[TusharK]

The Tata group is in talks with three state governments to invest up to US$ 300 million to set up a semiconductor assembly and test unit.



According to a media report, Tata is having discussions with the governments of Tamil Nadu, Karnataka and Telangana. The company is also said to be scouting for land to set up the new factory.

Tata plans to set up an outsourced semiconductor assembly and test (OSAT) plant. The facility will assemble and test foundry-made silicon wafers and turn them into finished semiconductor chips.

Tata's new semiconductor unit could employ up to 4,000 workers and is expected to come online by late 2022.

Source: ET Auto

Link to Team-BHP News

[nitkel]

Seems like a good idea, but will it be too late to be online by end 2022?

[Electromotive]

This is just assembly operation. They are not going to fab the wafers, just create finished chips from wafers sent by the foundries. Setting up a semi fab requires billions of dollars in investment.

[livetodrive]

Wafer to final chip is a long way. Starting from photolithography, etching , packaging and huge amount of ATE (Automatic Test Equipment) testing is required to role out a chip to the OEM. Even if TATA plans to just package and test a chip, that itself will cost millions and truly is a complex engineering to setup first time in India. This should be the ideal start point before we think to create wafers. Kudos to TATA.

[amay1283]

Tata leads again! Good starting point. Hopefully, in the long run they will be able to expand or create partnerships to have the entire manufacturing cycle running out of India. Not to mention the additional employment opportunities this is going to open up.

[Electromotive]

Let me drop some interesting tidbits:

- Most semiconductors (except high frequency applications - radar, satellite) are made from mono-crystalline silicon. A very large crystal of silicon is grown from a seed (small silicon crystal, generally from a previous batch).

- The large crystals are sawed into thin circular disks, these are called wafers.

- These wafers are doped and etched via photo-lithography in multiple sequential operations creating layers of features that form the transistor gates and other circuits.

- The circuits are grouped into dies. Each die is the actual working component of a chip.

- A single wafer can contain hundreds of dies. At this point the etched silicon disk is still called a wafer.

- The wafer is then diced to cut out the dies.

- The dies are then tested. Functional dies get wires bonded to them and then packaged into plastic or ceramic substrate to form a ready-to-use chip.
A die, when packaged, results in what we call a chip. The wires connect the die to the external pins of a chip. Refer image below.





TATA is getting the finished wafers and will be packaging them. This is all done by automated machinery. This is still controlled technology though not as critical as the fab tech. Safe to think the GOI is pulling some strings to make this happen.

Now, one might ask, isn't it logistically inefficient to create the finished wafer in one country and have them shipped to another just to package them?

Yes, it is inefficient but chip manufacturers do this to save on customs.
Look at the Ryzen chip below, it says Diffused in USA, Diffused in Taiwan but Made in China.



So, what happens is - Global Foundries in USA made the I/O controller die inside the Ryzen CPU and TSMC in Taiwan made the processing cores - chiplets in AMD speak. Hence the "diffused in USA and Taiwan".
Now the finished wafers containing the dies were shipped to China where the dies were cut out, tested and bonded to a PCB and a heat spreader put on resulting in the finished CPU shown in the image, hence Made in China.

[Vid6639]

Originally Posted by livetodrive

Wafer to final chip is a long way. Starting from photolithography, etching , packaging and huge amount of ATE (Automatic Test Equipment) testing is required to role out a chip to the OEM. Even if TATA plans to just package and test a chip, that itself will cost millions and truly is a complex engineering to setup first time in India. This should be the ideal start point before we think to create wafers. Kudos to TATA.

The CEO is a veteran in the industry. He was the first to setup an ATE test facility in India way back in 2004 from scratch.

He has been trying to setup a mass volume test and packaging facility in Sriperumbudur but wasn't able to find much traction.

Now with the backing of Tata they are planning to have a facility in Hosur.

The first phase will be to have an OSAT (Outsourced Assembly and Test) facility for mass volume testing and final supply to OEM's.

After that they will move backwards to fabricaton which is a long while away.

[neeraj0272]

These type of news only confirms the fact that Indian business houses are mere 'facilitators', 'advisors' 'consultants' etc who know how to work our bureaucratic and political systems in their favor and without any actual in-depth technical or even business knowledge required of an industry.

[livetodrive]

Originally Posted by Electromotive

Let me drop some interesting tidbits:

So, what happens is - Global Foundries in USA made the I/O controller die inside the Ryzen CPU and TSMC in Taiwan made the processing cores - chiplets in AMD speak. Hence the "diffused in USA and Taiwan".

