
right So in which case you have to take the technology and have it productionized elsewhere. So that's that's what I meant when I said it's a pilot production facility. Okay. And so you have found product market fit. Yeah.
00:50:30
Speaker
Which area are you doing? So we are primarily doing calium neri gallium for the strategic sector. By strategic sector, I mean defense space and aerospace. So we have about three products now which are running field trials with customers.
00:50:42
Speaker
So that's a good good space to be in. all right And so we expect that this will ramp into volume production over the coming year. So that's that's where we are. So that's also the sector that we started to work on.
00:50:53
Speaker
We also started working with the sector for a bunch of reasons, mainly because, you know, um so one of the reasons is that gallium nitride is also export restricted. Gallium nitride components in radio frequency are also export control.
00:51:03
Speaker
So what it means is that, let's say you are the Indian armed forces, you could go to one of the large, you know, e equipment manufacturers from the US, from US or from America, and then you can buy a radar, next generation radar, which has gallium nitride components.
00:51:17
Speaker
But if we want to design and build that radar here, then your access to gallium nitride components is not as issue. So you don't get the components that you want in the volumes that you want in the timelines that you want.
00:51:28
Speaker
So that's the reason that we started working you know with the strategic sector. And you know it was also the need of the hour. So so that's something that we that we kind of value. So it was the response was good. And you know so that's where we have the first chips in.
00:51:43
Speaker
But you know if you also think about it, and so that's where we are today. We also have a small development ongoing where we are trying to build a prototype mainly for a telecom product. And this is going to be more useful for things like five g right private networks, etc.
00:51:58
Speaker
So that should hopefully also be out in the next nine to 10 months. So hopefully we'll have something slowly diversifying to other sectors as well. So you have three CHEPS currently being prototyped and tested on field. Correct.
00:52:11
Speaker
Just tell me about each of them. What exact use case or which device is it powering? Sure. One of them is for a jammer, right? So this is more for a jammer use case. So, and of course, jammers have become all the all the news these days after recent developments, right? So, you know, there are, gallium nitride is primarily in radio frequency systems. Gallium nitride is a power amplifier.
00:52:33
Speaker
So what I mean by that is, let's say you have a small information signal that comes in and typically this could be, let's say a few milliwatts and then you want to, you know, you have an amplifier that as the name indicates, it takes a small signal and then it kind of boosts it up and then you can connect it to an antenna and the antenna will broadcast that signal.
00:52:49
Speaker
Okay, so that's what a power amplifier does. So the power amplifier in all the next generation wireless systems is moving towards gallium nitrate, again, for the same reasons that we discussed earlier. It's compact, it's efficient, it can handle higher amounts of power, right so which is something that everybody likes. So you know so that's the reason that this this kind of transition has happened.
00:53:09
Speaker
And in in a jammer, for example, you're trying to, let's say you have a bunch of drones coming at you, right? So you want to interrupt all of their communications so that they don't get signals anymore. like So that's what a jammer will do. right So there again, you need a power amplifier, which is able to kind of blanket the spectrum, so to speak, for a given range.
00:53:26
Speaker
So that's what you know gallium nitride is again very good at, because again, very compact, very efficient systems. right So so we we have one pilot there. The other pilot is more for a radio link.
00:53:38
Speaker
right So let's say you want to communicate today between two points. Again, you have all the same things. right So you would you would like as high a range as possible, number one. Number two, again, you want these systems to be extremely compact. So you maybe you want to mount the system on a drone or on some portable system or somebody who's kind of a soldier who's kind of wearing it and then you know running around. So those kind of things where, again, compactness is important, efficiency is important. right So again, gallium nitride is good there. So that's another use case that we have.
00:54:04
Speaker
I didn't get this radio link. So you're saying like, say, a walkie-talkie. Yeah, it could be a walkie-talkie. So walkie-talkie is actually a very good example. It could also be something, let's say that, you know, you have a field unit out there which is trying to communicate with other field units. So you don't have communication. So you'll to put your own, let's say, mast and then you have ah whatever is the equal equivalent end of a walkie-talkie on a larger scale.
