The market for high end is much smaller than the low end. So India's fab plants are initially aiming for the low end and then scaling up to compete with the best, at least that's the plan.
Indian Electronics Manufacturing Developments : News, Updates and Discussions
- Thread starter RISING SUN
- Start date
The market for high end is much smaller than the low end. So India's fab plants are initially aiming for the low end and then scaling up to compete with the best, at least that's the plan.
TSMC, samsung, micron, intel..etc all of them are done with their investments for next gen in node technology. Unless you give them $10 billion on a platter and molly coddle them nothing will happen. Even then it will likely become like HAL where we will only assemble/manufacture but cannot absorb the technology if at all they transfer it
nearly, unless we develop some path breaking technology that will put us on the top there is zero chance.
As usual we are the only ppl who are talking about ourselves nobody else is. Here samsung has decided to invest in US.
Samsung expands Texas chip plant in investment race with TSMC
South Korean company angles for top spot amid US-China tech war
asia.nikkei.com
Only place where we might have chance is designing, every major company from amazon,google,apple ...etc have started designing their own chips. Probably we can expect some companies to start R&D center or bring some of that work to India (some of the companies are already doing it like amd and intel). But manufacturing is still far away, in another few years tsmc will be moving to 3nm node technology while we dont even have capability for 22nm or 30nm tech which is 10 years old.
India was more industrialized than south korea in 1980's but they have absorbed technology much better than India and taken that forward.
India should stop daydreaming and do the things right way. There is no substitute to innovation , persistence and hard work. . Ever one says India is a big market to make money not a country to invest in critical R&D.
We are not going to manufacture their IP. We are only going to set up a fab plant mainly to manufacture our IP, and the IP of many other companies around the world who think the more established players are more expensive than the Indian plant or those who are looking at selling to the Indian market and other third world countries. There are plenty of designs around that are not high end and there's plenty of space around for indigenisation as well.
For example, 90% of the Indian smartphone market does not need expensive high end Qualcomm processors that you find in flagship phones. And that's only a very small segment of the industry. More than 90% of the processor market is low end that uses 32 nm and 65 nm processes.
The ones you are referring to is the current market for high end. No one in India is going to touch that market.
Do you know the margin in old technologies ?? You literally can't buy peanuts with that kind of money. Most chips are sold at less than $1 and that includes R&D, manufacturing and marketing costs. Chip manufacturing is dead and here in India we are so obsessed about chip fabrication !!
How can chip manufacturing be dead when it is the building block of every electronic system? Chip manufacturing is evergreen like coding the problem is complexity has increased but low end cpu for IoT and other general use machines is still open for competition and that market will only expand. We will never win or lead in the race because either China will replace the Americans or the taiwanes,Soko and U.S will continue to dominate. But the lack of interest to introduce electronic manufacturing is a blunder that we are paying for badly. Though the phone market is still open for competition provided we can offer quality infrastructure and ease of business to the companies and maintain law and order. We will have bite the bullet and take risks if we plan to succeed.
Well, it’s dead because investment is high and margins are waferthin. That’s the reason you see so many consolidations happening in semiconductor industry. As far chip fabrication is concerned, the latest company to stop investment is Chartered Semiconductor. Apple is world’s second largest chip maker, but doesn’t own any fabrication plants and same is the case for Qualcomm, Nvidia and Mediatek.
The margin can be made up with volumes. A billion chips at $1 is $1 billion in sales. What if the market is for 10 billion chips?
India has scale. And because of that scale, we can't afford imports. The market for very high end desktop processors is very small in India. But imagine 10 years down the line if there's a market for at least 10 million a year for a $2000 processor and 100 million a year for a $500 processor. The forex outflow will be $70B a year. We can't afford such significant outflows. So we have to start somewhere first.
We can't afford to import the electronics that are needed in electrical appliances like washing machines, air conditioners, heaters and microwave ovens either, even toys. And the microprocessors and microcontrollers in these devices do not cost $1. Wish they did, but they don't. And you most definitely don't use a Core i7 here. Let's not forget about all the other processors present in your desktop, like in your keyboard, mouse, headphones, printer, webcam etc, none of these use i7s either.
There's a massive market for electronics in India, and India on its own can afford a fab plant for all such needs, not to mention countries in Africa and other poor parts. And I'm sure companies may find some advantages in making the Arm family in India rather than give the market away to an Indian designer because GoI has imposed a 20% tax on their chip-imports from China giving the Indian designer a 15% price advantage. No one here is considering Intel will contract Xeons from this plant. Indian companies will definitely license foreign processors or make their own designs if a fab plant shows up here.
Once the fab plant is set up, where it goes from there is up for debate, but the minimum capability needed has a profitable market already. And when Indian designers come up with their own designs, there's a fab plant right next door.
