Trump wants TSMC to take over Intel’s plants. That’s a terrible idea—here’s what needs to happen instead

[hist78]

"David B. Yoffie is a professor at Harvard Business School. Reed Hundt is a former chair of the FCC. Charlene Barshefsky is a former U.S. Trade Representative. James Plummer is the former Dean of Engineering at Stanford. Professor Yoffie, Mr. Hundt, Ambassador Barshefsky, and Professor Plummer all served as longtime directors of the Intel board."

https://finance.yahoo.com/news/trump-wants-tsmc-over-intel-184726097.html?guccounter=1&guce_referrer=aHR0cHM6Ly93d3cuZ29vZ2xlLmNvbS8&guce_referrer_sig=AQAAALZbokDJOnyHaPqDXpr672wPKtgnX42VufAC7q7_1lW2JKdRg6r_IkBVQFAKKhQQ0MFeV21lagFsAveT8h4GkgVvu9ivrUbhSCWyVrQ-UvwbFoS69F2JIXIdNqAmCWNPuHoMLlgRhfsTBT0eqfszrS9BuJCSoiSYc4eFdl0J7bhU

 

[hist78]

During a recent interview with the YouTube channel "Acquired", TSMC founder Morris Chang stated that TSMC's policy is to courteously respond to every customer's or potential customer's request, even if some of those requests are irrational or crazy.

In this case, the request for TSMC to save or buy part of Intel is both irrational and crazy.

 

[Jert]

In the big picture, strategically to America, Intel the Foundry is far more important than Intel the Design House. There are plenty of superb (or better) design houses already, as the article has listed, one more or one less does not matter.

Also, being #1 Foundry is nice, but perhaps an extremely close #2 should also work for Intel. There are plenty of demands for N-1. No?

 

[Artificer60]

This quote had me scratching my head.

" The problem is that Intel’s technology in 2025 is at least a generation behind TSMC, and customers such as Nvidia and Qualcomm are unwilling to work with a potential competitor that’s lagging in technology."

With 18A scheduled to begin risk production sometime in the next 3 months (my estimate to launch by EOY), how on earth can you say Intel is at least a generation behind. I think the reason that Qualcomm and Nvidia haven't jumped on board the Intel bandwagons is far more likely to be an lack of willingness to bet their company on an unproven entity.

Now I could be all wet and 18A will not live up to expectations, but the technical papers certainly don't point to a process that is a generation behind. Oh, and did I mention that if it does launch by EOY, Intel's leading edge process will be available before TSMC's 2nm process. That doesn't sound behind to me.

When the authors make such a blatantly incorrect statement in their argument, it kind of calls their credibility into question in my mind even if other parts of their argument may be sound.

 

[siliconbruh999]

Recently Intel negative rumors have spiked heavily everyone is trying to discredit them

 

[Fred Chen]

While Intel Foundry is not split off from Intel Product, there's always the concern for AMD, Nvidia, or Apple that their designs are not completely safe when handed over to Intel Foundry, or that their business will not get sufficient priority compared to Intel Product. After all, Intel Product and Intel Foundry financials are still reviewed together as Intel at the end of the year.

On the other hand, splitting off Foundry will make Intel (Product or Foundry) lose the CHIPS money.

So Intel 18A value to potential customers can only be demonstrated on the upcoming Panther Lake.

 

[blueone]

This quote had me scratching my head.

" The problem is that Intel’s technology in 2025 is at least a generation behind TSMC, and customers such as Nvidia and Qualcomm are unwilling to work with a potential competitor that’s lagging in technology."

With 18A scheduled to begin risk production sometime in the next 3 months (my estimate to launch by EOY), how on earth can you say Intel is at least a generation behind. I think the reason that Qualcomm and Nvidia haven't jumped on board the Intel bandwagons is far more likely to be an lack of willingness to bet their company on an unproven entity.

I suspect they are looking at TSMC 2nm process as a foundry process, and 18A as likely an internal-only process for high volume production for, well, I'm not sure how long. I think you hit the nail on the head with the guess that customers are leery of IFS and 18A because both are unproven as foundry-ready.

Now I could be all wet and 18A will not live up to expectations, but the technical papers certainly don't point to a process that is a generation behind. Oh, and did I mention that if it does launch by EOY, Intel's leading edge process will be available before TSMC's 2nm process. That doesn't sound behind to me.