Very true Electromotive, every small step here is so complex that its worth and definitely needs to be handled by a specialized team. Since I am into Chip testing (not ATE but functional validation), I can see the design, verification, synthesis, PD, test development and delivery to Taiwan for production testing, all are done across the globe. With this architecture, as you mentioned there are complexities like customs and also different time zones involved. But somehow, we feel we are tightly combined to achieve the same goal.

Sorry for being off topic. What I am trying to say is, yes there are several steps and barriers involved in the process. But even if TATA picks up a small piece of the cake, it will be a great step ahead to put ourselves in the map of semiconductor manufacturing across the globe.

[Electromotive]

Originally Posted by livetodrive

What I am trying to say is, yes there are several steps and barriers involved in the process. But even if TATA picks up a small piece of the cake, it will be a great step ahead to put ourselves in the map of semiconductor manufacturing across the globe.

Yes, it is a step in the right direction.
Honestly, I don't have a lot of respect for most of the Indian business houses because all of them have a trader mindset - buy low, sell high, pocket the profits. I don't see them spending on R&D or innovating or even encouraging innovation. Which is why instead of product companies, we have service based companies. And whatever product companies we do have, all use imported tech to manufacture their product.

These, of course, are only the symptoms of the disease, the root is our stagnant, retarded culture that refuses to evolve. Less said the better, lest I get lynched.

Also, when it comes to critical technologies, it is the govt. pulling the strings behind the scene to make things happen. The current central govt. has correctly identified (a decade late) the need for self sufficiency in semi fabrication, so thing are now being set in motion. Noticed the newly developed friendship with Israel? Russia is being sanctioned so a new friend has been found for the weapons tech and it is uncle Sam approved.

[neeraj0272]

Originally Posted by Electromotive

Also, when it comes to critical technologies, it is the govt. pulling the strings behind the scene to make things happen. The current central govt. has correctly identified (a decade late) the need for self sufficiency in semi fabrication, so thing are now being set in motion. Noticed the newly developed friendship with Israel? Russia is being sanctioned so a new friend has been found for the weapons tech and it is uncle Sam approved.

There is a major reason that Israel is being proactively befriended by this govt and it is not what you mention above.

The problem with govt influencing business/industrial decisions is that it almost always result in failures. Corruption is always the end-product of these decisions.

And you are right that lack of innovative mindset is so deep-rooted in us that it can be called 'cultural'.

[Tucker48]

I would say it's too little and too late. By 2023, industry expects no chip shortages and we might even see overcapacity. It's just an assembly plant completely dependent on foundries who inturn are dependent on ASML(next trillion dollar company in my opinion) for advanced machines. Also geopolitics play a big role here. A while back US stopped the sale of advanced machines from ASML to China even though it's a Dutch company.

What I think tata is targeting is old age chips only which are used in vehicles as space is not an issue in vehicles like in computers, smartphones, etc. But in future, it might change as well with more tech flowing into vehicles.

[Electromotive]

Wow, lot of people seem interested in this stuff. Thank you for your interest.

Here is another tidbit:
India has a small scale govt. owned fab, now under the control of ISRO.
It is the Semi-Conductor Laboratory (SCL) in Chandigarh. Website

SCL is used for experimentation and small scale production. The most advanced process node they have is 22nm. They mostly make CMOS sensors, application specific chips and Micro-Electro-Mechanical Systems (MEMS). Not exactly a cutting edge setup, but it is an integrated setup with design and manufacturing capability.

The govt. has also established STARC (Semiconductor Technology & Applied Research Centre) at Bangalore and GAETEC (Gallium Arsenide Enabling Technology Centre) at Hyderabad, mostly for defense and space applications.

As mentioned in a previous post Gallium Arsenide (GaAs), Gallium Nitride (GaN), Indium phosphide (InP) are used as semiconductors for high frequency applications like radar, satellite and anything involving microwave radio communications.

[AZT]

Just to put things in perspective

Texas wins contest to host Samsung's new $17 bln chip plant

Read more: https://www.reuters.com/technology/s...sj-2021-11-23/

[OffRoadFun]

Good start with testing and may be packaging tech, lower end of IC value chain.
This way may be we can move up the value chain a bit.

Any hopes of an actual fab for wafers are just a dream, India missed the bus long back.

With pathetic infrastructure, water shortage, frequent power cuts. lack of availability of skilled labor in the area, fickle govt policies, land acquisition issues and labor unrest (remember what happened in Apple assembling factory in Neelamangala), India has ZERO chance at a fab. Who will put $10bn with so many factors to start a state of the art fab?
Even places like Vietnam are more attractive from that perspective.

Nevertheless, trying to capture lower end of value chain is good idea and good start.