00:54:23
Speaker
Doesn't satellite give signals everywhere? like Yeah, and it does. It does give signals everywhere, but it again depends. Do you want to, whose satellite are you going to use? Okay. If it's only yours, then you know you have to, there's a certain amount of wait time. right so And then maybe you want like ah to communicate with, let's say, two units which are kind of deployed a few kilometers apart, want to communicate with each other in the secure way.
00:54:46
Speaker
Okay. So this is this is also, it's kind of the walkie-talkie example, you know, scaled up. Scaled up. Got it. So that's there. and um And of course, with drones, like communicating with drones. Absolutely. Communicating with drones, right? So that's ah that's an important one.
00:55:01
Speaker
And the last one is mainly for, ah you know, like ah a use case where, which is more like ah like video trans transmission, right? So which again, if you think about it, it's like a point-to-point radio. So let's say you are today on a drone and you want to actually, ah let's say, record video and you want to send that back to base. right And let's say you want to do this over, let's say, 100 kilometer range.
00:55:22
Speaker
right So if you use a conventional silicon based solution today, depending on the sizing, etc, maybe you know an equivalent sized silicon part, maybe will give you about 20-25 kilometers of range. But what do you do if you want, let's 40 kilometers, 70 kilometers, 80 kilometers, 90?
00:55:36
Speaker
So in that case, you know using a gallium nitrate power amplifier makes a lot of sense. And again, if you think about it, you're on board a drone, you're presumably running off a battery pack of some kind. So you really need very, very efficient power amplifiers. These are very power hungry. right And you need to be extremely compact. Again, you're on top of a drone. You need the kind of you don't need that kind of rage anxiety as well. right so And this you know the broader use case is not just for things like surveillance. right So it's it's for things like monitoring.
00:56:02
Speaker
You could be monitoring an oil pipeline. You could be monitoring, let's say, a grid or or you know highway construction or whatever it is. Provided you're allowed to transmit you know a certain amount of power. Then you want the you want to transmit more amount of power in roughly the same footprint on board your drone.
00:56:16
Speaker
So, Gallium Nitrate again is very good as a power amplifier solution for all of this. So again, if you think about it, right, so the same thing that we discussed last time. So these systems also have a lot of control circuitry. They also do logic and memory.
00:56:27
Speaker
All of that stuff is still on silicon, right? So that's what's telling the Gallium Nitrate, you turn on, you turn off, etc. But Gallium Nitrate is what's doing all the amplification. right So that's primarily where we are. So these are three use cases for which we have systems which are currently being piloted.
00:56:42
Speaker
And so you know initial feedback has been good. So we are we are hoping that all of this will ramp into volume production shortly. so How far can a radio signal travel? Depends on what frequency you're transmitting it at and how much power you start putting out. Right. So, you know, that's, I mean, a good example is ah your your FM radio, right? So, you know, if you're at a few hundred megahertz, et cetera, you can hear an FM radio signal for you tens of kilometers, if not more.
00:57:08
Speaker
okay So, and of course, if you have repeaters, you can always boost the signal. So, and if you go lower, you can the signal will transmit. If it's a lower frequency, then normally you're able to transmit for much longer.
00:57:19
Speaker
But you know if you reduce the frequency, right so it becomes much harder to transmit. But at the same time, you can get much better resolution. So it's kind of what happens in ah in a radar at the end of the day. right So if you if you want a radar, which is picking out, let's say a larger target, you can operate in, let's say, a few gigahertz out of a month.
00:57:37
Speaker
But let's say now you want to pinpoint it drone. or you want to pinpoint smaller targets, right? So then you have to have a finer beam, so to speak. And that happens when you kind of lower the frequency. And sorry, when you increase the frequency, so you lower the wavelength of the beam, right? So it's able to pinpoint.
