Opinion: To make IDM 3.0 a success, Intel must make other companies IDM

[delong.height]

Why would the entire semiconductor industry, including supply chains, customers, and even Intel’s competitors, wait for Intel to figure out IDM 2.0, IDM 3.0, IDM X, or whatever “IDM” version comes next? The train left a long time ago, and most advanced logic players are already on board, except Intel and Samsung.

This “IDM xyz” approach worked in the past, when IDM players were the dominant force. Advanced logic IDMs like Intel and IBM used to be at the center of the universe. But those days are long gone.

Because foundry need to evolved, and foundry pure-play is unlikely to take place now given TSM has absorb all major revenues. For any others to survives (and they do deserve to), IDM plays a pivotal role, in which it's an evolution to existing business model.

You can argue that the train already left, but Intel has been the top contributor to the semiconductor industry.

Also, if I am Nvidia, I'm going to be super mad if TSM is giving same level of access to AMD, and I would want to beat it to death, so that even I lose all of competitive edge (developer mind, software support, etc). I can still win in terms of supply. And that is how Intel was able to win against AMD assuming I'm offering a competitive silicon and solution. I would want privilege, a better supply chain than everyone else.

 

[tomatoma]

Exactly, go on the route that make TSMC difficult to compete in by having a deeper specialization into things.

And that is true when most chips are becoming a commodity, the ability for fabless companies to be more vertically integrated is the most important way.

For example, in power semiconductors, vertical integration seems to be dominant over Foundry.
After all, I realize that Foundry is something that has specialized in logic.

 

[Daniel Nenni]

IDM is not a good idea unless you cannot get support for the manufacturing required for your field. You need scale and the flexibility of multiple customers. Both Pat and Swan 100% agreed that internal IDM cannot work (Its the math). Swan wanted outsource, Pat wanted IDM2.0 (or as I call it IBM2.0).

The Plan now is to get major customers just to allow 14A to ramp. Intel is using a 5 year old technology (Intel 7/10) to manufacture the majority of its wafers today and it has 70% CPU market share (including all wafers used for all products).

if no customers, 14A doesnt ramp (licensing agreement and codevelopment agreement with someone will follow... like IBM)

Here is a simple response. If you split Intel the design part will not be competitive with AMD, Nvidia, Amazon, Google, etc... Intel Design will be a cow out to pasture. AMD and Nvidia have close relationships with TSMC that Intel does not have. Pat Gelsinger made that happen. Can Lip-BU change that? Maybe but you will not see the changes for years to come. Meanwhile Intel design becomes less relevant.

Apple defined what a close foundry relationship should be 15 years ago with TSMC. TSMC has since expanded that to include other exclusive to TSMC customers like AMD and Nvidia. These companies have first line input into the process and PDK development plus packaging. We now call this "Process-Design Co-Optimization". That is a big advantage!

Intel will do the same for 14A, Intel Design will be included in that inner circle of course but there must be others and that is what Lip-Bu is doing. He has the relationships to make this happen, absolutely.

Samsung is now doing the same with Tesla, Process-Design Co-Optimization. Elon Musk probably has a different name for it... Walking the fab floor?

Bottom line: The IDM model is evolving to better compete with TSMC. How long will it take? I do not know. Maybe 5-10 years but either it will happen or we will only have TSMC which is a very big risk all things considered.

 

[hist78]

Here is a simple response. If you split Intel the design part will not be competitive with AMD, Nvidia, Amazon, Google, etc... Intel Design will be a cow out to pasture. AMD and Nvidia have close relationships with TSMC that Intel does not have. Pat Gelsinger made that happen. Can Lip-BU change that? Maybe but you will not see the changes for years to come. Meanwhile Intel design becomes less relevant.

Apple defined what a close foundry relationship should be 15 years ago with TSMC. TSMC has since expanded that to include other exclusive to TSMC customers like AMD and Nvidia. These companies have first line input into the process and PDK development plus packaging. We now call this "Process-Design Co-Optimization". That is a big advantage!

Intel will do the same for 14A, Intel Design will be included in that inner circle of course but there must be others and that is what Lip-Bu is doing. He has the relationships to make this happen, absolutely.

Samsung is now doing the same with Tesla, Process-Design Co-Optimization. Elon Musk probably has a different name for it... Walking the fab floor?

Bottom line: The IDM model is evolving to better compete with TSMC. How long will it take? I do not know. Maybe 5-10 years but either it will happen or we will only have TSMC which is a very big risk all things considered.

