Starting a fab company

[Aking2713]

Let me preface this post by stating I am not in this industry but am just curious about it. I will also say my knowledge of the field is limited as well so if I say something stupid just call me out and correct me.

I've always been interested in manufacturing and creating goods for consumers. Just the automation, supply chain, and all the moving parts create fun problems to solve. I dream one day of owning my own manufacturing company. With the recent news on the semiconductor shortage I started researching why there is a shortage and then looked into the manufacturing process and loved the complexity of it all.

I am curious the feasibility of one day starting a semiconductor production company. As I have read in order to compete at the bleeding edge you need to have an insane amount of capital in excess of billions of dollars. But what about one not specializing in the latest and greatest semiconductors. Surely, there has to be plants that specialize in semiconductors for things like small electronics like vacuums and toaster ovens. Are these still super expensive to start or could one feasible set one up for a more reasonable starting price like a few million.

Also, if it is still super expensive to set up a smaller, low end semiconductor fab do the bigger companies outsource certain steps like wafer production that one could possibly specialize into or is everything done in house?

I know the notion of starting one of these companies might be pretty niave but I always like to have a conversation before I dimiss ideas. Someone has done it in the past so why can't someone do it in the future.

Thanks for reading!

 

[smeyer0028]

This is not my area. I worked at LSI Logic back in the early
1980s. Wilf Corrigan and the fab VP from HP had no trouble
getting capital and setting up a leading edge at the time
high yield fab.

LSI Logic sort of failed because of lack of EDA tools so there
were no chip orders to fill the fab. I think even now main
problem in semiconductor manufacturing is long term
realiable orders to fill the fab. Get an order from Apple
and I am sure you will have no trouble setting up a fab.

 

[prime007]

I'm a software developer so my knowledge of the semiconductor industry is VERY limited. That said my question to you is...why would you bother throwing/risking millions of dollars to start a fab to produce a low-margin commoditized product (like small electronics for vacuums and toaster ovens)? What would be your competitive advantage? It certainly won't be price (a tiny start-up fab won't have the volume needed). Why would any company trust and choose your no-name fab vs a large foundry that would provide them with additional services that you can't provide? The established players have continuously optimized their manufacturing process and costs that I imagine it'll be extremely difficult (impossible?) for a start-up fab to make a profit.

 

[Aking2713]

I'm a software developer so my knowledge of the semiconductor industry is VERY limited. That said my question to you is...why would you bother throwing/risking millions of dollars to start a fab to produce a low-margin commoditized product (like small electronics for vacuums and toaster ovens)? What would be your competitive advantage? It certainly won't be price (a tiny start-up fab won't have the volume needed). Why would any company trust and choose your no-name fab vs a large foundry that would provide them with additional services that you can't provide? The established players have continuously optimized their manufacturing process and costs that I imagine it'll be extremely difficult (impossible?) for a start-up fab to make a profit.

Honestly, you are probably right. But doing things that are hard is a worthwhile endeavor or at least asking the question to see if there is any untapped market. Thank you for your honest opinion.

 

[Aking2713]

This is not my area. I worked at LSI Logic back in the early
1980s. Wilf Corrigan and the fab VP from HP had no trouble
getting capital and setting up a leading edge at the time
high yield fab.

LSI Logic sort of failed because of lack of EDA tools so there
were no chip orders to fill the fab. I think even now main
problem in semiconductor manufacturing is long term
realiable orders to fill the fab. Get an order from Apple
and I am sure you will have no trouble setting up a fab.

Thank you for the insights! So even with this semiconductor shortage you think the main problem would be demand?