00:57:51
Speaker
If you increase the frequency though, you know, the amount of attenuation that the signal encounters will increase. So this is the reason that light doesn't bend over, bend across corners, right? But you can hear your FM radio even if you sit inside a building.
00:58:03
Speaker
Right. So because light is a few terahertz sort of frequency, so it doesn't, it cannot really turn a corner. Right. So it's more straight line communication. Okay. um So who is ah piloting your chips? Are these ah suppliers to defense forces, like say, like say an idea forge which supplies drones would be ah testing your chips or like who's who's testing the chips?
00:58:26
Speaker
These are all suppliers to defense forces, right? So I won't be able to name names because these are again early days and given the nature of what we are doing, we'd like to you know stay quiet about that. But you know what we've actually seen is that there's a lot of you know um enthusiasm, especially also from the defense private sector, in terms of running running pilots with companies like ours. so And they're also kind of in a similar board where they're trying to develop solutions quickly.
00:58:51
Speaker
So you know even things like the energy levels match sometimes, which is very good. right So that's where we are. and again, in terms of the defense PSUs that we have, they've been extremely supportive in the kind of you know constant support and encouragement and you know kind of product roadmapping that needs to happen.
00:59:06
Speaker
So the defense sector has been you know a bit of
Building Customer and Market Credibility
00:59:09
Speaker
a revelation. I think before we got into it, people always kind of warned us saying that, look, be careful because you know historically things have been delayed and you know there's always a complaint that people don't move quickly.
00:59:18
Speaker
but I think our experience has been something different. I think there are always constraints in any, in any system, but you can see that the intent of the policy push is very real, right? So, which is what we, we kind of like about this. And we've seen this over the last couple of years, right? So quite encouraging, honestly.
00:59:33
Speaker
How did you learn to sell? And, you know, how did you get these pilots in place? You're like a bunch of academicians. You've never had to sell other than maybe selling a research paper or something like that. Uh, How did you track crack that as a business, as a company?
00:59:51
Speaker
Yeah, I mean, ouch, but but hey, you know, actually, I have a different take on it. I think that academics always kind of sell because they also, you know, for example, when I was when i was in when i was doing my PhD or postdoc, you know, if something very fundamental. For example, you go to conferences and you talk about your work.
01:00:09
Speaker
yeah So let's say you've done all your work in the lab, you're going to go and talk about it. So what is that, if not salesmanship? You're not going to hear those. I mean, true sales is when you hear... So somebody will probably stand up in the audience and say, you're completely wrong. So I've been in conferences where that has happened.
01:00:26
Speaker
right So it's pretty unfiltered and the feedback is very raw and immediate as well. Right. So it's not that we are we ah polite. I mean, academics necessarily aren't a polite bunch of people. Right. So, you know, ah so that's that. What?
01:00:37
Speaker
So, you know, whether it's on, if you think about the other thing that normally happens, right. So scientists will typically do their work and then they'll write a paper because you want to disseminate what you do. And then you'll send it to, let's say, a journal. You'll get rejected a bunch of times.
01:00:50
Speaker
right So whether it's a rejection, whether it's trying to communicate your ideas in written form or oral form is actually something that we go through. It's just that you know we probably don't think about it along this lens. right But this is this is what it is fundamentally.
01:01:01
Speaker
But what is more important, you know what i mean the the mind shift is that at the end of the day, you know in in let's say in the business sense, where we are is that we want to add some value to somebody else. right So you to be able to put yourself into your shoes rather than kind of narrowly say that, okay, I've done this, I'm trying to you know ah get this out you know in in, let's say, published form of conferences. that's ah That's probably the different mindset. You're trying to put it into you know you're trying to put yourself into your customer's shoes and not just from that from the moment of that purchase thing. It's what does it mean for you? Why do you want to buy?