Thats not completely. AMD uses 7nm FF process of TSM to only manufacture a part of the processor (Zen 2 architecture) which is a Multi Chip Module. CPU cores are fabricated in 7nm TMSC FF. The rest of the chip (Memory controller, cache SRAM, PCI-E controller etc) are still based on older 14 nm GlobalFoundries process (which is their own spinoff). 7 nm has proved to be pretty expensive so they have not yet moved to it.
Its possible to have a basic semicondutor industry even with processes like 180 nm or 90 nm. Lots of microcontrollers are actually based on it.
Besides, there is more to chip than just fabrication. Packaging chips is a massive sub-industry and thats how worthless countries like Malaysia got into chip manufacturing business.
India should start at one of these places.
To be honest there are few things we should consider in terms of semicondutors.
Why do we want to have our own semiconductor industry at scale?
1. To reduce our import bill? There are other ways.
- May be we can improve our export account?
2. To gain strategic freedom from getting bugged semiconductors / denial regimes? Not really needed, atleast at the taiwanese scale.
- For vast majority of semi-conductors used in sensitive area like defence etc can be fabricated on 1998's fabrication technology (180 nm): Things like controllers, memory etc.
- There are niche areas like Radar which actually requires specialized semiconductor processes based on GaAs and GaN instead of silicon as semiconductor. These are low volume high cost products which can be manufactured without market price constrains, so again, you don't need multi billion dollar investment there.
- Even in areas which use modern high performance compute devices for AI applications, the inference part of AI is actually tame in its requirement for computation and is often optimized for lowering compute requirement still. Remember, improvements in semiconductor often saturate by twice or thrice in each generation (Even at Moore's law 18 months for 1.5x), while improvement in software like algorithms can have 10x, 100x, 1000x from one implementation to another (Compare RCNN(x), Fast RCNN (10x), Faster RCNN(100x) and YOLO (1000x) source : R-CNN, Fast R-CNN, Faster R-CNN, YOLO — Object Detection Algorithms).
- Sensors? Again it is not really required to have 8 nm process. 180-90 nm process is enough.
- Supercomputers? Yesterday's super computers are todays workstation with GPUs. Commodity hardware which is seldom denied.
3. Employment? Not really. Semiconductor manufacturing is not a source of job. Not anymore. May be PCB fabrication might be a better job provider industry.
- Chip manufacturing has become almost a very sublime art of automation. Jobs are far and few.
4. Enabling innovation else where?
- Innovation and value addition these days is a lot more concentrated in software. To give you an example, the interconnect in modern CPUs and motherboards are much more defined by software than by hardware. In storage, SSD have a lot of value derived from its controller algo than just the semiconductor.
- Renting exotic hardware as commodity is becoming more common. You can rent cloud servers with FGPA and exotic high performance GPUs.
- Again for a number of applications, having the best and the biggest fabrication tech is not needed.
The upshot is :
Having the greatest and frontline fabrication technology is not really needed. Its a low margin business. The areas where we can focus are :
1. Solar cells : This is a totally different niche and it does not require SAME type of investment as a top of the line chip process. We should invest a lot in it because it is actually a purely capital industry. Its products produce electricity. And for India it is a boon as sunshine is all we have got. All sun no oil.
2. LEDs : Again, not the top of the line fabrication process needed neither it is as expensive. It is a cost saver industry.
3. Chip packaging : It is needed for all chips and it is much less money intensive. After all if village idiots like Malaysia can do it, how hard it can be?
4. PCB fabrication: While world is going towards SoC, PCB will still be needed.
5. UltraCaps and Energy storage : This is another specialized type of electronic manufacturing with applications in electric vehicles, weapons, drones, lasers.
Why do we want to have our own semiconductor industry at scale?
1. To reduce our import bill? There are other ways.
- May be we can improve our export account?
2. To gain strategic freedom from getting bugged semiconductors / denial regimes? Not really needed, atleast at the taiwanese scale.
- For vast majority of semi-conductors used in sensitive area like defence etc can be fabricated on 1998's fabrication technology (180 nm): Things like controllers, memory etc.
- There are niche areas like Radar which actually requires specialized semiconductor processes based on GaAs and GaN instead of silicon as semiconductor. These are low volume high cost products which can be manufactured without market price constrains, so again, you don't need multi billion dollar investment there.
- Even in areas which use modern high performance compute devices for AI applications, the inference part of AI is actually tame in its requirement for computation and is often optimized for lowering compute requirement still. Remember, improvements in semiconductor often saturate by twice or thrice in each generation (Even at Moore's law 18 months for 1.5x), while improvement in software like algorithms can have 10x, 100x, 1000x from one implementation to another (Compare RCNN(x), Fast RCNN (10x), Faster RCNN(100x) and YOLO (1000x) source : R-CNN, Fast R-CNN, Faster R-CNN, YOLO — Object Detection Algorithms).
