Intel Delays its Ohio fabs

[TBiggs]

https://www.wsj.com/tech/intel-delays-20-billion-ohio-project-citing-slow-chip-market-713bde9e?mod=hp_lead_pos10

Which is it? The chip market is slowing, or Intel just trying to squeeze the Commerce Department? I still can't believe Intel is going to agree to the government's restrictions on stock buybacks, opening it's facility-level books, and potentially sharing profits anyway.

A leading edge fab in Ohio was all about politics and getting money for votes.

Ohio has no infrastructure for the major vendors ASML, LAM, AMAT, TEL and so many other. All have to build vendor support neither cheap and easy to find people who want to live in Suckeye country, LOL.

Yes there is “The World Famous. Ohio State” but I am skeptical you can find a willing talent if hundreds of engineers and technicians who want to make that place their long term home. A high tech fab isn’t like setting up a car factory yet!

 

[blueone]

A leading edge fab in Ohio was all about politics and getting money for votes.

Intel is a US corporation. Everything is about money. Every government action in a democracy is about votes.

Ohio has no infrastructure for the major vendors ASML, LAM, AMAT, TEL and so many other. All have to build vendor support neither cheap and easy to find people who want to live in Suckeye country, LOL.

Oregon was light on semiconductor infrastructure until Intel became successful in the 1980s. I have no doubt that vendors will follow Intel to Ohio.

Yes there is “The World Famous. Ohio State” but I am skeptical you can find a willing talent if hundreds of engineers and technicians who want to make that place their long term home. A high tech fab isn’t like setting up a car factory yet!

As for people wanting to live in central Ohio, it might be preferable to Upstate New York, Idaho, Austin, (the cost of living and traffic are annoying), New Mexico, Oregon (high cost of living)... I'd probably pick Phoenix if I wanted to work in a fab, but Phoenix has its own challenges and the climate surely isn't for everyone. In the big picture of where fabs are in the US, Ohio doesn't look like the worst place to live of the bunch.

 

[mozartct]

A leading edge fab in Ohio was all about politics and getting money for votes.

Ohio has no infrastructure for the major vendors ASML, LAM, AMAT, TEL and so many other. All have to build vendor support neither cheap and easy to find people who want to live in Suckeye country, LOL.

Yes there is “The World Famous. Ohio State” but I am skeptical you can find a willing talent if hundreds of engineers and technicians who want to make that place their long term home. A high tech fab isn’t like setting up a car factory yet!

By that yard stick, all fabs should still be in California.

There is no semi infrastructure in OH but there is manufacturing. There are also people, lots of them if you include nearby states.

In principle, I do not see why it could not work long term. Even TSMC has run out of Taiwan labor and is building in Kaohsiung. Key point is long term. I do not think that we have the political or financial fortitude to see it through. And intel is not a sound partner with everything else going on (Israel, AZ, Germany...).

So a vote grabber perhaps but not a (completely) nonsensical one. I think the larger issue is that manufacturing within the US is not strong, wherever you look. Costs too high, regulatory delays etc. Fabs are where they are for a reason and public monies is a smaller part of the story than people think (except for China of course)

 

[hskuo]

Fabs normally get GDS-II files. You can get much more from a GDS-II file than a mask. correct?

The true story known was about Oak Technology and UMC in ~1997 or 1999.

https://www.usitc.gov/publications/337/pub3186.pdf

 

[count]

I have worked on a variety of mega projects in multiple industry. They pretty much always get delayed. What ends up happening is complexity increases exponential with these massive projects of this scale but managements ability to handle complexity does not and linear thinking is applied to cost/schedules ect.

There are all kinds of logistical issues when you are building something on this scale that are non issues on smaller projects like:
"can the local power/water utilities support a project this large?"
"Is the labor market robust enough to hire the people we need"
"will the extra political attention on a project this scale lead to extra regulatory steps, public consultation, and legal challenges and are we prepared for those?"
"Can we even source this much concrete?"

These are things you don't even need to consider on a project at the $10-100mm capex scale but can cause massive problems when dealing with $1b+ projects, and catastrophic delays on $10b+ projects

 

[MKWVentures]

Incentives can cover / subsidized / offset partially the cost of labor, materials, and infrastructure. A larger long term challenge is setting up a capable, stable, workforce that can work within the local culture.

No question that between labor, culture and infrastructure certain countries and regions hold huge advantages that subsidies will be hard to overcome. If Asian didn’t have an overwhelming advantage wonder why they all went there, LOl

The number one reason for building in other countries is incentives.... by a long shot, Not labor, not culture.... infrastructure also. I have a pie chart of fab building and running costs. labor cost difference is <5% of the total fab cost. Incentive on most fabs in other countries average 25-30%.

have you looked at the cost breakdown of a Fab in Texas vs in Japan or Singapore?

