For those who are wishing for a split of Intel, and to Intel management.

[delong.height]

To those who are praying for Intel selling IFS or split the company into 2.

Intel shouldn't even be splitting the company into 2. IDM has always and will be the biggest possibility for Intel to come back to the top. What they need to think is how to lower the cost of manufacturing. To me, they shouldn't be freezing Germany fabs, they should stopped it.

Building the fabs in preparation of Intel 18A while they are the lifeblood of next-gen products are just detrimental to the health of the company. They are going to move majority of the products in house. That means a ton of wafer that is going to be needed. Where can they be build? Currently it seems like much investment have to be made ahead of time, which hurts profit margin so much.

Intel should just license their 18A process node to Samsung, and let Samsung compete with TSMC. What Intel needs the most, is to have a better cost structure. Outsourcing the manufacturing part to Samsung, and keep low-footprint of manufacturing and R&D is the only way to go.

Samsung is backed by a nation, Intel doesn't. Samsung is making profit from the current memory shortage, Intel doesn't.
Samsung sucks at latest process node technology and PDK development, Intel doesn't, or at least better than Samsung.

Hence, Intel should just let Samsung to be the franchisee of 18A, Intel 3, and possibly any future gen.

It's too costly for them to even consider to build in US. None of them (govn't, companies) are supporting it with real dollars. So why should Pat Gelsinger even has the thought of doing so? CHIPS act is not enough until there is a second bill. And Intel can no longer afford it.

Franchise 18A has a couple of benefits.

1. 1. Samsung offers the cheapest price out of the 3, Intel can pay the best price for each wafer manufactured while being design compatible because Samsung want their first whale customer just like Intel beside itself. Intel has the volume and partnership to negotiate an even lower price than maybe Apple does with TSMC.
2. 2. It's costly to expand new fabs, and next-gen products need 18A capacity so bad. Count the products, Panther lake, Clearwater forest, Falcon Shores, next Rapid, existing 3-party fabless customers
3. 3. Beside TSMC, Samsung has the most EUVs. And Intel doesn't need high-NA for their 18A. Current EUV should be enough. Samsung can just 'Copy Exactly' Intel's fabs. The whole process can be done literally in months, not waiting for fab shell readiness, equipments, technicians.
4. 4. Intel should keep growing its packaging revenue as that is easier to scale. It requires less money and less resources while there are plenty of demands. Put less money in fab; while more in packaging. Don't give out everything to Samsung. Outsource a portion of your business to grow your revenue and margin more quickly. To Intel management, investors aren't going to give you more times for what you want to accomplish. Do you want a 50/50 split of manufacturing footprint? Or do you want 60/40 margin ahead of 2030. You are fighting 3 wars at same time, and that won't be easy. All requires resources, while IFS requires the most. You can reorient your business while being IDM. TSMC isn't going to let your margin be saved, find a partner, Samsung. Don't bet foolishly on Germany, or Ohio. Outsource your 18A to Samsung to keep your future products saved. And to have your margin grows ahead of 2030. That's what investors are asking for. Not some 50/50 split on footprint. That is just foolish!
5. 5. Don’t cut Altera, and you don’t need the money to build fabs. Who know where you are in four years? Maybe Intel you are dead. Don’t over promise the fabs. Think smart. Every investment you have reinvested so far has not been that good.

You are in a very dynamic market, Pat Gelsinger, you didn’t anticipate PC crash in sales after COVID, you didn’t anticipate ongoing AI boom. Don’t lock yourself inside a box. You need flexibility. Being in this debt burden isn’t going to help you, Intel, or us the investors.

For Intel management again. Act fast, act quick, reduce workforce if you need to, STOP THE CAPEX, FIND A PARTNER, DELIVER YOUR PRODUCTS, and achieve that 60/40 margin goal by 2026. You need one year from now on to have a competitive products in all segments to be able to get that margin again. Don’t miss your schedule.

 

[hist78]

To those who are praying for Intel selling IFS or split the company into 2.

Intel shouldn't even be splitting the company into 2. IDM has always and will be the biggest possibility for Intel to come back to the top. What they need to think is how to lower the cost of manufacturing. To me, they shouldn't be freezing Germany fabs, they should stopped it.

Building the fabs in preparation of Intel 18A while they are the lifeblood of next-gen products are just detrimental to the health of the company. They are going to move majority of the products in house. That means a ton of wafer that is going to be needed. Where can they be build? Currently it seems like much investment have to be made ahead of time, which hurts profit margin so much.

Intel should just license their 18A process node to Samsung, and let Samsung compete with TSMC. What Intel needs the most, is to have a better cost structure. Outsourcing the manufacturing part to Samsung, and keep low-footprint of manufacturing and R&D is the only way to go.

