Over 20 semiconducter companies submit EOIs for PLI scheme in India

[CaptainSoap42]

Over 20 semiconductor cos submit EOIs for PLI Scheme

The government will be open to negotiating new incentives

www.thehindubusinessline.com

@Chhatrapati @koolzberg @rEd cHiLLi @HostileInsurgent @MirageBlue

 

[CaptainSoap42]

Government should start work on infrastructure needed for it.

 

[Bilal9]

Well Good luck on this.

“The government is looking at at least 3-4 companies leveraging this incentive or showing expression of interest. They do not want to fix what kind of incentives they want to provide. Once they have EOIs, they will customise incentives based on the company’s area of work, type of fab, and requirement. It will not be a one-size-fits-all solution and that is going to be USP of fab incentives this time compared to the traditional approach of take it or leave it,” Rajeev Khushu, chairperson, IESA told BusinessLine declining to name the companies.

Why the effort to decline naming the companies...if this is a public sector taxpayer-funded thing. Doesn't India have a FOIA law that says the members of the public have to know everything going on?

And what is stopping local Indian business folks from spending a few crores to set up their own semiconductor shops for System on Chip type stuff....FABs are not a matter of joke, they need investments of three to five Billion and the advantages evaporate in two years (planned obsolescence), needing re-investment at the same level again by that time.

Knowing Indian govt. takes away manufacturing incentives after a few years (like they did for the Indian auto sector recently which caused a crash in that sector), I'd like to see how many companies remain interested. Not wishing bad luck, but there are countries which offer more competitive incentives than India does.

Please don't respond with troll comments, no need.

 

[FairAndUnbiased]

Well Good luck on this.



Why the effort to decline naming the companies...if this is a public sector taxpayer-funded thing. Doesn't India have a FOIA law that says the members of the public have to know everything going on?

And what is stopping local Indian business folks from spending a few crores to set up their own semiconductor shops for System on Chip type stuff....FABs are not a matter of joke, they need investments of three to five Billion and the advantages evaporate in two years (planned obsolescence), needing re-investment at the same level again by that time.

Knowing Indian govt. takes away manufacturing incentives after a few years (like they did for the Indian auto sector recently which caused a crash in that sector), I'd like to see how many companies remain interested. Not wishing bad luck, but there are countries which offer more competitive incentives than India does.

Please don't respond with troll comments, no need.

I don't think India will get a real semiconductor industry within the next 10-20 years but actually that's not how fabs work. They don't get obsoleted when a new node comes, they are repurposed. That's why companies like Texas Instruments, Global Foundries, etc. are still around and actually make tons of money despite having only older nodes.

Very few chips use the latest node because costs rise exponentially for each die shrink. The most cost effective node are 130 nm - 28 nm. FinFET even costs more to design, much less fabricate.

As of right now India's fabs are basically the state of the art from 1990's. They still have a ton of low hanging fruit to pick before getting to the 1st "hard" barrier at 65 nm (transitioning to immersion lithography), then a 2nd "hard" barrier at 45 nm (transitioning from poly cate to metal gate and from silicon dioxide gate dielectric to high-K), then finally a 3rd "hard" barrier at 14 nm (transitioning to FinFET).

 

[Bilal9]

I don't think India will get a real semiconductor industry within the next 10-20 years but actually that's not how fabs work. They don't get obsoleted when a new node comes, they are repurposed. That's why companies like Texas Instruments, Global Foundries, etc. are still around and actually make tons of money despite having only older nodes.

Very few chips use the latest node because costs rise exponentially for each die shrink. The most cost effective node are 130 nm - 28 nm. FinFET even costs more to design, much less fabricate.

As of right now India's fabs are basically the state of the art from 1990's. They still have a ton of low hanging fruit to pick before getting to the 1st "hard" barrier at 65 nm (transitioning to immersion lithography), then a 2nd "hard" barrier at 45 nm (transitioning from poly cate to metal gate and from silicon dioxide gate dielectric to high-K), then finally a 3rd "hard" barrier at 14 nm (transitioning to FinFET).