Could Mr. Pat Gelsinger invite Jensen Huang, Lisa Su, Tim Cook, and CC Wei, each separately, for a personal dinner? He could use the occasion to apologize for the not-so-kind words and videos directed at them during those stressful and challenging years as Intel's CEO.

Or has Mr. Li-Pu Tan already conveyed the apology on behalf of Intel?

 

[count]

China will never catch TSMC. Big diverse customers and the ecosystem drives TSMC and China does not have that. If you say 20-30 years? Sure, anything could happen. We already know what will happen 5-10 years out and it is all TSMC.

If China takes Taiwan that is another story but that is even more incentive to get Intel back on the leading edge.

Mature nodes are anyone's for the taking. TSMC has the best yield and economies of scale so if they want the mature node business it would be theirs.

Foundry Gross Margins for 2024:

TSMC 56.1% (Mature node gross margins are actually higher / Leading edge nodes are lower)
SMIC 18%
GF 24.5%
UMC 32.6%

But yes, the China Government has told China companies that they must buy China wafers if possible and still stay competitive. But the rest of the world will buy on wafer price.

Never say never. The reason I have confidence in this is the slowdown in the pace of Moore's law. If leading edge was still doubling transistors every 2 years, China would never be able to catch up, but that's not what's happening. All China has to do is keep up with N-2 for a few years, and leading edge economics will crumble.

What happens when N-2 is giving 80% performance at 30% of the cost? The demand for N will start to drop. When the demand for N drops, it's harder and harder to keep pushing N forward so N slows down further and the gap between N-2 and N grows narrower and narrower until there is no real distinction between N and N-2, and then competition becomes purely on price and not performance.

We are not that far away from this scenario, and the only thing pushing N forward now is AI.

 

[samwilde]

China will never catch TSMC. Big diverse customers and the ecosystem drives TSMC and China does not have that.

China has a lot of customers. SMIC's ecosystem isn't as large as TSMC's but there are plenty of clients with advanced designs. HiSilicon, Phytium, Loongson, Zhaoxin, Hygon, on CPUs. On AI chips you have HiSilicon, BIREN, Cambricon, etc.

If you say 20-30 years? Sure, anything could happen. We already know what will happen 5-10 years out and it is all TSMC.

It depends on if the Chinese can break the tools blockade or not. I would say they have the critical mass to do it.

Foundry Gross Margins for 2024:

TSMC 56.1% (Mature node gross margins are actually higher / Leading edge nodes are lower)
SMIC 18%
GF 24.5%
UMC 32.6%

GF and UMC are basically stuck on a time warp where fabs aren't built and processes aren't designed anymore. Not the case for SMIC. They are making massive investments in capacity.

But yes, the China Government has told China companies that they must buy China wafers if possible and still stay competitive. But the rest of the world will buy on wafer price.

Except the Chinese wafers are cheaper. At least on "legacy" nodes they are pushing the market prices down across the board.

 

[IanD]

Here is a simple response. If you split Intel the design part will not be competitive with AMD, Nvidia, Amazon, Google, etc... Intel Design will be a cow out to pasture. AMD and Nvidia have close relationships with TSMC that Intel does not have. Pat Gelsinger made that happen. Can Lip-BU change that? Maybe but you will not see the changes for years to come. Meanwhile Intel design becomes less relevant.

Apple defined what a close foundry relationship should be 15 years ago with TSMC. TSMC has since expanded that to include other exclusive to TSMC customers like AMD and Nvidia. These companies have first line input into the process and PDK development plus packaging. We now call this "Process-Design Co-Optimization". That is a big advantage!

Intel will do the same for 14A, Intel Design will be included in that inner circle of course but there must be others and that is what Lip-Bu is doing. He has the relationships to make this happen, absolutely.

Samsung is now doing the same with Tesla, Process-Design Co-Optimization. Elon Musk probably has a different name for it... Walking the fab floor?

Bottom line: The IDM model is evolving to better compete with TSMC. How long will it take? I do not know. Maybe 5-10 years but either it will happen or we will only have TSMC which is a very big risk all things considered.

The problem for Intel is that it's increasingly difficult for any IDM to compete with TSMC as far as process/IP/PDCO is concerned -- not just "I can build this transistor, see the ISSCC paper, mine's smaller than yours!" but being able to build it in vast volumes with good yield, which needs excellent PDCO and a big throughput of different designs to shake out the bugs in the process/PDK -- and then have a big variety of crucial IP available early enough in the PDK rollout to be useful, generated by a huge ecosystem of IP suppliers.