- Sensors? Again it is not really required to have 8 nm process. 180-90 nm process is enough.
- Supercomputers? Yesterday's super computers are todays workstation with GPUs. Commodity hardware which is seldom denied.
3. Employment? Not really. Semiconductor manufacturing is not a source of job. Not anymore. May be PCB fabrication might be a better job provider industry.
- Chip manufacturing has become almost a very sublime art of automation. Jobs are far and few.
4. Enabling innovation else where?
- Innovation and value addition these days is a lot more concentrated in software. To give you an example, the interconnect in modern CPUs and motherboards are much more defined by software than by hardware. In storage, SSD have a lot of value derived from its controller algo than just the semiconductor.
- Renting exotic hardware as commodity is becoming more common. You can rent cloud servers with FGPA and exotic high performance GPUs.
- Again for a number of applications, having the best and the biggest fabrication tech is not needed.
The upshot is :
Having the greatest and frontline fabrication technology is not really needed. Its a low margin business. The areas where we can focus are :
1. Solar cells : This is a totally different niche and it does not require SAME type of investment as a top of the line chip process. We should invest a lot in it because it is actually a purely capital industry. Its products produce electricity. And for India it is a boon as sunshine is all we have got. All sun no oil.
2. LEDs : Again, not the top of the line fabrication process needed neither it is as expensive. It is a cost saver industry.
3. Chip packaging : It is needed for all chips and it is much less money intensive. After all if village idiots like Malaysia can do it, how hard it can be?
4. PCB fabrication: While world is going towards SoC, PCB will still be needed.
5. UltraCaps and Energy storage : This is another specialized type of electronic manufacturing with applications in electric vehicles, weapons, drones, lasers.
$2000 processor? Man what are you smoking ??
Is that all you got out of it?
I wrote the price off of my memory. And it was for i9.
Intel Core i9: Everything we know about Intel’s hypercharged PC processor
Intel has filled in the rest of the details on its Intel Core i9 microprocessor family for PCs, and started shipping all of them, too. Now with a review of the 18-core Core i9-7980X and 16-core Core i8-7960XE!
www.pcworld.com
Off the mark by $1. The same price as those chips you were talking about.
It's easier for new industries since rules and regulations are yet to be formulated. The govt can make it convenient for the next 10-20 years, including supporting the industry from their end. Messing with other established businesses and old laws are called reforms and that comes with its own unique set of challenges. It's always good to focus on something that has a clean slate if we are to make it a pillar for either import substitution or export.
Most strategic systems in defence will not benefit from the fab plant. Most of it is made by DRDO itself, with little to no non-defence civilian involvement. As for most other systems, like aircraft, radars, the numbers required are too small, plus they rely on COTS anyway.
Where it's going to be important is our civilian communication systems. The US and Chinese have a habit of bugging electronics before export.
Our population is big enough that we need to put our hands into every product in our market. We will be the only major country with a manpower surplus for the next 30 years. And even fab plants requires presence since we will have colleges and universities exclusively catering to this discipline. If the industrial base doesn't exist, the education sector will also suffer and our manpower in this sector will be forced to work elsewhere.
The govt considers fab plants as a strategic sector, so the idea isn't merely to make a profit. All the govt is looking for is some level of feasibility to get two plants running, both competing with each other. The rest can be taken care of using tariffs and NTBs. The main thing the govt is looking for is whether the Indian market has reached the maturity level needed to run 1 or 2 fab plants. Apart from 2 fab plants for silicon, the govt is also aiming for a govt-controlled fab plant for germanium. So whether it happens now or after 5 years is subject to domestic market maturity and they don't care about how established it is elsewhere. There's the same thrust towards building gigafactories as well.
Remember how quickly Huawei was sanctioned? We can't afford such a thing happening to us. The Soviet Union also existed with a complete denial of technology in this sector. They made their own reverse-engineered copies of western processors instead.
=
Higher price processor are needed in workstations. Those are not items of scale. Those are speciality equipment needed by professionals.
There are things which are even bigger in scale. Think petroleum. We have always afforded oil. Besides, there are other types of electronics that does not require such a massive investment: packaging, solar cells.
LOL! If anything, world is moving towards cheaper processors. An iPad processor holds nothing on a previous gen threadripper or even an i5 but is more than enough for powering an iPad.
Higher price processor are needed in workstations. Those are not items of scale. Those are speciality equipment needed by professionals.
New industries? If anything, mass scale fabrication is a well established industry with consolidated players. Once China bleeds its way to currency in state of the art semiconductor business, there will be very less space left for a commercially viable third or fourth player. Problem with China is that they have a chip on their shoulder. Pun intended.
