Foundry Monopoly in leading-edge Manufacturing: is it a real problem?

[user nl]

I have a question to this community of experts regarding Foundry Monopoly. I read quite some posts (also by our host) that people are worried that TSMC has entered into "monopoly" status regarding leading-edge semi manufacturing. And that people wonder how Intel will ever develop a meaningful Foundry business.

It reminds me a little to how ASML, during the 2006-2016 decade, gradually knocked out Canon and Nikon in the DUV lithography market (https://www.asml.com/en/news/stories/2021/twinscan-20-years-innovation) and successfully followed this "effective DUV-monopoly" by establishing a true EUV-monopoly during the last decade. So, ASML is now a real monopolist in the lithography market. Does anyone in the West care about ASML's monopolist status?
Is anyone in the West interested in building EUV (or DUV) litho-machines? It seems there are enough interesting other problems to work on.

I read some interesting stories recently about TSMC-Manufacturing, really impressive,
https://cwnewsroom.substack.com/p/why-intel-never-caught-up-to-tsmcanswer
https://cwnewsroom.substack.com/p/key-figure-behind-tsmc-us-expansion-yinglang-wang

It seems that TSMC-Foundry manufacturing is truly something special. It is not just the Foundry eco-system, not just about the leading-edge R&D. It seems that Manufacturing (leading-edge) semi is something very very special. And TSMC over some 40+ years has entered now effective monopolist status.

TSMC is rapidly expanding Manufacturing to other places during the last 5 years or so (USA, Japan, Europe) to keep all the clients in the West happy. And it seems that TSMC global Manufacturing expansion will only grow, now that TSMC has overcome its initial hesitance regarding increased costs, supplier-ecosystem and the "silicon shield" issue. TSMC is spending some 35-40% of revenue now in 2025 on capital investments, which semi-company can do that as well?
https://www.tomshardware.com/tech-i...tious-plans-detail-nine-production-facilities

So, as long as TSMC serves all their clients (and even Intel) well by efficient and trusted local Manufacturing, for a reasonable price (say gross margin 55-60%), what will be the real added value of Intel Foundry besides US national pride (and some national security related chip production). Like perhaps Canon and Nikon were for some time for Japan in lithography?

How will Intel be able to keep up in Foundry Manufacturing to sustain leading edge R&D (and now also in packaging) and the huge capital investments needed in a business where economy of scale and a true Manufacturing mindset seems so crucial?

And will Intel ever stay stuck in a Foundry business with say <5-10% leading-edge market share?

Any ideas and comments are appreciated.

 

[Xebec]

My personal take is ASML is a monopoly in a market niche -- EUV is only used / cost effective for "7nm" and below nodes, while DUV is still used by the vast majority of transistors fabricated today. It would certainly be ideal to have more competition here, but it's a bit different.

TSMC has the potential to become a full top to bottom foundry monopoly due to scale and being exclusive for leading edge nodes. Because of their general scale and size, and profitability from modern nodes, TSMC can theoretically outdo anyone on cost for any given node. All they need to do then is lower prices on older nodes (while building capacity) to effectively lock others out of the entire fab market.

There's also the innovation pipeline to consider. Intel has led quite a few innovations over the decades, and TSMC and Intel (and Samsung) have helped push each other forward through competition. If only TSMC is pushing the leading edge then we'll likely see slower progress than before (and we're already slowing down in terms of density scaling and performance improvements per node due to complexity, cost, etc.)

 

[count]

Monopolies are inevitable in Tech. They aren't necessarily bad.

Tech has winner take all dynamics, and having a monopoly is the prize for winning. And you win by being the best.

The thing about tech monopolies is they only get to be monopolies for a short period of time before some other disruptive tech comes along. Intel was effectively a monopoly for about 20 years, yet here we are talking about can Intel even compete. IBM was a monopoly in mainframe, and now the company is largely irrelevant.

The truth is TSMC and ASML monopolies won't mean much 20-30 years from now.

 

[Arthur Hanson]

As a long-term investor in TSM, they have entered a zone where only a few companies in other fields can operate. TSM is a one of a kind in the semi sector where they not only support themselves, but others in their ecosystem from suppliers to customers have a vested interest in TSM continuing to dominate the semi world. I see TSM dominating the leading edge until someone or some company figures a way to leapfrog their technology which I consider doubtful due to the ecosystem TSM has built from suppliers to work force to customers that all have a vested interest in continuing to dominate the semi sector. Even if there is a technological breakthrough, TSM will be the one to find it.

 

[Paul2]

Foundry Monopoly in leading-edge Manufacturing: is it a real problem?​

No one previously found any issue with Intel's monopoly in the leading edge

 

[Yunus]

No one previously found any issue with Intel's monopoly in the leading edge

I'm sure AMD people have a different opinion on that.