Without this they risk ending up with a process that might be theoretically available at the same time as TSMC, but in reality costs more per wafer, has lower yield, and has longer product TTM because the IP libraries are either not available or delivered later. All of which means non-Intel customers are put off because these are all significant downsides or increased risk, with no real positive side except being "not TSMC".

Your last line is true, but customers have to compare the definite risks with Intel Foundry (or Samsung...) described above with the possible long-term risk of a TSMC monopoly -- and TBH that's pretty much where the industry is right now anyway for advanced nodes. We've looked at Intel Foundry several times over the years (from inside more than one company), and even when they offered lower wafer prices than TSMC (to try and bring in business) still made the decision that this didn't compensate for the risk. Others who made the opposite decision got very badly burned (e.g. by the 10nm fiasco), it nearly destroyed one big telecomms company... :-(

 

[IanD]

Never say never. The reason I have confidence in this is the slowdown in the pace of Moore's law. If leading edge was still doubling transistors every 2 years, China would never be able to catch up, but that's not what's happening. All China has to do is keep up with N-2 for a few years, and leading edge economics will crumble.

What happens when N-2 is giving 80% performance at 30% of the cost? The demand for N will start to drop. When the demand for N drops, it's harder and harder to keep pushing N forward so N slows down further and the gap between N-2 and N grows narrower and narrower until there is no real distinction between N and N-2, and then competition becomes purely on price and not performance.

We are not that far away from this scenario, and the only thing pushing N forward now is AI.

Actually that's not true, what's also pushing N forward is the need for lower power consumption per bit/operation as demand for bandwidth and processing power keeps rising but power budget doesn't.

That doesn't just apply to AI but to mobile CPUs, networking, tablets/laptops -- anywhere where power/battery life is a constraint and/or higher throughput is needed. Cost is less of a driver, in fact often the next-generation chip is more expensive than the previous one.

If this wasn't true there wouldn't be such a big number of NTOs (New Tape Outs) in N2, and IIRC this is higher at the same stage of process maturity as any other previous recent node...

 

[count]

Actually that's not true, what's also pushing N forward is the need for lower power consumption per bit/operation as demand for bandwidth and processing power keeps rising but power budget doesn't.

That doesn't just apply to AI but to mobile CPUs, networking, tablets/laptops -- anywhere where power/battery life is a constraint and/or higher throughput is needed. Cost is less of a driver, in fact often the next-generation chip is more expensive than the previous one.

If this wasn't true there wouldn't be such a big number of NTOs (New Tape Outs) in N2, and IIRC this is higher at the same stage of process maturity as any other previous recent node...

I don't entirely disagree but even gains is performance/watt are slowing down. I think the underlying thesis of a narrowing gap between N and N-2 still holds, but it will take longer for the gap to close on power vs density/performance.

 

[tedc]

• Capital and Governance Integration: A core component of this model is allowing strategic investors like Arm, fabless companies, and SoftBank to invest in Intel Foundry Services (IFS) with shared governance rights. This joint governance structure transforms the relationship from a simple transaction into a strategic partnership of shared interests. With a seat on the board, clients can mitigate intellectual property concerns.
• 
  
• Business Model Redefinition: As Intel has already implemented, its product divisions (like CPU and GPU) are now treated as external clients, paying for foundry services and committing to wafer orders. This internal marketization provides a credible reference for external clients. More importantly, IDM 3.0 adopts the profit allocation logic of the IDM model: IFS would use rebates to return a portion of the "manufacturing profit" to key clients. This mechanism would boost client profitability, creating a mutually beneficial ecosystem and, through economies of scale, lowering Intel's own costs.

Interesting ideas! The shared investment and governance piece is probably more of a double-edged sword: Companies would need to invest to mitigate the conflicts of interest of IDM? Also, more stakeholders = more interests, and that's not always a good thing for a strategic business. The best way for this model to work might be to set up a new, co-invested company (i.e., similar to what Rapidus did...).

Having clients receive a portion of manufacturing profits is a bold idea. Maybe I missed it, but how would this lower Intel's own costs? Does this mean clients would pay (comparatively) more upfront to stand up a new process? If the goal is to drive volume and economies of scale, that might be a hurdle. Rebates would also take away money that should be re-invested into new processes and R&D.