Well Thanks brother - good to know.

I thought they had only research FABs in India, not commercial FABs and certainly not the latest 7 or 9 nm variety. Of course you are talking to a total greenhorn so pardon my ignorance.

What good would a FAB do to the Indian market? It seems they are only looking to make SoC products for Cellphones, Tablets and Laptops (or even TVs). You have seen Bangladesh Volume too for cellphones - could they profitably make SoCs locally? Right now they import these from Helio (Octacore) and the Graphics chips may be from Mediatek. This Bangladeshi Tablet was OEM for US market and used the Helio P60.

 

[FairAndUnbiased]

Well Thanks brother - good to know.

I thought they had only research FABs in India, not commercial FABs and certainly not the latest 7 or 9 nm variety. Of course you are talking to a total greenhorn so pardon my ignorance.

What good would a FAB do to the Indian market? It seems they are only looking to make SoC products for Cellphones, Tablets and Laptops (or even TVs). You have seen Bangladesh Volume too for cellphones - could they profitably make SoCs locally? Right now they import these from Helios (Octacore) and the Graphics chips may be from Mediatek.

yep Indian fabs are still running on 200 mm wafers and 180 nm. Even though there's still a ton of 200 mm fabs, they're not being used for logic but rather for analog, power, etc (which can still have powerful microcontroller logic capabilities for things like cars, robots, etc but not for computers). using 200 mm wafers for logic is 1990's level tech.

a mobile SoC is far out of reach of India. 1st gen SoCs like Snapdragon 1 require 65 nm fab capabilities.

 

[Bilal9]

yep Indian fabs are still running on 200 mm wafers and 180 nm. Even though there's still a ton of 200 mm fabs, they're not being used for logic but rather for analog, power, etc (which can still have powerful microcontroller logic capabilities for things like cars, robots, etc but not for computers). using 200 mm wafers for logic is 1990's level tech.

a mobile SoC is far out of reach of India. 1st gen SoCs like Snapdragon 1 require 65 nm fab capabilities.

Well in that case, it is probably out of reach for Bangladesh too. Especially at this particular time with the local demand and export volumes for mobile devices like Cellphones and tablets. Probably similar for upmarket IoT devices too.

It is probably better to source these mobile SoCs from Samsung, companies sourcing from TSMC (like Mediatek) or other low cost suppliers.

A lot of the local manufacturers are importing and integrating Snapdragon 6 series into motherboards for their inexpensive phones, including those in Bangladesh. These are popular among gamers.

 

[CaptainSoap42]

Well Good luck on this.



Why the effort to decline naming the companies...if this is a public sector taxpayer-funded thing. Doesn't India have a FOIA law that says the members of the public have to know everything going on?

And what is stopping local Indian business folks from spending a few crores to set up their own semiconductor shops for System on Chip type stuff....FABs are not a matter of joke, they need investments of three to five Billion and the advantages evaporate in two years (planned obsolescence), needing re-investment at the same level again by that time.

Knowing Indian govt. takes away manufacturing incentives after a few years (like they did for the Indian auto sector recently which caused a crash in that sector), I'd like to see how many companies remain interested. Not wishing bad luck, but there are countries which offer more competitive incentives than India does.

Please don't respond with troll comments, no need.

You don’t reveal name of companies wanting to invest and the EoI form is still there on MEITY website. We already have our small scale fabs used by ISRO and IIT Bombay but those are just 100nm. Indian goverment is not planning to take away incentives, infact is going to increase it.

The countries which are giving higher incentives is for high performance chips used in Laptops or other highesnd devices, what I can expect is setting up of fabs for low end to mid end chips, used in smartphones, automobiles, aviation, etc.

200

100

 

[FairAndUnbiased]

Well in that case, it is probably out of reach for Bangladesh too. Especially at this particular time with the local demand and export volumes for mobile devices like Cellphones and tablets. Probably similar for upmarket IoT devices too.

It is probably better to source these mobile SoCs from Samsung, companies sourcing from TSMC (like Mediatek) or other low cost suppliers.