 

[count]

The number one reason for building in other countries is incentives.... by a long shot, Not labor, not culture.... infrastructure also. I have a pie chart of fab building and running costs. labor cost difference is <5% of the total fab cost. Incentive on most fabs in other countries average 25-30%.

have you looked at the cost breakdown of a Fab in Texas vs in Japan or Singapore?

Labor isn't a huge cost, but lets say someone gave you unlimited money to build a fab in Antarctica, you are not going to be able to hire the people you need to run it at any cost.

 

[blueone]

Labor isn't a huge cost, but lets say someone gave you unlimited money to build a fab in Antarctica, you are not going to be able to hire the people you need to run it at any cost.

I've lived in Ohio, and it is nothing like Antarctica. Within 300 miles of Intel's site are numerous major universities in PA, IN, MI, and NY with excellent engineering and physics schools. When Intel expanded to Oregon I think it was a lot more challenging. Also, the cost of living in Central Ohio is very reasonable, which will be an attraction even compared to Phoenix. My sense is that Intel's challenges with the Ohio site will have more to do with process and manufacturing readiness and their ability to attract enough customers to make the investment pay off. The process problems are in Oregon, and the customer attraction problems are site independent.

IMO, the one site in the US that makes no human geographical sense at all is TSMC Washington. Even Micron's and GF's siting decisions make more sense.

 

[hist78]

I've lived in Ohio, and it is nothing like Antarctica. Within 300 miles of Intel's site are numerous major universities in PA, IN, MI, and NY with excellent engineering and physics schools. When Intel expanded to Oregon I think it was a lot more challenging. Also, the cost of living in Central Ohio is very reasonable, which will be an attraction even compared to Phoenix. My sense is that Intel's challenges with the Ohio site will have more to do with process and manufacturing readiness and their ability to attract enough customers to make the investment pay off. The process problems are in Oregon, and the customer attraction problems are site independent.

IMO, the one site in the US that makes no human geographical sense at all is TSMC Washington. Even Micron's and GF's siting decisions make more sense.

The "old" Intel has the volume and market potential to convince its suppliers and partners to follow Intel to a new place. Do you think Intel still have that absolute power to do that? Especially does the competition allow Intel to have enough time to do that?

For IFS' external customers, the fab location is important because it will have the effects on the cost, yield, volume, time for delivery, and political consequence.

 

[hist78]

The number one reason for building in other countries is incentives.... by a long shot, Not labor, not culture.... infrastructure also. I have a pie chart of fab building and running costs. labor cost difference is <5% of the total fab cost. Incentive on most fabs in other countries average 25-30%.

have you looked at the cost breakdown of a Fab in Texas vs in Japan or Singapore?

Yes, the subsidies/incentives is the number one consideration for selecting a new fab site. But quickly all other number two criteria will move in and decide the fate of a potential site.

Remember a typical subsidy or incentive only helps the initial construction and public infrastructure cost of a fab. Other incentives, such as property tax, school district tax, and R&D tax credit, can be more long term but are more limited and may have some strings behind it. The long term operation cost and impacts on business of a particular fab is too important to be ignored. Although I believe Intel lost sight on the long term effects on evaluating a fab site in the past two years.

 

[MKWVentures]

Yes, the subsidies/incentives is the number one consideration for selecting a new fab site. But quickly all other number two criteria will move in and decide the fate of a potential site.

Remember a typical subsidy or incentive only helps the initial construction and public infrastructure cost of a fab. Other incentives, such as property tax, school district tax, and R&D tax credit, can be more long term but are more limited and may have some strings behind it. The long term operation cost and impacts on business of a particular fab is too important to be ignored. Although I believe Intel lost sight on the long term effects on evaluating a fab site in the past two years.

Intel looks at all the costs and incentives and strings and makes the correct decision. I am pretty sure Intel knows the details of what the financials are better than the people blogging online. I have seen the contracts and I can state that they are more detailed than people on here understand..... but Intel actually knows all the details and the financial numbers. Where I think Intel is incorrect is in their ability to be successful in foundry and ramp .... although it is becoming clear that their strategies are more complex and nuanced than most people thought.

 

[MKWVentures]

Labor isn't a huge cost, but lets say someone gave you unlimited money to build a fab in Antarctica, you are not going to be able to hire the people you need to run it at any cost.

I think intel might know more about hiring where they have sites than we do. There are strategies to staff new fabs that intel will use and Intel already has a plan on sources of employees, timing, pay etc. Example: Boise and New Mexico can be very difficult to hire in. Fabs are successfully staffed multiple times with tried and true strategies. But again, Intel probably knows more about this than we do.