1. Balancing "Partners" and "Neutrality": With limited resources, IDM 3.0 prioritizes a pragmatic strategy: focus on the most critical clients. This mirrors the influence Apple and Nvidia have on TSMC. The model acknowledges Intel’s financial and growth constraints by deeply binding with key players in the Arm ecosystem, securing crucial initial orders and capital while accepting a loss of "neutrality" for the broader market.

I'd argue that this is what Intel is doing now, and it's not working. The whole plan of "whale" clients sounds nice, but overlooks all the work leading up to it (e.g., establishing customer trust, building foundry services, nurturing partner ecosystems) that can't be created overnight.

1. Complexity of Investment Returns: The primary goal of IDM 3.0 is to lock clients into Intel’s end-to-end services, from wafers to packaging, testing, and assembly. This deep integration makes switching costs very high, securing stable returns. The ROI should be measured not only by the foundry's direct profits but also by the indirect enhancement of Intel's own and its partners' product competitiveness.

Any potential client would be wary of being locked in to a single company with high switching costs. The investment vs. risk balance just doesn't make sense when TSMC and Samsung are options too.

 

[hist78]

Capital and Governance Integration: A core component of this model is allowing strategic investors like Arm, fabless companies, and SoftBank to invest in Intel Foundry Services (IFS) with shared governance rights. This joint governance structure transforms the relationship from a simple transaction into a strategic partnership of shared interests. With a seat on the board, clients can mitigate intellectual property concerns.

SoftBank owns Arm Holdings PLC, and it recently invested $2 billion in Intel. But will Arm Holdings follow SoftBank’s step? Probably not.

The reason is simple: Arm is an important company in the semiconductor industry, but its revenue is relatively small, around $4 billion for its fiscal year 2025. Most of Arm’s revenue comes from licensing its technology and collecting royalties. Many of Arm’s customers are industry giants such as Apple, Qualcomm, MediaTek, Samsung, Nvidia, Broadcom, and TSMC.

Because most of these companies are direct competitors of Intel, Arm is unlikely to get involved too closely with Intel. Doing so would risk damaging its relationships with its customer base, which is the foundation of its business model.

 

[hist78]

Because foundry need to evolved, and foundry pure-play is unlikely to take place now given TSM has absorb all major revenues. For any others to survives (and they do deserve to), IDM plays a pivotal role, in which it's an evolution to existing business model.

You can argue that the train already left, but Intel has been the top contributor to the semiconductor industry.

Also, if I am Nvidia, I'm going to be super mad if TSM is giving same level of access to AMD, and I would want to beat it to death, so that even I lose all of competitive edge (developer mind, software support, etc). I can still win in terms of supply. And that is how Intel was able to win against AMD assuming I'm offering a competitive silicon and solution. I would want privilege, a better supply chain than everyone else.

We all agree that more competition is needed in the foundry business. But let’s not forget why fabless companies turned to pure-play foundries like TSMC in the first place: they wanted to compete against IDMs, or capture market opportunities that IDMs had no interest in pursuing.

So why should we suddenly believe these same fabless companies would be happy to prop up their long time rivals—Intel or Samsung, just so the IDM business model can survive? Why should the fabless industry put hard earned money into a “bundle deal” that locks Intel’s products and its foundry together?

With a pure-play foundry, fabless companies face challenges such as delivery schedules, capacity allocation, and profit swings sometimes tied to TSMC’s pricing. But when forced into a partnership to help an IDM like Intel or Samsung, there is a far bigger risk: live or death.

In Disney cartoons, a tiger, a monkey, and a rabbit might live happily together. But in the real world, even if the tiger offers his kitchen or bedroom for free, the monkey and rabbit will run away in no time. Unless, one day, the tiger truly transforms into a vegetarian.

 

[MKWVentures]

Seems like all of these comments make two assumptions:

1) Intel must exist. Everyone must make adjustments so Intel can exist
- I do not believe this is true. It was true 10 years ago. It is not now IMO, World moved on without Intel.
- This assumption leads to arrogance

2) Intel is a unmatched great and talented company whose management just screwed up. "Intel Exceptionalism"
- Intel has very smart people and a lot of assets. And a lot of baggage. Intel is not a top 10 tech company now.
- This assumption leads to arrogance


LBT knows this and will fix these if other people get out of his way.
A fabless Intel using outside manufacturing is worth 200B and has 50K employees and EPS of $3+.