A lot of the local manufacturers are importing and integrating Snapdragon 6 series into motherboards for their inexpensive phones, including those in Bangladesh. These are popular among gamers.

The biggest issue is that today it's actually harder to get into top tier semiconductors than before due to the cost and the low profit margins. There used to be 10+ companies competing at the highest level, now it's down to TSMC and Samsung at the highest levels with SMIC, Intel, UMC and Global Foundries following.

Fabs weaker than tier 1 can't compete directly in leading edge consumer digital logic which has the highest volume and profit. They settle for tier 2 products (SMIC, UMC, GloFo), use architecture innovation to boost performance instead of raw process power (Intel), or compete in industrial fields (all of them - even Intel sells tons of industrial products).

The other thing is diminishing returns. Beyond 14 nm, things get incredibly expensive to even design and for many 7 nm chips, most of the transistors aren't even used because of heat constraints. Intel and GloFo realized this and they decided to stabilize at 10-14 nm.

You don’t reveal name of companies wanting to invest and the EoI form is still there on MEITY website. We already have our small scale fabs used by ISRO and IIT Bombay but those are just 100nm. Indian goverment is not planning to take away incentives, infact is going to increase it.

The countries which are giving higher incentives is for high performance chips used in Laptops or other highesnd devices, what I can expect is setting up of fabs for low end to mid end chips, used in smartphones, automobiles, aviation, etc.

100

200 mm refers to wafer size. Current state of the art is 300 mm. 300 mm wafers have much higher yield than 200 mm wafers due to their larger surface area and less dead space at the wafer edge.

180 nm refers to process size. you can have small processes on 200 mm wafers i.e. automotive 65 nm chips on 200 mm or large processes on 300 mm wafers i.e. 180 nm mixed signal microcontroller on 300 mm.

The problem with 300 mm is that it is way more expensive to build equipment to handle wafers that big and you invest more into a single wafer (making them a poor choice for low volume or highly customized chipsets).

 

[CaptainSoap42]

The biggest issue is that today it's actually harder to get into top tier semiconductors than before due to the cost and the low profit margins. There used to be 10+ companies competing at the highest level, now it's down to TSMC and Samsung at the highest levels with SMIC, Intel, UMC and Global Foundries following.

Fabs weaker than tier 1 can't compete directly in leading edge consumer digital logic which has the highest volume and profit. They settle for tier 2 products (SMIC, UMC, GloFo), use architecture innovation to boost performance instead of raw process power (Intel), or compete in industrial fields (all of them - even Intel sells tons of industrial products).

The other thing is diminishing returns. Beyond 14 nm, things get incredibly expensive to even design and for many 7 nm chips, most of the transistors aren't even used because of heat constraints. Intel and GloFo realized this and they decided to stabilize at 10-14 nm.


200 mm refers to wafer size. Current state of the art is 300 mm. 300 mm wafers have much higher yield than 200 mm wafers due to their larger surface area and less dead space at the wafer edge.

180 nm refers to process size. you can have small processes on 200 mm wafers i.e. automotive 65 nm chips on 200 mm or large processes on 300 mm wafers i.e. 180 nm mixed signal microcontroller on 300 mm.

The problem with 300 mm is that it is way more expensive to build equipment to handle wafers that big and you invest more into a single wafer (making them a poor choice for low volume or highly customized chipsets).

Not all who submitted EoI will set up fab, most others will manufacture equipment used in semiconductor industry or related stuff.

 

[nahtanbob]

Well Good luck on this.



Why the effort to decline naming the companies...if this is a public sector taxpayer-funded thing. Doesn't India have a FOIA law that says the members of the public have to know everything going on?

And what is stopping local Indian business folks from spending a few crores to set up their own semiconductor shops for System on Chip type stuff....FABs are not a matter of joke, they need investments of three to five Billion and the advantages evaporate in two years (planned obsolescence), needing re-investment at the same level again by that time.

Knowing Indian govt. takes away manufacturing incentives after a few years (like they did for the Indian auto sector recently which caused a crash in that sector), I'd like to see how many companies remain interested. Not wishing bad luck, but there are countries which offer more competitive incentives than India does.