01:01:35
Speaker
right So those kind of things, you know, it's a little bit of that that human element, that little bit of psychology is not something that, you know, I suppose academics generally think about. But you learn, you have to think about this and you do you do learn. right So it is, I wouldn't say that it's easy. It's not easy for anybody.
01:01:52
Speaker
I think that once you do that, i think it's it's better. How did you open those doors and like, you know, I mean... when you want to get your work published, there's like a fair straightforward process where you submit a form, whatever. But when you want to sell, there's no form you fill up to send your picture across. like It's a lot of like hustle to get in front of the right people and get them to hear you.
01:02:18
Speaker
Absolutely. I think what we've seen is that firstly, you know, um if people realize that this is a serious effort, right So I think that you are naturally, you know they want to talk to you, they want to know what you're doing, they want to know how they can help right from a customer standpoint. And that also you know kind of validates the problem that you're trying to solve.
01:02:37
Speaker
So I can honestly say that there's nobody who is working in in, let's say, the strategic sector whom we wanted to get a meeting with, whom we did whom we did not get a meeting with. I think that just goes to show that the problem that we are trying to solve is a very immediate problem for them.
01:02:50
Speaker
They really need a solution. Then the next question is, you know, are you the right person to solve? Why should I talk to you? Are you an academic effort? Right. So those kind of questions we we got a lot and we continue getting them even today, by the way, given that we are still located on ISE's campus. Right.
01:03:03
Speaker
So that's something that there's no getting around it. So the sooner that you can communicate that you are, you know, kind of your own independent entity and you are developing a product that's of some value to them, right? And the sooner that that can get across and that this is not being done in an academic sort of a scale, this is a business. And if you don't sell, you will die, right?
01:03:22
Speaker
So I think once once people realize that and you're able to communicate that very clearly, then i think that the conversation, of course, you know, is is a lot more productive, right? No doubt about it. So we we have gone through that. So that has been kind of what it's both a negative and a positive to be incubated at an academic institution.
01:03:39
Speaker
You know, the the positive is that, you know, there's like everyday stuff of running operations wise, etc. Right. So the everyday indignities of running a company, put it that way. Right. So, you know, am I going to get electricity today? Is somebody going to pull my water connection? um ah you know You know, what happens to my solid waste disposal?
01:03:56
Speaker
Those kind of things are all handled more at an IAC sort of a level here for us. Not just for us, for anybody who gets to incubate an incubator of some kind. So you don't have that. But at the same time, you know when you go and have conversations with customers, you know the quicker that you can communicate, you are not this is not a research project.
01:04:12
Speaker
You don't have students working on it. This is a serious team that knows what it's doing. Here here is some proof that you can put on the table. It's very important do you have that proof. and If you typically go with pitch decks, I mean, pitch decks are good for investors. they're not good for customers.
01:04:24
Speaker
You need to have something that's workable, something that's working. So when we started putting data sheets and putting like evaluation boards on the table with our customers, we got taken a lot more seriously. thanks It took us like ah you know slightly more than two years to do start doing that. Does IIC have a stake? IIC does have a stake. IIC has incubated us, supported us. you know and kind of so that is That has been very important for us. IIC does have a stake in the company as well. They also gave us a very nice seed grant initially to start out with. and Today, I think we we use the facilities, but anyway, we pay for the facilities that we use.
01:04:58
Speaker
right And in the initial days, you also get some amount of subsidized use of all the facilities. And slowly, you know, in kind of like year five, you're paying what market pays. But that initial period is also very important because that's when you want to do your MVP, that kind of initial support, that viability gap funding that people call it. and So especially in deep tech, that's a very real thing.
01:05:16
Speaker
yeah And so ISE incubation solved a lot of that for us. ah What's the way forward now?
Scaling Challenges for Gallium Nitride
01:05:23
Speaker
Let's say these three chips, if they see success and you need to manufacture that scale, what kind of scale do you need to manufacture that? Can the IAC foundry handle it or where will you manufacture it from?