 

[delong.height]

SoftBank owns Arm Holdings PLC, and it recently invested $2 billion in Intel. But will Arm Holdings follow SoftBank’s step? Probably not.

The reason is simple: Arm is an important company in the semiconductor industry, but its revenue is relatively small, around $4 billion for its fiscal year 2025. Most of Arm’s revenue comes from licensing its technology and collecting royalties. Many of Arm’s customers are industry giants such as Apple, Qualcomm, MediaTek, Samsung, Nvidia, Broadcom, and TSMC.

Because most of these companies are direct competitors of Intel, Arm is unlikely to get involved too closely with Intel. Doing so would risk damaging its relationships with its customer base, which is the foundation of its business model.

My assumption is that Arm want to move up in the value chain, to extract more value from its clients, or the end customers as well.

One of the way that they are doing now is by adding more IPs to Arm V9, and things like CSS, to create more value for the hyper-scalers. They are willing to attack its own customer, Qualcomm, to maximize its return. But by doing so, it is incredibly risky as well. So I don't worry that its partnership will put Arm in a dangerous spot because Arm/Softbank want to maximize its return. And lately I searched Softbank still has 90% of stake in Arm, and they also control Ampere.

So one of the better speculation that I would suggest to Arm CEO, is that they can work with IFS, negotiate a favorable deal for them and for its customers. In this way, they either get rebate from better participation throughout different part of manufacturing by referring customers toward Intel Foundry, or they can, perhaps do a crowdfund to secure a big wafer commitment. In such way, if they want to sell the chips (instead of design) or do custom silicon with companies like Amazon, Ampere, Google, and Microsoft and earn those portion of outsourced revenue like Broadcom and Marvell, they can do it.

I guess you may ask why they cannot go to TSM. I think they can, but TSM 's priority isn't on non-Apple, Nvidia for now. Will TSM bent the knee to optimize for Arm silicon, it is a huge if.

 

[delong.height]

We all agree that more competition is needed in the foundry business. But let’s not forget why fabless companies turned to pure-play foundries like TSMC in the first place: they wanted to compete against IDMs, or capture market opportunities that IDMs had no interest in pursuing.

So why should we suddenly believe these same fabless companies would be happy to prop up their long time rivals—Intel or Samsung, just so the IDM business model can survive? Why should the fabless industry put hard earned money into a “bundle deal” that locks Intel’s products and its foundry together?

With a pure-play foundry, fabless companies face challenges such as delivery schedules, capacity allocation, and profit swings sometimes tied to TSMC’s pricing. But when forced into a partnership to help an IDM like Intel or Samsung, there is a far bigger risk: live or death.

In Disney cartoons, a tiger, a monkey, and a rabbit might live happily together. But in the real world, even if the tiger offers his kitchen or bedroom for free, the monkey and rabbit will run away in no time. Unless, one day, the tiger truly transforms into a vegetarian.

1. Because only IDM can still afford the cost of doing foundry at leading edge.

2. If you look at Intel and Samsung, they have one thing in common. And that is, they are incredibly weak on design. Lip-Bu Tan said that they are not even top ten designers, and that is correct. The only reason that Intel still survived, instead of crushed by AMD in 2019/2020 is that they had great supply power. So they can go into a price war with AMD. And not every players in this industry have the luxury to tap into supply chain to have a complete control. Nvidia now does, but when it goes weak, they should wish that they have developed new moat so that they can still occupy 90% of the market.

3. To succeed, you need boldness from both side, it is not a game for weak. Only the paranoid survived! And I think that this applies to SoftBank and Intel, or perhaps more companies. And Masa is a very bold person, willing to take the risk.

4. Over time, Intel Foundry can grow increasingly independent from Intel Product.

 

[delong.height]

Seems like all of these comments make two assumptions:

1) Intel must exist. Everyone must make adjustments so Intel can exist
- I do not believe this is true. It was true 10 years ago. It is not now IMO, World moved on without Intel.
- This assumption leads to arrogance

2) Intel is a unmatched great and talented company whose management just screwed up. "Intel Exceptionalism"
- Intel has very smart people and a lot of assets. And a lot of baggage. Intel is not a top 10 tech company now.
- This assumption leads to arrogance


LBT knows this and will fix these if other people get out of his way.
A fabless Intel using outside manufacturing is worth 200B and has 50K employees and EPS of $3+.

a fabless Intel without a dedicated foundry partner is doomed for failure. It may work six years ago when AMD launched Zen, and Intel had competitive products back then.