There's technical behind and foundry business behind. I think IFS is indeed business behind in multiple ways, most likely PDKs and IP readiness.

 

[blueone]

While Intel Foundry is not split off from Intel Product, there's always the concern for AMD, Nvidia, or Apple that their designs are not completely safe when handed over to Intel Foundry

I have always thought this concern was nonsense.

or that their business will not get sufficient priority compared to Intel Product. After all, Intel Product and Intel Foundry financials are still reviewed together as Intel at the end of the year.

This concern is more legitimate, but is easily solvable with production capacity contractual commitments from Intel. The situation with TSMC is really no different. If I were a chip design company contracting with TSMC for production in a process Apple or Qualcomm was using, I would need capacity commitments to make sure TSMC didn't sacrifice my volume needs in favor of increased wafer demand (or packaging demand) from a whale.

 

[BruceA]

I have always thought this concern was nonsense.

This concern is more legitimate, but is easily solvable with production capacity contractual commitments from Intel. The situation with TSMC is really no different. If I were a chip design company contracting with TSMC for production in a process Apple or Qualcomm was using, I would need capacity commitments to make sure TSMC didn't sacrifice my volume needs in favor of increased wafer demand (or packaging demand) from a whale.

If I think about it during N5/N4 at height of shortage in one place all the worlds supply of Apple, Nvidia, AMD, MediaTek, Qualcomm and others was decided by one company and a likely one person. The world needs more than one source. Like Airbus and Boeing we need choices

 

[benb]

"First, a world-leading Western chipmaker has to have its research arm located in the U.S. or the West, not Taiwan."

Portland advanced logic R&D could be converted to a national lab, like Sandia, Los Alamos or Fermi. The goal would be like Sematech, develop and implement a national semiconductor process, for national fab operators, not just Chandler; if GF Malta wants the process they get it too. Free. But not Israel or Ireland, and not Samsung or TSMC fabs.

I think going forward, a leading, advanced, free (for US operators) semiconductor process would be a huge advantage for the US. The main thing that drove fabs out of the US was the cost. R&D cost is one of the those costs that the US government can subsidize without picking winners. All the other options including the Yoffie solution ("guaranteed orders") are picking winners which leads to moral hazard.

 

[hist78]

"First, a world-leading Western chipmaker has to have its research arm located in the U.S. or the West, not Taiwan."

Portland advanced logic R&D could be converted to a national lab, like Sandia, Los Alamos or Fermi. The goal would be like Sematech, develop and implement a national semiconductor process, for national fab operators, not just Chandler; if GF Malta wants the process they get it too. Free. But not Israel or Ireland, and not Samsung or TSMC fabs.

I think going forward, a leading, advanced, free (for US operators) semiconductor process would be a huge advantage for the US. The main thing that drove fabs out of the US was the cost. R&D cost is one of the those costs that the US government can subsidize without picking winners. All the other options including the Yoffie solution ("guaranteed orders") are picking winners which leads to moral hazard.

Great idea, except a semiconductor R&D program that is not directly linked to or supported by business, revenue, profits, and results will be hard to sustain in the long term. The US government, under different leadership, can and will shrink or cancel this government subsidized R&D initiative overnight.

The DoE's national labs you mentioned have been facing the same problems over the years.

 

[benb]

There are some issues with the national labs, tracing back to the publicly listed operators of said labs (SAIC for one). As with other defense contractors, they know how to get paid, get good results for SAIC, while the mission is less important.

National labs do attract and retain good people, and despite the parasitic operator, the end result comes closer to the ideal than you would think. The mission still matters, somehow.

 

[blueone]

There are some issues with the national labs, tracing back to the publicly listed operators of said labs (SAIC for one). As with other defense contractors, they know how to get paid, get good results for SAIC, while the mission is less important.

National labs do attract and retain good people, and despite the parasitic operator, the end result comes closer to the ideal than you would think. The mission still matters, somehow.

This idea reminds me of IBM and their 2nm process. It doesn't sound like a winning plan.

 

[Artificer60]

"First, a world-leading Western chipmaker has to have its research arm located in the U.S. or the West, not Taiwan."