Most of these security problems are solved by end to end encryption and authentication. In design of any secure communication system, it is always assumed that communication will be captured. Even if you implement your own equipment, it will get broken into or sabotaged. Communication media is always assumed to be insecure.
Its more important if we standardize our requirements and algorithms for encryption, authentication, non-repudiation than to trust the ones coming out of USA/ANSI. China is already doing that, except they are foolish enough to define theirs from scratch. We should standardize ones for our need and review them. That requires a body of smart folks to analyze and challenge protocols and a mechanism to safely challenge and fix the standards. Its more of a policy thing and not a manufacturing thing.
I remember someone use to say "we are too poor to afford war". Well we are poorer to afford burning money in a vanity project like our own fab ecosystem. China is doing that and learning it the hard way. Good for them. Let them have this perpetual chip on the shoulder.
Last edited:
Actually if an Indian company gets sanctioned in the same way Huawei is getting sanctioned, having all the manufacturing capability is not going to do a damn thing. Huawei is not short of manufacturing muscle. What is denied to them is something totally different. It is among other things.
1. Market access.
2. OS / App marketplace access.
3. Software and standards to synthesize cores : Think Mentor graphics softwares like CADENCE.
4. IP Cores from fabless fellows like Qualcomm
None of these things require a manufacturing base. All of these are policy and IP denial things. A better idea for India will be to adopt open technologies instead of propriety ones here. Like adopting open ISA like RISC-V as standard in all of our government purchases. Or say fund research in developing open EDA tools. Or say invest in developing open microarchitecture for open ISA like RISC-V. Promote open market places free from such embargos. None of these are manufacturing thing and all of these are design/software/policy things.
BTW, if russia wants a microprocessor, I can get one for them for free, LOL! UltraSPARC microarchitecture is free and opensource. In future, there will be free and opensource implementations of likes of RISC-V, making entire point of such technology denial useless. Couple it with open EDA toolchains and standards and you are done... It will take time but thats where we are going. Just like US government cannt put an OS technology denial regime or browser technology denial regime.
That has been happening forver since the microprocessor was created. The only thing that is happening the amount of cores will keep increasing. CpU size is going to get smaller. We have just broken another threshold with the 7nm chips and it can still go smaller. It can still go smaller but then increasing the amount of cores becomes a problem. And we will have expensive chips still. Also people keep forgetting how much chip requirements will massively increase once IoT becomes mainstream...
There's a huge difference between the two. Oil grows at 3-10% pa, whereas electronics can easily grow at 10-30%. Also oil is volatile, it can rise as well as drop, and adjustments are made accordingly. But electronics will always go up. And we import oil simply because we do not have a choice. We can't enter the 2030s importing both electronics and oil. You can imagine what will be the import demand for gold as the middle class grows as well.
It doesn't matter. India is expected to have 1.5 billion people by 2035. Even if 200 million people achieve very high standards of living by 2035, the govt won't be able to handle the import bill. If the procs become cheaper, then it gets worse because more people will be able to afford them.
Look at the sheer rate of growth.
India: value of electronic product imports 2023 | Statista
In fiscal year 2023, India imported electronic products valued more than 6.2 trillion Indian rupees.
www.statista.com
And this is only going to increase. 3 times growth in a span of 8 years, when electronics was such a small component of everyday life. You can expect the next phase of growth for the next 8 years to be at least 5 times. We could very well be talking about 19-20 LCr by 2028. At the same rate of growth, we could be at 100 LCr. That's $1.3T by today's standards, could easily be around half of GoI's yearly expenditure. We can't afford such a large import bill.
In order to prevent such a large import bill, the govt will have to introduce tariffs, which in turn will affect GDP growth if people are denied cutting edge tecnologies. The alternative is to bring technologies into India, protect the market with tariffs and have production run in rupees, so people are not affected by the large amount of taxes.
Last edited:
Not in India though. I am talking about govt regulations in India. Since it's a clean slate, all sorts of new laws necessary for the next 10-20 years can be passed today. This will give us an edge over other countries that are dealing with regulations that require reforms to keep up.
GoI's main goal is to control both the hardware and software, not just the software part.
Our current supercomputer expansion involves a large indigenisation component where the desire is to indigenise every component possible.
India Building Its Own Supercomputers, Says Centre
India is building supercomputers instead of procuring them from foreign countries, Department of Science and Technology Secretary Ashutosh Sharma said on Thursday.
www.ndtv.com
A program to develop our own chips for supercomputers has also begun.
Same thing. We are most definitely too poor to afford war today. Similarly, we are too poor today to have our own fab plants functioning today. But if we start today, we may just be able to compete 5-10 years down the line, when we will be able to afford war as well.