01:05:35
Speaker
Sure, for the strategic and the telecom sectors, the ISC foundry has enough capacity to actually, you know, manufacture here, right. so So, that's okay. In the, like I said, the more the consumer facing use cases, whether it's fast charges, whether it's things like, you know, if we start making a solar PV inverter that's extremely compact.
01:05:52
Speaker
So chips for all of those use cases, you know, they really numb the volumes are tens of millions a year, if not hundreds. So there we will not be able to manufacture here, but it's also a good place. So you're not going to get to 100 million, let's say 10 million units right off the bat.
01:06:06
Speaker
right You first need to make sure that you're making the right kind of gallium nitride chip that can fit into that application to begin with. right So, like I said, there's a reasonable amount of customization that goes into that. And that's again because gallium nitrate today is more expensive than silicon as you can imagine.
01:06:19
Speaker
So, you know you need the correct price performance trade-off. right So, it's not because people are like, okay, why am I paying twice for this gallium nitrate component compared to a silicon component? and So you have to solve for that adoption problem. You have to show them that, look, if you use this in your system, the sheer savings in your bill of materials by doing this and the functionality enhancement, plus the fact that you need lesser cooling cost and all of that means that your total system cost actually comes down if you use gallium nitrate.
01:06:45
Speaker
And that's a conversation that you need to have with the you know customers across designs, across products. So there the prototyping initial volumes can can always be done here. So once you establish that, you know, ah that this is a good product for this particular use case, then I think you can you can think about where that scaling is going to happen.
01:07:02
Speaker
And there are foundries worldwide where you could, you know, potentially take the technology, have it scaled up, get back your products. And then that operates more on the standard sort of a foundry model, right? So that's where we are.
01:07:14
Speaker
So you told me that... I've been calling it Agnet and you pronounce it Agnet. Sorry, I, yeah, yeah. What's the right pronunciation? Please call it Agnet. Agnet is good.
01:07:26
Speaker
ah So sometimes I lapse into, you know, my Americanism. yeah It's not a not a good thing. Okay, okay got it. Agnet, this comes from that Agni, the Hindi word Agni. Okay, got it. Yeah, Agnet also means, you know, uncountable.
01:07:41
Speaker
It's an old Sanskrit word, uncountable, infinite. it So which is actually a good reflection of you know the kind of number of use cases that gallium nitride could potentially do. right So again, if you take two steps back and think about it, I'm telling you that wherever you have to transmit some amount of information wirelessly or wherever you you are consuming energy in the form of electricity, gallium nitride will have a role to play in those areas.
01:08:04
Speaker
We've just seen the first manifestation of that in terms of your phone charger and laptop charger. right It's literally everywhere that you use electricity, gallium nitrite could be a very competitive solution. so And same goes with the wireless transmission case.
01:08:17
Speaker
so I guess even for like say music, like like airport, et cetera, gallium nitrite would allow you to have more more output in smaller form factor. Presumably, yeah. So, and, but again, there also there's a question of safety, you know, how much power do you want to put, you know, close to your ears? And that's also the reason that GAN hasn't yet got into cell phones, but it's getting there, right? So it could be a much better integrated solution. But today, most of the cell phones will use gallium arsenide-based power amplifiers in their front end, which is what does the radio communication, right?
01:08:47
Speaker
so But it's a matter of time before that you know gets swapped out because gallium nitride will start addressing lower power requirements as well. so And that's why the technology is kind of also headed. So you can integrate a lot more and maybe even do like small bits of logic, and so which will help you know maybe even controlling when the GAN is on and off.
01:09:05
Speaker
That could be done on GAN. right So all of that will will pack more functionality. right so And we've seen that this is very possible, again, given all the extensive experience with silicon and with previous generation technologies like gallium arsenide.
01:09:17
Speaker
So, will it take over completely? Absolutely not. right So, there is space to play for everybody. But it is going to do an increasing number of things and it is also going to give you ah kind of differentiated functionality.