 

[blueone]

The "old" Intel has the volume and market potential to convince its suppliers and partners to follow Intel to a new place. Do you think Intel still have that absolute power to do that? Especially does the competition allow Intel to have enough time to do that?

Yes.

For IFS' external customers, the fab location is important because it will have the effects on the cost, yield, volume, time for delivery, and political consequence.

I get the political consequence part, and somewhat for cost (like water and power), but for Intel's Copy Exactly strategy, how are these other factors affected by geography?

 

[blueone]

Although I believe Intel lost sight on the long term effects on evaluating a fab site in the past two years.

You can't be serious. Intel has groups of people who do nothing but financial modeling and analysis. Unless these groups are specifically suppressed, I don't understand how the company as a whole loses sight of projects costing ten billion dollars or more.

 

[count]

I've lived in Ohio, and it is nothing like Antarctica. Within 300 miles of Intel's site are numerous major universities in PA, IN, MI, and NY with excellent engineering and physics schools. When Intel expanded to Oregon I think it was a lot more challenging. Also, the cost of living in Central Ohio is very reasonable, which will be an attraction even compared to Phoenix. My sense is that Intel's challenges with the Ohio site will have more to do with process and manufacturing readiness and their ability to attract enough customers to make the investment pay off. The process problems are in Oregon, and the customer attraction problems are site independent.

IMO, the one site in the US that makes no human geographical sense at all is TSMC Washington. Even Micron's and GF's siting decisions make more sense.

These days factories in the Midwest are struggling to attract people. In the US in general people don't see manufacturing as a good job anymore and Ohio is not where a lot of people want to live.

I have seen this firsthand at the last factory I spent time in at the Midwest. The local labor market was tapped out pretty quickly in spite of in theory having a large manufacturing presence and a skilled workforce to draw from, and very few people could be convinced to move there.

 

[blueone]

These days factories in the Midwest are struggling to attract people. In the US in general people don't see manufacturing as a good job anymore and Ohio is not where a lot of people want to live.

I have seen this firsthand at the last factory I spent time in at the Midwest. The local labor market was tapped out pretty quickly in spite of in theory having a large manufacturing presence and a skilled workforce to draw from, and very few people could be convinced to move there.

Fabs are not comparable to most other manufacturing jobs. Also, generalizing about "the Midwest" isn't much for accuracy. While Ohio as a state doesn't have impressive population growth, the area Intel chose is much better off:

Ohio is largely losing population and growing older except in the greater Columbus area, study shows - Ohio Capital Journal

The vast majority of the state of Ohio is growing older, losing population, and losing workers — a trend offset only by growth happening mostly in Columbus and surrounding areas, a new report shows.

ohiocapitaljournal.com

Note the map in the article.

If I were responsible for a new gigafab location in the US, and having thought about it after Intel made the Ohio announcement, I couldn't think of an obviously better place. Colorado Springs came to mind, but Intel must have thought of that already (Intel has a site there) and ruled it out. Dallas also came to mind, as did the Research Triangle, but Intel picked Ohio for the big site.

 

[staf]

Fabs normally get GDS-II files. You can get much more from a GDS-II file than a mask. correct?

Indeed, if you have the full mask set data you can derive the full functionality of the design from it; it's called 'circuit extraction'.

 

[blueone]

Indeed, if you have the full mask set data you can derive the full functionality of the design from it; it's called 'circuit extraction'.

But going from circuits to the functionality of the high-level logic design still sounds impossible. Is it?

 

[Tanj]

These are things you don't even need to consider on a project at the $10-100mm capex scale but can cause massive problems when dealing with $1b+ projects, and catastrophic delays on $10b+ projects

When you are Intel, who have build multi-Bn projects in multiple companies, there is probably depth of managerial understanding of how to do greenfield gigafab in a new country.

When you have done it only in Taiwan, there are probably multiple, even overwhelming, surprises in Arizona. Compare with Samsung, who are not on their first rodeo, and seem to be making good progress.

It makes sense that TSMC is much happier in Japan with a bunch of partners in a consortium who know how to navigate the politics and practicalities.

 

[Tanj]

But going from circuits to the functionality of the high-level logic design still sounds impossible. Is it?

Seems like an easy job for AI, especially for modern designs based on cell libraries. Once you factor out the low level clutter and the buffers the circuit net will be more obvious. It may be tough to reconstitute the high level logic that humans understand since early steps in EDA may convert clusters of logic to primitive gates like NAND which function correctly but are, in quantity, gibberish to humans. But there might be tools around to help, maybe not on the open market but developed by folks interested in sucking in outside IP.