Samsung is backed by a nation, Intel doesn't. Samsung is making profit from the current memory shortage, Intel doesn't.
Samsung sucks at latest process node technology and PDK development, Intel doesn't, or at least better than Samsung.

Hence, Intel should just let Samsung to be the franchisee of 18A, Intel 3, and possibly any future gen.

It's too costly for them to even consider to build in US. None of them (govn't, companies) are supporting it with real dollars. So why should Pat Gelsinger even has the thought of doing so? CHIPS act is not enough until there is a second bill. And Intel can no longer afford it.

Franchise 18A has a couple of benefits.

1. 1. Samsung offers the cheapest price out of the 3, Intel can pay the best price for each wafer manufactured while being design compatible because Samsung want their first whale customer just like Intel beside itself. Intel has the volume and partnership to negotiate an even lower price than maybe Apple does with TSMC.
2. 2. It's costly to expand new fabs, and next-gen products need 18A capacity so bad. Count the products, Panther lake, Clearwater forest, Falcon Shores, next Rapid, existing 3-party fabless customers
3. 3. Beside TSMC, Samsung has the most EUVs. And Intel doesn't need high-NA for their 18A. Current EUV should be enough. Samsung can just 'Copy Exactly' Intel's fabs. The whole process can be done literally in months, not waiting for fab shell readiness, equipments, technicians.
4. 4. Intel should keep growing its packaging revenue as that is easier to scale. It requires less money and less resources while there are plenty of demands. Put less money in fab; while more in packaging. Don't give out everything to Samsung. Outsource a portion of your business to grow your revenue and margin more quickly. To Intel management, investors aren't going to give you more times for what you want to accomplish. Do you want a 50/50 split of manufacturing footprint? Or do you want 60/40 margin ahead of 2030. You are fighting 3 wars at same time, and that won't be easy. All requires resources, while IFS requires the most. You can reorient your business while being IDM. TSMC isn't going to let your margin be saved, find a partner, Samsung. Don't bet foolishly on Germany, or Ohio. Outsource your 18A to Samsung to keep your future products saved. And to have your margin grows ahead of 2030. That's what investors are asking for. Not some 50/50 split on footprint. That is just foolish!
5. 5. Don’t cut Altera, and you don’t need the money to build fabs. Who know where you are in four years? Maybe Intel you are dead. Don’t over promise the fabs. Think smart. Every investment you have reinvested so far has not been that good.

You are in a very dynamic market, Pat Gelsinger, you didn’t anticipate PC crash in sales after COVID, you didn’t anticipate ongoing AI boom. Don’t lock yourself inside a box. You need flexibility. Being in this debt burden isn’t going to help you, Intel, or us the investors.

For Intel management again. Act fast, act quick, reduce workforce if you need to, STOP THE CAPEX, FIND A PARTNER, DELIVER YOUR PRODUCTS, and achieve that 60/40 margin goal by 2026. You need one year from now on to have a competitive products in all segments to be able to get that margin again. Don’t miss your schedule.

I'm afraid the scope you're talking about is still too huge and challenging, even during good times when revenue and profits are high.

 

[coldsolder215]

We've seen this logic play out before with IBM, in fact they have this exact strategy in place with Japan and it isn't looking good. Put up a wall between technological development and manufacturing and you'll get fun research projects with terrible economics. Intel might as well adopt the IBM death drive and continuously devour itself to prop up the dividend.

 

[delong.height]

We've seen this logic play out before with IBM, in fact they have this exact strategy in place with Japan and it isn't looking good. Put up a wall between technological development and manufacturing and you'll get fun research projects with terrible economics. Intel might as well adopt the IBM death drive and continuously devour itself to prop up the dividend.

The point I'm trying to argue is that for more expansion, they should outsource the manufacturing to Samsung. Intel can produce a limited or just a fraction of what Samsung can fab. They just don't need Germany or perhaps Ohio because it's expensive.

 

[MKWVentures]

The IBM route is:
1) Be a leader in IDM and processing with a wafer cost that is higher than competition
2) Claim that you will do foundry to get scale and lower costs.
3) change from foundry to partnerships on foundry where they utilize your knowledge
4) The partners realize they are not cost effective and they do not like your processes anymore.
5) Give away the factories in exchange for capacity, Become advisor to others on semiconductor technology. IBM is now very successful with this model.