01:09:27
Speaker
So, you have a compact fast charger. So, that is at the end of you can make a fast charger you know even even using silicon, it may not be as compact. so But that compactness matters for certain use cases. That efficiency matters for certain in use cases.
01:09:40
Speaker
Getting 3-4% higher efficiency, which is what your gallium nitride charger does today when compared to silicon, on a population scale actually saves you ah you know a lot of energy. So those are things that gallium nitride is becoming more and more meaningful for.
01:09:53
Speaker
So the number of use cases for gallium nitride is going to increase and we hope that we are able to you know cater to a wide variety of that and also push for adoption in certain use cases.
India's Semiconductor Strategy
01:10:03
Speaker
So, you know, I had asked you this question that if India wants a ah wants to build a foundry, ah does it need money or does it need IP? You told me ip is a bigger challenge and you want to be one of those people who provides that IP. What did you mean by that? Like what kind of an IP would you provide to a foundry?
01:10:26
Speaker
Sure. So let's say, so for example, we have the India semiconductor mission, right? So hopefully a version two also will will come out very shortly, right? So I think you have a lot of incentives in order to set up all the manufacturing infrastructure, you know, set up a foundry, so to speak, right?
01:10:41
Speaker
So if you think about it, what will make such an effort successful in the first place? Think about it like a three-legged stool, right? three legs of the stool. One is, of course, the capital. So let's say government will you know give you a 50% subsidy and say state governments will give you another additional 20% subsidy or so.
01:10:56
Speaker
You still need to find money to put up the 30% in the first place. and right so And whatever is the mode of reimbursement, etc. that they use. right So you need to find the money. That is the capital piece of the puzzle. Then you have the technology piece of the puzzle. okay You put up all this capacity. What...
01:11:11
Speaker
is it going to run? What technology is it going to run? Where is it going to come from? right And what does it do? What will it implement? right Like lithography, where will it get the lithography technology from? or No, I mean the entire the entire chain, not just specific pieces of it. So let's say you want, ah like I said, you know and let's say you want 50 nanometer gallium nitride node operating.
01:11:31
Speaker
or that Or for example, the fab that's coming up in Gujarat is going to be, you know I believe, a 180 nanometer process technology on silicon initially, which is capable of making certain kind of power power integration modules, PMICs, power management ICs and things like that.
01:11:45
Speaker
So that's what I mean by technology. So what is the node, so to speak, you know put it in very layman, what is the node on which this boundary will run? So that process technology. And number three is, okay so let's say that you have the infrastructure, it's running a certain technology,
01:12:00
Speaker
Who's going to do the offtake? Who's going to buy these chips? Who are those customers? you know what What are they looking for? you know Why would they buy The product technology. Yeah, correct. So that's the whole the whole offtake piece of it. is slightly more you know That's the business angle.
01:12:12
Speaker
okay So you have capacity. You have a certain capacity. You put up a certain size plant. It's running this technology. Products are going to come out. ah Who are they going to be relevant for? Who are the customers? Who's going to buy? So at Acnit, we are we're kind of solving for the second problem.
01:12:28
Speaker
And third piece of that, right? So we are solving for the technology, for being the technology provider to us foundry in the future. And we are solving for the offtake piece of it. So if you make these chips using this technology, who's going to buy it, right? So who are the customers who will actually buy this?
01:12:43
Speaker
So buy these chips, or who are the customer profiles who will buy these chips? Or what kind of chips can result out of this process technology, right? So that's what we are solving for. i once heard a talk by Raghuram Rajan, who kind of said that, uh,
01:13:01
Speaker
in India should not attempt to, and I'm maybe wrong in what I'm paraphrasing, but essentially his take was that ah while we so hung up on building our own semiconductor supply chains, ah we have countries who will provide it at much cheaper cost than we could ever build in-house. And we should focus on our strength, which is on services design, you know, instead of trying to get into manufacturing. ah Manufacturing is like there are,
01:13:31
Speaker
people who will do it cheaper and we would be like, we can buy a semi conductors as much as we want, because there are countries who are doing it cheaply. ah What is your take on this?