The money is the relatively easy part . I use the word relative because semiconductor is capital intensive business. The infrastructure (water, power) is the hard part.

India has already won the round 1. All the chip companies and EDA vendors have significant engineering operations in India. This really trying to graduate from a lower level to a higher level.

 

[CaptainSoap42]

The money is the relatively easy part . I use the word relative because semiconductor is capital intensive business. The infrastructure (water, power) is the hard part.

India has already won the round 1. All the chip companies and EDA vendors have significant engineering operations in India. This really trying to graduate from a lower level to a higher level.

Yes, we currently make 100nm and 180nm shakti chips here in small fabs in ISRO and IIT Bombay. It will be good if we make chips if not the higher end ones, moderate ones will be fine.

 

[CaptainSoap42]

US semiconductor maker Applied Materials in talks with Indian govt

Company could help with government's desire to have a semiconductor industry.

www.business-standard.com

 

[FairAndUnbiased]

The money is the relatively easy part . I use the word relative because semiconductor is capital intensive business. The infrastructure (water, power) is the hard part.

India has already won the round 1. All the chip companies and EDA vendors have significant engineering operations in India. This really trying to graduate from a lower level to a higher level.

the water and chemical delivery infrastructure is incredible. for semiconductor equipment, it is normal for it to be cleaned to ppb (parts per billion) spec. The water is so pure that even electropolished stainless steel or quartz glass (considered ultra clean for hospitals, food and pharmaceutical) is considered filthy because the water has to be 16+ MOhm purity - the tiny amounts of nickel that leach from stainless or trace sodium from quartz into ultrapure water is unacceptable.

 

[Paul2]

I don't think India will get a real semiconductor industry within the next 10-20 years but actually that's not how fabs work. They don't get obsoleted when a new node comes, they are repurposed. That's why companies like Texas Instruments, Global Foundries, etc. are still around and actually make tons of money despite having only older nodes.

Very few chips use the latest node because costs rise exponentially for each die shrink. The most cost effective node are 130 nm - 28 nm. FinFET even costs more to design, much less fabricate.

As of right now India's fabs are basically the state of the art from 1990's. They still have a ton of low hanging fruit to pick before getting to the 1st "hard" barrier at 65 nm (transitioning to immersion lithography), then a 2nd "hard" barrier at 45 nm (transitioning from poly cate to metal gate and from silicon dioxide gate dielectric to high-K), then finally a 3rd "hard" barrier at 14 nm (transitioning to FinFET).

What node is most economical depends on the production run size.

For the prime majority 1 man fablesses companies who have no volume at all, the only thing they can afford is 200mm, hence 180nm.

200mm is fantastically profitable given how how old the node is.

I don't think India will get a real semiconductor industry within the next 10-20 years but actually that's not how fabs work. They don't get obsoleted when a new node comes, they are repurposed. That's why companies like Texas Instruments, Global Foundries, etc. are still around and actually make tons of money despite having only older nodes.

Very few chips use the latest node because costs rise exponentially for each die shrink. The most cost effective node are 130 nm - 28 nm. FinFET even costs more to design, much less fabricate.

As of right now India's fabs are basically the state of the art from 1990's. They still have a ton of low hanging fruit to pick before getting to the 1st "hard" barrier at 65 nm (transitioning to immersion lithography), then a 2nd "hard" barrier at 45 nm (transitioning from poly cate to metal gate and from silicon dioxide gate dielectric to high-K), then finally a 3rd "hard" barrier at 14 nm (transitioning to FinFET).

A lot of fabs sell dry, planar litho on 55nm. Sometimes with SOI.

What the biggest difference after immersion is that Japanese tool makers completely dropped out of the race, hence tool cost starting to go stratospheric.

ASML itself may not be willing to babysit completely new entrants for little money.

Maximum what Indians can buy on short notice would be 2nd hand Japanese steppers, which, surprise, been nearly all bought by 3rd tier fabs 4-5 years ago in anticipation of the supply crunch on mature nodes.