You are currently viewing SemiWiki as a guest which gives you limited access to the site. To view blog comments and experience other SemiWiki features you must be a registered member. Registration is fast, simple, and absolutely free so please, join our community today!
Leading edge means the most advanced node, assuming the english words are used accurately. It's a lot more than national pride in my mind, it's about national security. China appears to be much more willing to do what's necessary to achieve their objectives than the US or its allies are, whatever is necessary. IMO, the US needs to wake up and compete.
What most people seem to miss in this conversation is that there is a distinct collection of skill sets required to design a leading edge process. And other than TSMC the only company that has demonstrated the ability to do that is Intel. Samsung has tried and shown how difficult it is despite throwing truckloads of money at the endeavor. Once that capability is gone the US would need another Apollo program to get it back. That took around a decade, cost $257B to carry out and didn't have to deal with an ever moving goal line.
I also fail to see what Intel foundry would have to offer that others aren't already providing if they don't provide leading edge offerings. IFS is just another also ran if they walk away from leading edge development. They don't need to oust TSMC from the top spot to be successful. They just need to make enough to continue funding leading edge development. I feel like the pie is plenty big enough to support two players here.
TSMC's gross margain is ~ 60%. and the operating margin was 49.6%.
This indicates that labor costs account for roughly 10%.
Assume U.S. labor costs are twice those in Taiwan. （TSMC’s cost in Arizona runs about 10% higher than TW.）
For every one dollar of sales, 40 cents is cost and 10 cents is labor.
If Intel can make the same wafer at 80 cents,with 20 cents labor, this will equal to 1 dollar.
In this case, Intel should not drop the foundary.
I haven’t factored in the potential impact of tariffs yet.
What most people seem to miss in this conversation is that there is a distinct collection of skill sets required to design a leading edge process. And other than TSMC the only company that has demonstrated the ability to do that is Intel. Samsung has tried and shown how difficult it is despite throwing truckloads of money at the endeavor. Once that capability is gone the US would need another Apollo program to get it back. That took around a decade, cost $257B to carry out and didn't have to deal with an ever moving goal line.
I also fail to see what Intel foundry would have to offer that others aren't already providing if they don't provide leading edge offerings. IFS is just another also ran if they walk away from leading edge development. They don't need to oust TSMC from the top spot to be successful. They just need to make enough to continue funding leading edge development. I feel like the pie is plenty big enough to support two players here.
I really agree with you
As a Japanese, I've seen the Japanese semiconductor industry stay away from semiconductor manufacturing for a long time.
And I feel like I've finally made a step up...
Losing cutting-edge technology and experience is very painful
I hope that the Japanese semiconductor industry will not repeat what the Japanese semiconductor industry has experienced.
TSMC's gross margain is ~ 60%. and the operating margin was 49.6%.
This indicates that labor costs account for roughly 10%.
Assume U.S. labor costs are twice those in Taiwan. （TSMC’s cost in Arizona runs about 10% higher than TW.）
For every one dollar of sales, 40 cents is cost and 10 cents is labor.
If Intel can make the same wafer at 80 cents,with 20 cents labor, this will equal to 1 dollar.
In this case, Intel should not drop the foundary.
I haven’t factored in the potential impact of tariffs yet.
Gross profit margin already accounts for direct labor costs associated with production. Although it does not include indirect labor costs, such as salaries for marketing or human resources personnel, I believe these represent a relatively small portion of expenses in semiconductor manufacturing.
TSMC's gross margain is ~ 60%. and the operating margin was 49.6%.
This indicates that labor costs account for roughly 10%.
Assume U.S. labor costs are twice those in Taiwan. （TSMC’s cost in Arizona runs about 10% higher than TW.）
For every one dollar of sales, 40 cents is cost and 10 cents is labor.
If Intel can make the same wafer at 80 cents,with 20 cents labor, this will equal to 1 dollar.
In this case, Intel should not drop the foundary.
I haven’t factored in the potential impact of tariffs yet.
When you factor in the following, this may be the case already:
1. USDTWD depreciation: TWD appreciated ~10% recently against the USD; so the labor cost should be equal now. If USD keeps dropping against TWD given that Taiwan has been running trade surplus against US for over 20 years, this (labor cost) gets even better for Intel.
2. Tariffs: This will directly and proportionally increase the cost of wafers coming in from Taiwan.
When you factor in the following, this may be the case already:
1. USDTWD depreciation: TWD appreciated ~10% recently against the USD; so the labor cost should be equal now. If USD keeps dropping against TWD given that Taiwan has been running trade surplus against US for over 20 years, this (labor cost) gets even better for Intel.
2. Tariffs: This will directly and proportionally increase the cost of wafers coming in from Taiwan.
Actually Singapore is a good choice. Their partner, UMC, has a fab there. Intel has packaging facilities in msia. Singapore has low tax, good infrastructure and reasonable incentives. However, Singapore PM mentioned publicly before that Intel is due to receive 10 billion incentives for their fab in Germany. So perhaps Intel requires very significant incentives. Beyond that , pat gelsinger is focusing on building new facilities in western hemisphere.
Personally, I think they should also consider how likely and how prompt the promised incentives will be delivered. For Singapore I think it will be very prompt.