01:13:42
Speaker
You know, it's, it's very simple. I mean, we, we live in the world we live in, right? so ah that also means that there's a certain geopolitical reality to all the conversations that we have now.
01:13:53
Speaker
I think in large piece, you can see what's happening between you know the two superpowers of the world now. So that is ah there is a lot of contention, especially over technology. right And so some amount of manufacturing efforts, whether it's in semiconductors, et cetera, is also, you know again, even in the semiconductor space, you go and put up foundries, they're not going to be 100% self-reliant. Somebody could say that I won't sell you this particular chemical and then you're dead.
01:14:17
Speaker
right So that's the reality. right So you will never be 100% indigenous. So that's number one. Number two is, you know how do how is it that you'll move up the value chain even in this kind of manufacturing sort of an ecosystem? Let's say you want to do that.
01:14:28
Speaker
So that's an interesting piece of it. right So it's not just driven by things like you know pure economic reasons. It's also driven by things like geopolitical reasons. And then what is it that you can do in the future?
01:14:40
Speaker
So let's say that what I mean by that is, let's say that you've not put in the hard yards, you're not putting the blood-sweat into years to develop today's technology, you know, you're not going to be able to upgrade it to tomorrow's technology.
01:14:51
Speaker
So let's say you've got a technology handout today, you'll get a technology handout tomorrow, and then you'll expect a technology handout the day after tomorrow, right? So do you really want to be in that position as a large economy?
01:15:02
Speaker
So if you were a small you know bit player somewhere, it maybe it does not matter. right So nobody is suggesting that you know a much smaller country, go put up semiconductor boundary, do this, do that. Even though there are quite good examples of of countries who have done that. Like Korea and in Japan. Yeah, Korea, Japan, Taiwan. right So they all they've all done that. And they've all been like significantly industrialized economy. So if you're going to be a significantly industrialized economy,
01:15:24
Speaker
um So you'll never do 100% of semiconductor manufacturing in-house. But you know we need to have areas which we say, look, these are important. In this amount of capacities, these are important to us as a country.
01:15:37
Speaker
And we'd like to keep some part of that internal. So that means that you know for you said that we can go and buy anything from anywhere. right That's clearly not true, for example, in the gallium nitride RF space today.
01:15:47
Speaker
Like I said, these are export control technologies. So it's a restricted technology. So you're not able to get it for certain, you declare defense as your end use, you probably won't get it. Right. So do we want to be in that position?
01:15:58
Speaker
is it's At the same time, it's not true for fast charges. Let's say you want to do fast charges. Yeah, people probably sell you as many chips as you want. Right. So there the the customer pain point is different and it needs to be solved for in a different way.
01:16:10
Speaker
Right. So there is some truth to both sides. So we've historically been very strong in the design and the services driven piece of it. But you know i'm I'm also very curious. So this is also not something that I know very well. right So i'm also kind of I may also be very wrong. But in a world where you know ah many of the services that we that we are currently doing, whether it's in terms of software services, et cetera, there's always this uncertainty in terms of what's going to happen to them now that AI is coming up.
01:16:34
Speaker
So, you know, does this not make a case for having some amount of manufacturing so that some amount of job creation will be there? Because currently we have this demographic bulge, right? So, and and you didn't need to find productive employment. Not everybody is going to be able to retrain in a way that, you know, your jobs will be AI approved, right? So not everybody is going to pick up prompt engineering and then and then get started, right?
01:16:54
Speaker
So you do need a certain amount of jobs in the manufacturing sector to exist, right? And they also provide very gainful employment to a lot of people. And more importantly for efforts like Acton, for example, it's not just about you know the employment piece of it. right So that's that's kind of secondary from where we are sitting. It's the IT piece of it that is primary.