 

[Jert]

A most basic fab today probably requires about $400-500 million to start with for maybe capacity of 5,000-10,000 wafers per month depending on device and technology levels. The facilities (the building, the cleanroom, various pipes for safe handling of chemicals, gases, automation, etc.) probably uses up about 30% of that budget, the remaining for buying 200-300 production machines costing perhaps $2-5M a piece. And the products can be made with this level of technology will be likely low end, which means primarily price competition with competitors who've been doing this for a lot longer and more experienced. An advanced technology node fab for similar production capacity will cost $3B - $5B or even more for most leading edge. All semiconductor fabs runs 24 x 7 non-stop including all holidays, that's a given. Each down day can cost many millions including equipment depreciations. It's a big gamble. Therefore smaller companies will almost always need some sort of backing including government's to take on such risks, which is what many governments in Asia do and many are hoping the US/EU governments will do similarly, unless you are big giants like Intel, TSMC, Samsung who have the chips to gamble (even that is very dangerous as we are witnessing the troubles that Intel is facing).

 

[Dougwithau]

Do you have an idea for a part or chip?
You talk about building a fab, just to build a fab. What would it make?
Most fab companies started 40-50 years ago, when the jump from 0->1 was not huge as it is now. If you want to start something, you need a niche with a decent margin. A tight niche or new technology has a chance. Tiny parts for white goods ain't it.
There is not really a "fab" anyhow. Memory process for DRAM is a specialized game, as are the large logic parts fabbed by TSMC and Intel. There are smaller fabs for mems, and specialists in analog. These are distinct technologies, moving a part from one to a different fab technology is insanely difficult to impossible.
If you have a niche technology that needs a chip, and you can't get it made at TSMC, or Global Foundries, then you may consider a fab. If your just kicking the tires, get a job working at a fab first. Try looking for product engineering jobs, or even fab technician. There are also many jobs dealing with the logistics, complexity and management of the fab. These jobs keep you busy and pay very well, without the billion $$ risks.

 

[Dougwithau]

Or, try this if your interested. https://www.zerotoasiccourse.com/

 

[Staf_Verhaegen]

But what about one not specializing in the latest and greatest semiconductors. Surely, there has to be plants that specialize in semiconductors for things like small electronics like vacuums and toaster ovens. Are these still super expensive to start or could one feasible set one up for a more reasonable starting price like a few million.

Problem is that in that space there is a lot of competition and the existing fabs are all written off. So you will need to compete with your fab that you still need to write off. The reduction of the shortage will need to come from expansion of existing fabs.

 

[Staf_Verhaegen]

Or, try this if your interested. https://www.zerotoasiccourse.com/

This teaches how to design chips, not the processing of the chips itself. It's like following driver lessons with the goal of starting a car manufacturing company.

 

[SunnyLining]

Small foundries are a thing, most of them specialized. Why don't you do some research on a few of them? Here's a couple of names to get started:
Tower/Jazz
Skywater

Those are both pure-play foundries. A lot of smaller, specialized foundries are attached to companies like Texas Instruments, NXP, and OnSemi to name a few. Sometimes they will have external customers, though there tend to be partnerships with lots of contracts involved.

You asked about outsourcing wafer production and that definitely does happen. There are whole supply chains needed to make a foundry run including wafer production, photoresist chemicals, electrolysis chemicals, etc. There are also post-production activities that tend to be contracted out like chip assembly and test.

Now, I haven't thought much about it, much less done any research, but if I wanted to find opportunities to build companies associated with a US or European push for semiconductor independence, I think weak spots in the supply chain would be where to look. It's when you look at the supply chains that semiconductor independence looks most problematic. For instance, as I understand it, Japanese companies dominate the global photoresist market.

As a disclaimer, I've never been particularly interested in starting a company, so I'll leave it to you to do the research, build a business case, get financing, attract customers, etc., etc., etc. Just typing it out makes it feel exhausting to me!