 

[delong.height]

Interesting ideas! The shared investment and governance piece is probably more of a double-edged sword: Companies would need to invest to mitigate the conflicts of interest of IDM? Also, more stakeholders = more interests, and that's not always a good thing for a strategic business. The best way for this model to work might be to set up a new, co-invested company (i.e., similar to what Rapidus did...).

Having clients receive a portion of manufacturing profits is a bold idea. Maybe I missed it, but how would this lower Intel's own costs? Does this mean clients would pay (comparatively) more upfront to stand up a new process? If the goal is to drive volume and economies of scale, that might be a hurdle. Rebates would also take away money that should be re-invested into new processes and R&D.


I'd argue that this is what Intel is doing now, and it's not working. The whole plan of "whale" clients sounds nice, but overlooks all the work leading up to it (e.g., establishing customer trust, building foundry services, nurturing partner ecosystems) that can't be created overnight.


Any potential client would be wary of being locked in to a single company with high switching costs. The investment vs. risk balance just doesn't make sense when TSMC and Samsung are options too.

1. They only can take selected customers, and I bet the list isn't long. Arm, Apple, and Nvidia, Qualcomm (maybe but I guess they will go with Samsung). Whomever agreed shall put dedication and resource to build up its supply chain, not Intel's but theirs with Intel's resources.

2. It should always be a market fair price, whatever TSM charges, the same goes for Intel. The rebates only happened after the yield is stable on that node. The idea is to give the selected customers preferential treatments, whatever they can do for Intel design, they can do it for other customers.


And I think the concept of rebate boils down to a good feedback loop. The idea for IDM to return rebate when the production becomes stable. So that Intel product and others fabless can collectively invest in the next process node, knowing that whatever they invest in the foundry will yield more as times go on and as PPE deprecated fully. And I think being locked in to a single company is the wrong intention that I described, I shall say 'owning and have a deeper control of supply chain' is a better description.

 

[tooLongInEDA]

a fabless Intel with a dedicated foundry partner is doomed for failure. It may work six years ago when AMD launched Zen, and Intel had competitive products back then.

I do not buy these counsels of despair.

If Intel design is as weak as people keep suggesting (their supposition, not mine), it won't survive regardless of who fabs the chips.

If LBT is as good as people keep suggesting (which I do go along with), he should be able to turn around any (or at least the worst) failings in Intel design (given enough time).

 

[MKWVentures]

I do not buy these counsels of despair.

If Intel design is as weak as people keep suggesting (their supposition, not mine), it won't survive regardless of who fabs the chips.

If LBT is as good as people keep suggesting (which I do go along with), he should be able to turn around any (or at least the worst) failings in Intel design (given enough time).

EXACTLY: Intel can have competitive products again. LBT will fix the issues. Plus not having to deal with churn caused by IFS capacity strategy changes and priorities will speed product development.

 

[hist78]

My assumption is that Arm want to move up in the value chain, to extract more value from its clients, or the end customers as well.

One of the way that they are doing now is by adding more IPs to Arm V9, and things like CSS, to create more value for the hyper-scalers. They are willing to attack its own customer, Qualcomm, to maximize its return. But by doing so, it is incredibly risky as well. So I don't worry that its partnership will put Arm in a dangerous spot because Arm/Softbank want to maximize its return. And lately I searched Softbank still has 90% of stake in Arm, and they also control Ampere.

So one of the better speculation that I would suggest to Arm CEO, is that they can work with IFS, negotiate a favorable deal for them and for its customers. In this way, they either get rebate from better participation throughout different part of manufacturing by referring customers toward Intel Foundry, or they can, perhaps do a crowdfund to secure a big wafer commitment. In such way, if they want to sell the chips (instead of design) or do custom silicon with companies like Amazon, Ampere, Google, and Microsoft and earn those portion of outsourced revenue like Broadcom and Marvell, they can do it.

I guess you may ask why they cannot go to TSM. I think they can, but TSM 's priority isn't on non-Apple, Nvidia for now. Will TSM bent the knee to optimize for Arm silicon, it is a huge if.

TSMC has collaborated with many companies to optimize their products—AMD, Broadcom, Qualcomm, MediaTek, Ampere Computing, Nvidia, Apple, Intel, and even Arm’s IPs. Given this broad ecosystem, why would Arm risk taking sides by endorsing Intel’s IDM model with its money, reputation, and trust?