01:17:13
Speaker
right So you can say that, look, we have this technology. Now what we choose to do with it, whether we want to make chips for our internal consumption, maybe we'll make you know ah products that also start getting exported globally and make global standard products. And by the way, there's no relaxation in that. So just because we are supplying to the Indian market, the benchmark is not that, oh, you know you're an Indian supplier, we'll cut you some slack. You'll be 20% worse than somebody else. right So people are going to tell you, you need to be as good as the other guy, the the global products I'm able to access, and you need to be 20% cheaper. right So that's typically the way the conversation goes.
Legacy Tech Nodes and Future Prospects
01:17:45
Speaker
You said currently we have a 150 nanometer node a foundry coming up in India. 180, I believe. So Tata's fab is just coming up. in I think it's also, it's going to do a bunch of nodes. It's going to do 180, 90, 45, believe. So, um but yeah, that's there's more information in the public domain.
01:18:03
Speaker
it So it's eventually going to build up to a 45 nanometer. I think so. Yeah, I think so. Is that ah like if I was to take telecom, like the world is on 5G now, so that 45 nanometer is what? Is it like a 4G or a 3G equivalent? Yeah.
01:18:21
Speaker
Actually, it's it has a lot of, not everybody's, again, like I said, it's mainly your iPhone processor and your your kind of, you know, like ah laptop processor, which is all pushing into the cutting edge, right? So older generation technology nodes have a lot of value.
01:18:35
Speaker
Even doing 180 nanometer, for example, a lot of power management ICs, etc. get made on 180 nanometer. A lot of automotive chips, you know, we've heard this time and again during the co-built chip shortages. A lot of automotive chips are not at the bleeding edge.
01:18:46
Speaker
Okay, because you need something that's, you know, in a given form factor, which is reliable, which is tested and the economics need to work out. So there's a lot of value in doing like, you know, like older generation nodes.
01:18:57
Speaker
And I think that pragmatism is what we are seeing now. So, which is very important. But do you think India could... go ah like sub 10 meters, ah sub 10 nanometers? I don't know. I think there are more informed minds out there which will have an opinion. I think our biggest problem is going to be something like manpower shortage, right?
01:19:14
Speaker
It's again not about you could potentially get technology and you could, you know, find the capital to put up all of this. And of course, no doubt if you are able to make it, there'll be a market for it, right? But let's say you have 10,000 people who are working in a fab.
01:19:26
Speaker
So you're not going to find 10,000 people who've done this before. So you're not even going to find 10,000 people who have done, let's say, 45 nanometer before in terms of manufacturing, in terms of running a plant, in terms of.
01:19:38
Speaker
So in many ways, it's ah it's ah it's an effort that's breaking ground there. right So even in terms of the whole manpower development. This is a decades-long project, which we must go. It's a multi-decadeable project. We must embark on it. And again, we mustn't try to do everything.
01:19:53
Speaker
We can't do everything under the sun. We we can pick things in which you know we as an ecosystem have a certain amount of technology capability or some prior experience in adjacent areas. So we have these you know really... like giant, globally giant chemical companies, right?
01:20:06
Speaker
So we could you know get into the semiconductor chemical supply chain, right? So these are all you know kind of viable possibilities as well. And in and in cases in certain cases, we may even have the technology, the semiconductor technology in order to do this, right? So there's work being done excellent work being done across academic institutions in the country.
Conclusion and Acknowledgments
01:20:24
Speaker
SEL is you know now being repurposed into you know a center that will now support a lot of design efforts, especially on silicon, etc. right And they have a manufacturing base which they've kept running for 30 odd years.
01:20:35
Speaker
So there are always these pockets of technology excellence that you can find and maybe we build around those. yeah Thank you so much for your time, Harish. It's been a fascinating chat.
01:20:47
Speaker
Great. Thank you. Thank you so much