 

[Aking2713]

A most basic fab today probably requires about $400-500 million to start with for maybe capacity of 5,000-10,000 wafers per month depending on device and technology levels. The facilities (the building, the cleanroom, various pipes for safe handling of chemicals, gases, automation, etc.) probably uses up about 30% of that budget, the remaining for buying 200-300 production machines costing perhaps $2-5M a piece. And the products can be made with this level of technology will be likely low end, which means primarily price competition with competitors who've been doing this for a lot longer and more experienced. An advanced technology node fab for similar production capacity will cost $3B - $5B or even more for most leading edge. All semiconductor fabs runs 24 x 7 non-stop including all holidays, that's a given. Each down day can cost many millions including equipment depreciations. It's a big gamble. Therefore smaller companies will almost always need some sort of backing including government's to take on such risks, which is what many governments in Asia do and many are hoping the US/EU governments will do similarly, unless you are big giants like Intel, TSMC, Samsung who have the chips to gamble (even that is very dangerous as we are witnessing the troubles that Intel is facing).

Thank you for the information. Seems like a pretty brutal market!

 

[Aking2713]

Do you have an idea for a part or chip?
You talk about building a fab, just to build a fab. What would it make?
Most fab companies started 40-50 years ago, when the jump from 0->1 was not huge as it is now. If you want to start something, you need a niche with a decent margin. A tight niche or new technology has a chance. Tiny parts for white goods ain't it.
There is not really a "fab" anyhow. Memory process for DRAM is a specialized game, as are the large logic parts fabbed by TSMC and Intel. There are smaller fabs for mems, and specialists in analog. These are distinct technologies, moving a part from one to a different fab technology is insanely difficult to impossible.
If you have a niche technology that needs a chip, and you can't get it made at TSMC, or Global Foundries, then you may consider a fab. If your just kicking the tires, get a job working at a fab first. Try looking for product engineering jobs, or even fab technician. There are also many jobs dealing with the logistics, complexity and management of the fab. These jobs keep you busy and pay very well, without the billion $$ risks.

Ya you are probably right. I plan on trying to get a job at one of these companies to learn the ropes and maybe get some ideas of a business of my own. Thank you for the info!

 

[Aking2713]

Small foundries are a thing, most of them specialized. Why don't you do some research on a few of them? Here's a couple of names to get started:
Tower/Jazz
Skywater

Those are both pure-play foundries. A lot of smaller, specialized foundries are attached to companies like Texas Instruments, NXP, and OnSemi to name a few. Sometimes they will have external customers, though there tend to be partnerships with lots of contracts involved.

You asked about outsourcing wafer production and that definitely does happen. There are whole supply chains needed to make a foundry run including wafer production, photoresist chemicals, electrolysis chemicals, etc. There are also post-production activities that tend to be contracted out like chip assembly and test.

Now, I haven't thought much about it, much less done any research, but if I wanted to find opportunities to build companies associated with a US or European push for semiconductor independence, I think weak spots in the supply chain would be where to look. It's when you look at the supply chains that semiconductor independence looks most problematic. For instance, as I understand it, Japanese companies dominate the global photoresist market.

As a disclaimer, I've never been particularly interested in starting a company, so I'll leave it to you to do the research, build a business case, get financing, attract customers, etc., etc., etc. Just typing it out makes it feel exhausting to me!

THIS. This is sort of what I have been thinking about. Maybe not building a whole fab but a small company that specializes in something to help, in my case, the US to become more independent from the supply chain. I guess that is what I need to look into. Focus on the supply chain on some of these big companies and see where the weak links are. Thank you for the information!

 

[Paul2]

I'm a software developer so my knowledge of the semiconductor industry is VERY limited. That said my question to you is...why would you bother throwing/risking millions of dollars to start a fab to produce a low-margin commoditized product (like small electronics for vacuums and toaster ovens)? What would be your competitive advantage? It certainly won't be price (a tiny start-up fab won't have the volume needed). Why would any company trust and choose your no-name fab vs a large foundry that would provide them with additional services that you can't provide? The established players have continuously optimized their manufacturing process and costs that I imagine it'll be extremely difficult (impossible?) for a start-up fab to make a profit.