Perhaps Intel should skip steps 2,3,4 unless they existing fabs are already cost effective (which they do not seem to be)

 

[delong.height]

The IBM route is:
1) Be a leader in IDM and processing with a wafer cost that is higher than competition
2) Claim that you will do foundry to get scale and lower costs.
3) change from foundry to partnerships on foundry where they utilize your knowledge
4) The partners realize they are not cost effective and they do not like your processes anymore.
5) Give away the factories in exchange for capacity, Become advisor to others on semiconductor technology. IBM is now very successful with this model.

Perhaps Intel should skip steps 2,3,4 unless they existing fabs are already cost effective (which they do not seem to be)

My proposal is to keep IFS, but keep the footprint low. Franchise your technology to Samsung. Let Samsung be the only alternative for Intel's internal design team. What this does is 5 benefits:

1. Keep wafer cost low, while the TD and IFS and Samsung keep working to get the yield better and PDK to be on track.
2. Get scale of manufacturing from Samsung. Samsung is the only one that can compete with TSMC truthfully in terms of cost because it has a strong support from a nation. Korea need more jobs, and Samsung can offer that.
3. Samsung need Intel's volume. It is a whale customer, bigger than AMD at least. Samsung does not have any more customers for their leading edge (3nm or under it) at least not in near future. If IFS is having such a problem getting new customers, so does Samsung because their process is just not good.
4. The design is basically compatible across IFS and Samsung if copy exactly. You don't pay TSMC and let them get bigger, you pay Samsung to make TSMC weaker as a result.
5. Customers who favor 18A but turned down by its cost can transition to Samsung and have their wafer done, while packaging done at IFS. But if you hand over the opportunity to TSMC, TSMC is going to eat you up end-to-end.

 

[MKWVentures]

My proposal is to keep IFS, but keep the footprint low. Franchise your technology to Samsung. Let Samsung be the only alternative for Intel's internal design team. What this does is 4 benefits:

1. Keep wafer cost low, while the TD and IFS and Samsung keep working to get the yield better and PDK to be on track.
2. Get scale of manufacturing from Samsung. Samsung is the only one that can compete with TSMC truthfully in terms of cost because it has a strong support from a nation. Korea need more jobs, and Samsung can offer that.
3. Samsung need Intel's volume. It is a whale customer, bigger than AMD at least. Samsung does not have any more customers for their leading edge (3nm or under it) at least not in near future. If IFS is having such a problem getting new customers, so does Samsung because their process is just not good.
4. The design is basically compatible across IFS and Samsung if copy exactly. You don't pay TSMC and let them get bigger, you pay Samsung to make TSMC weaker as a result.
5. Customers who favor 18A but turned down by its cost can transition to Samsung and have their wafer done, while packaging done at IFS. But if you hand over the opportunity to TSMC, TSMC is going to eat you up end-to-end.

I like the concept.... but Samsung got technologies from IBM years ago.... I dont think they are happy with the results. GF wasnt too happy either.

IMO the key challenge for Intel is: Intel is still very very good at compute architecture. Intel used to use a very expensive fab technology to provide leadership since others could not keep up. It all worked out well in the end. Now, Intel doesnt have process leadership and so it is not worth having an expensive process. AMD and Nvidia dont even use TSMCs leading node. So the only way to make it worth using internal processes is if Intel wafer cost matches TSMC wafer cost.... which seems difficult. We shall see.

My only model to make this work: Intel forms separate JV with other companies "US Fabs" or "AmFab" ... they all can license technology if they want it. Or not. If they can get CLOSE to TSMC price (10-20% higher), they will have a very strong business IMO.

 

[delong.height]

I like the concept.... but Samsung got technologies from IBM years ago.... I dont think they are happy with the results. GF wasnt too happy either.

I don't think IBM can help Samsung much, Samsung need a better partner that is not out of touch. IBM is known for its research expertise in semiconductors, but it doesn't' do manufacturing. So the end result is it may works well on paper or in the lab, but isn't practical or can't be a good resource when HVM is the most necessary portion. Intel can do it in this regard.

So the only way to make it worth using internal processes is if Intel wafer cost matches TSMC wafer cost.... which seems difficult.

The economies of scale only go down over time if Intel doesn't have good volume as TSMC. TSMC well benefit hugely while as Intel may benefit a little. So what might be good right now may no longer be or hold true 2 years from now. So it's always going to be for Intel to be at a disadvantage spot because they cannot expand any further as the cost of doing foundry in US/west is worse/higher than that of Asia. Hence any more capacity needed after the initial ramp, outsource those to Samsung, keep cost of the wafer lower, lower than TSMC and IFS because TSMC needs a good margin while Samsung doesn't and IFS won't have enough capacity. They need wins, wins, and wins. But Intel need to be at a health margin (60/40 as promised by the management), 50//50 of manufacturing footprint sounds good on paper, but will never be practical.