Actually Singapore is a good choice. Their partner, UMC, has a fab there. Intel has packaging facilities in msia. Singapore has low tax, good infrastructure and reasonable incentives. However, Singapore PM mentioned publicly before that Intel is due to receive 10 billion incentives for their fab in Germany. So perhaps Intel requires very significant incentives. Beyond that , pat gelsinger is focusing on building new facilities in western hemisphere.
Personally, I think they should also consider how likely and how prompt the promised incentives will be delivered. For Singapore I think it will be very prompt.
You say you have a great next node. No one wants it. you cannot fund it so you stop work (you buy chips from a different foundry). You write papers how it is an awesome node. You try to license it to someone else.... they dont want it even if it is free. IBM2.0. The difference? IBM actually had a reasonable large foundry business before giving it away. Intel hired the people that saw the end of IBM foundry and then GF deciding to stop work on advanced nodes. They can help.
I think this is going the way of DEC, or AMD more than IBM! Not enough volume and really doesn’t seem LBT has any attachment to intel’s legacy as the keeper of Moore’s Law. He is a cold calculated venture capitalist with no experience nor prospective of the complexity of technology development nor manufacturing.
Don’t think a whale is showing up just because they know LBT, they are as cold and calculating and understand technology better than him.
I think this is going the way of DEC, or AMD more than IBM! Not enough volume and really doesn’t seem LBT has any attachment to intel’s legacy as the keeper of Moore’s Law. He is a cold calculated venture capitalist with no experience nor prospective of the complexity of technology development nor manufacturing.
Don’t think a whale is showing up just because they know LBT, they are as cold and calculating and understand technology better than him.
Yeah, I feel like this is the kiss of death for Intel.
If I'm a customer looking at who I want to fab my chips, why would I spend a second of my time to design on a process where the process designer hasn't committed to spend a dime to actually ramp that process. I just don't see this as a winning strategy. No risk, no reward and Mr. Tan has indicated that he isn't willing to take the risk.
Then there is the issue of the timeline. Intel has to be on a schedule that is at least as fast as TSMC, earlier would be better. A14 is supposed to go into production in the 2028 time frame. That means Intel 14A has to do the same. So customers need 2 years to design for the process at a minimum. That means they have to commit in 2026 for a 2028 launch and Intel has to start filling a shell with tools at about the same time to have ramped capacity in place.
My assumption is that in order to do this Intel needs to have a quality 1.0 PDK out by 2026? Given that assumption I'm not sure I see that happening given how little I've heard on Intel 14A progress. One single misstep and this house of cards will collapse.
Actually Singapore is a good choice. Their partner, UMC, has a fab there. Intel has packaging facilities in msia. Singapore has low tax, good infrastructure and reasonable incentives. However, Singapore PM mentioned publicly before that Intel is due to receive 10 billion incentives for their fab in Germany. So perhaps Intel requires very significant incentives. Beyond that , pat gelsinger is focusing on building new facilities in western hemisphere.
Personally, I think they should also consider how likely and how prompt the promised incentives will be delivered. For Singapore I think it will be very prompt.
UMC have just completed their Expansion of there 12i FAB here.
Sadly Singapore Govt doesnt have the pockets to drop "Leading Edge" money on a Foundry when the public perception here is that all the jobs will go to foreigners.
In Singapore are tearing up the golf courses for building projects, world leader in water recycling [Newater] and can get any amount of skilled labour needed from the surrounding region for peanuts!!
Am pretty sure Govt just launched a study on Nuclear Power too
UMC have just completed their Expansion of there 12i FAB here.
Sadly Singapore Govt doesnt have the pockets to drop "Leading Edge" money on a Foundry when the public perception here is that all the jobs will go to foreigners.
That’s true. I’m sure the Singapore govt has that money. But they are sonewhat frugal and won’t spent 10 billion incentives unless there are very concrete roadmap and commitment. I am guessing 1 to 2 billion is the max
UMC have just completed their Expansion of there 12i FAB here.
Sadly Singapore Govt doesnt have the pockets to drop "Leading Edge" money on a Foundry when the public perception here is that all the jobs will go to foreigners.
In Singapore are tearing up the golf courses for building projects, world leader in water recycling [Newater] and can get any amount of skilled labour needed from the surrounding region for peanuts!!
Am pretty sure Govt just launched a study on Nuclear Power too
When you factor in the following, this may be the case already:
1. USDTWD depreciation: TWD appreciated ~10% recently against the USD; so the labor cost should be equal now. If USD keeps dropping against TWD given that Taiwan has been running trade surplus against US for over 20 years, this (labor cost) gets even better for Intel.
2. Tariffs: This will directly and proportionally increase the cost of wafers coming in from Taiwan.
Bravo to LBT. Essentially, LBT is saying Intel’s cost at 14A is close to twice of TSMC’s 1.4nm —so who foots the bill? TSMC is about 60% margain. I don't know the cost of 14A at all For now
Gross profit margin already accounts for direct labor costs associated with production. Although it does not include indirect labor costs, such as salaries for marketing or human resources personnel, I believe these represent a relatively small portion of expenses in semiconductor manufacturing.
Office supplies
Insurance
Rent
Utilities
Salaries and wages (indirect ones)
Legal fees
Property taxes
Marketing
Repairs
Travel expenses
Accounting fees
License fees
Advertising cost
Bank charges Depreciation <-- Huge in semiconductor industry
Maintenance
Accounting expenditures
Other overhead costs Research and development <-- Huge in semiconductor industry