I see a huge opportunity still around for small order fabs.

Anything mass manufacturing as you justly remarked is off limits. It's crowded, it's expensive, and it's low margin.

200mm is very profitable today, but people who are reaping money today only do so because they spotted the trend 5-6 years ago.

Can you open a yet another 200mm mini-fab today? If you can get the equipment. 200mm equipment has been vacuumed clean off the second hand market by all people rushing to build their 5th, 6th, 7th tier mini fabs.

 

[Tanj]

There was a Japanese research project 10 years ago to develop a small scale fab. The idea was rapid turn around of low volumes of modern devices, probably in a research setting. I recall they used a single reticle shot on wafers maybe 50mm diameter? A lot of the work would have been simplifying the chemistry. Some processes like advanced etching may never even have been built at such small scale, since all the industrial uses would e 200mm or 300mm and the hard work is in perfecting the machines, which no-one would have done at smaller sizes. I have not seen any news of it for years, though.

I have sometimes wondered if there is a production process possible based upon the square epitaxial wafers (cleaved from a rectangular ingot with low waste) used in the solar industry. These are 120mm square, cheap and available in high volume. They are much thinner than 300mm standard, which might make them easier and cheaper to stack. However, pretty much every machine would need to be rebuilt for the size. Some processes might be relatively simple, there are already machines which load up multiple wafers into a chamber so modifying that for 30 or 40 squares instead of 6 disks might be done. Probably some of the chemistry and cleaning is scalable too. Planarizing would be a whole new development since it needs exquisite control and has decades of refinement with big wafers. The wafer stages used for stepper lithography would be very different.

It would be an enormous effort that pays back only slowly on cheaper materials.

 

[hist78]

Ya you are probably right. I plan on trying to get a job at one of these companies to learn the ropes and maybe get some ideas of a business of my own. Thank you for the info!

Or, you can try it at home ...

Garage Semiconductor Fab Gets Reactive-Ion Etching Upgrade

It’s a problem that few of us will likely ever face: once you’ve built your first homemade integrated circuit, what do you do next? If you’re [Sam Zeloof], the answer is clear: bu…

hackaday.com

 

[Staf_Verhaegen]

I see a huge opportunity still around for small order fabs.

I think this also mostly covered by the MPW (multi-project wafer) offered by a lot of the regular foundries (including TSMC).

 

[VCT]

Let me preface this post by stating I am not in this industry but am just curious about it. I will also say my knowledge of the field is limited as well so if I say something stupid just call me out and correct me.

I've always been interested in manufacturing and creating goods for consumers. Just the automation, supply chain, and all the moving parts create fun problems to solve. I dream one day of owning my own manufacturing company. With the recent news on the semiconductor shortage I started researching why there is a shortage and then looked into the manufacturing process and loved the complexity of it all.

I am curious the feasibility of one day starting a semiconductor production company. As I have read in order to compete at the bleeding edge you need to have an insane amount of capital in excess of billions of dollars. But what about one not specializing in the latest and greatest semiconductors. Surely, there has to be plants that specialize in semiconductors for things like small electronics like vacuums and toaster ovens. Are these still super expensive to start or could one feasible set one up for a more reasonable starting price like a few million.

Also, if it is still super expensive to set up a smaller, low end semiconductor fab do the bigger companies outsource certain steps like wafer production that one could possibly specialize into or is everything done in house?

I know the notion of starting one of these companies might be pretty niave but I always like to have a conversation before I dimiss ideas. Someone has done it in the past so why can't someone do it in the future.

Thanks for reading!

Most small fabs are not profitable untill recent shortage. Therefore the number of players just gets less and less.

 

[Paul2]

I think this also mostly covered by the MPW (multi-project wafer) offered by a lot of the regular foundries (including TSMC).

The gap in between MPW, and first minimally viable mass manufacturing is gigantic. This is where 200mm people dug in, and do extremely well.