Portland advanced logic R&D could be converted to a national lab, like Sandia, Los Alamos or Fermi. The goal would be like Sematech, develop and implement a national semiconductor process, for national fab operators, not just Chandler; if GF Malta wants the process they get it too. Free. But not Israel or Ireland, and not Samsung or TSMC fabs.

I think this misses a key feature for semi-conductor development. Unlike most of the results coming out of the national labs your need to run the process at a reasonable volume just to debug it and verify the process works as intended. This won't work if the intent is to create a 'process' and then let others adopt it if they want to. Someone has to run their silicon on it to get the process HVM ready. Intel products have done that historically, you would need someone to fill that role in order to successfully replicate the semiconductor R&D process.

 

[Artificer60]

There's technical behind and foundry business behind. I think IFS is indeed business behind in multiple ways, most likely PDKs and IP readiness.

Completely agree. I wouldn't have batted an eye if they hadn't specifically called out the Intel was technically behind. From a foundry perspective I believe they are probably 2 generations beyond 18A, at best, from being able to offer a good (rather than just acceptable) foundry experience.

 

[Arthur Hanson]

I think this misses a key feature for semi-conductor development. Unlike most of the results coming out of the national labs your need to run the process at a reasonable volume just to debug it and verify the process works as intended. This won't work if the intent is to create a 'process' and then let others adopt it if they want to. Someone has to run their silicon on it to get the process HVM ready. Intel products have done that historically, you would need someone to fill that role in order to successfully replicate the semiconductor R&D process.

Having worked at a national lab, you are dead on. Governments are not the best on cost control or results in an efficient manner.

 

[blueone]

Completely agree. I wouldn't have batted an eye if they hadn't specifically called out the Intel was technically behind. From a foundry perspective I believe they are probably 2 generations beyond 18A, at best, from being able to offer a good (rather than just acceptable) foundry experience.

The other thing I wonder about with IFS... who are the voices of the customer? Specifically, I mean senior people who were experienced foundry customers, both people with business and technical expertise, who will know specifically what foundry customers will want from their foundry. IMO, you can't just guess at these requirements. Slides with IP lists and schedules aren't enough, IFS needs to know in detail up front what foundry customers really need. I hope IFS has people like this in senior roles. If not, I'll be negative on IFS's chances for success.

 

[Artificer60]

The other thing I wonder about with IFS... who are the voices of the customer? Specifically, I mean senior people who were experienced foundry customers, both people with business and technical expertise, who will know specifically what foundry customers will want from their foundry. IMO, you can't just guess at these requirements. Slides with IP lists and schedules aren't enough, IFS needs to know in detail up front what foundry customers really need. I hope IFS has people like this in senior roles. If not, I'll be negative on IFS's chances for success.

I would hope that Naga will be able to help fill that role. I believe his experience at Micron certainly makes him more qualified that most of Intel's management in that regard. I also think that is why it is critical that the next CEO be someone with manufacturing experience. Since Intel is the last IDM almost anyone they choose should bring something to the table there if they come from a manufacturing environment outside of Intel.

If only the board had a clue and had a succession plan in place. So much of this drama could have been avoided.

 

[siliconbruh999]

I would hope that Naga will be able to help fill that role. I believe his experience at Micron certainly makes him more qualified that most of Intel's management in that regard. I also think that is why it is critical that the next CEO be someone with manufacturing experience. Since Intel is the last IDM almost anyone they choose should bring something to the table there if they come from a manufacturing environment outside of Intel.

If only the board had a clue and had a succession plan in place. So much of this drama could have been avoided.

If board had a plan they would not be in this place at first

 

[blueone]

I would hope that Naga will be able to help fill that role. I believe his experience at Micron certainly makes him more qualified that most of Intel's management in that regard. I also think that is why it is critical that the next CEO be someone with manufacturing experience. Since Intel is the last IDM almost anyone they choose should bring something to the table there if they come from a manufacturing environment outside of Intel.

If only the board had a clue and had a succession plan in place. So much of this drama could have been avoided.

I'm not following your line of reasoning. Chandrasekaran has no foundry experience whatsoever, either working for a foundry or being a user/customer of one, according to his Intel bio. For manufacturing, I'm not qualified to judge his experience. As for leading an IDM, I believe that the only reasonable org structure is as a conglomerate, where manufacturing and products are operated as two separate companies. IMO, there's no evidence that one CEO can be found who can be an effective ultimate decision-maker for both.