 

[count]

As a long-term investor in TSM, they have entered a zone where only a few companies in other fields can operate. TSM is a one of a kind in the semi sector where they not only support themselves, but others in their ecosystem from suppliers to customers have a vested interest in TSM continuing to dominate the semi world. I see TSM dominating the leading edge until someone or some company figures a way to leapfrog their technology which I consider doubtful due to the ecosystem TSM has built from suppliers to work force to customers that all have a vested interest in continuing to dominate the semi sector. Even if there is a technological breakthrough, TSM will be the one to find it.

It's unlikely anyone will leapfrog TSMC as a semiconductor foundry, just like no-one leapfroged Intel in PCs and no-one leapfrogged IBM in mainframes. At some point a new computing paradigm will come along (maybe quantum) as semiconductor based computing reaches it's limitations, and at that point being the best semiconductor manufacturer will no longer have the same relevance in the same way that no one really cares who the worlds leading vacuum tube manufacturing company is today.

But for the next 5-10 years I think TSMC has a license to print money.

 

[siliconbruh999]

It's unlikely anyone will leapfrog TSMC as a semiconductor foundry, just like no-one leapfroged Intel in PCs and no-one leapfrogged IBM in mainframes. At some point a new computing paradigm will come along (maybe quantum) as semiconductor based computing reaches it's limitations, and at that point being the best semiconductor manufacturer will no longer have the same relevance in the same way that no one really cares who the worlds leading vacuum tube manufacturing company is today.

But for the next 5-10 years I think TSMC has a license to print money.

Tbh Intel shot themselves In the foot big time and I don't think they will get their top of the world positive back unless something drastic happens. TSMCs license to print money is unlike any other as of rn.

 

[user nl]

It's unlikely anyone will leapfrog TSMC as a semiconductor foundry, just like no-one leapfroged Intel in PCs and no-one leapfrogged IBM in mainframes. At some point a new computing paradigm will come along (maybe quantum) as semiconductor based computing reaches it's limitations, and at that point being the best semiconductor manufacturer will no longer have the same relevance in the same way that no one really cares who the worlds leading vacuum tube manufacturing company is today.

But for the next 5-10 years I think TSMC has a license to print money.

I find it very interesting how photons are now entering more and more the leading-edge chips for (high-speed) data transport, not only laser-produced photons but also it seems LED-produced photons. Interesting new bet by TSMC: https://spectrum.ieee.org/microled-optical-chiplet
Much more energy efficient (and fast) to transmit bits by photons than by electrons.

If TSMC can make this successful integration of advanced packaging with (cheap) photons at scale they may have a couple of more years of "printing money". Energy efficiency in high-performance compute and bit-transport appears to be key for large scale affordable implementation of AI.

My two cents: semiconductor including photons may have a longer lifetime than a couple of decades........

 

[tedc]

Monopolies become a problem when companies abuse that position (e.g., to cut off competitors, to coerce suppliers, etc.). Intel was investigated multiple times by the FTC and fined by the European Commission for this.

Right now, we're talking about the potential for TSMC to abuse their position. As long as they continue to deliver what customers and partners want in a reasonable manner, it's difficult to fault them.

 

[hist78]

Tbh Intel shot themselves In the foot big time and I don't think they will get their top of the world positive back unless something drastic happens. TSMCs license to print money is unlike any other as of rn.

Sadly, Intel shot itself multiple times over the years. Every time Intel told people that they know what they are doing

 

[siliconbruh999]

Sadly, Intel shot itself multiple times over the years. Every time Intel told people that they know what they are doing

They had such a horrible management lol and since otlenni I don't know what Intel has been doing until PG took over.

 

[Arthur Hanson]

It's unlikely anyone will leapfrog TSMC as a semiconductor foundry, just like no-one leapfroged Intel in PCs and no-one leapfrogged IBM in mainframes. At some point a new computing paradigm will come along (maybe quantum) as semiconductor based computing reaches it's limitations, and at that point being the best semiconductor manufacturer will no longer have the same relevance in the same way that no one really cares who the worlds leading vacuum tube manufacturing company is today.

But for the next 5-10 years I think TSMC has a license to print money.

TSM is working on quantum computing technology. I have little doubt they will take the lead in this area to, maybe with partners if needed. Any company working in quantum computing would want TSM as a powerful and rich partner.

 

[Daniel Nenni]

Monopolies become a problem when companies abuse that position (e.g., to cut off competitors, to coerce suppliers, etc.). Intel was investigated multiple times by the FTC and fined by the European Commission for this. Right now, we're talking about the potential for TSMC to abuse their position. As long as they continue to deliver what customers and partners want in a reasonable manner, it's difficult to fault them.

I agree completely. As long as TSMC does not do predatory pricing I think they are okay. I do not see that happening at all with their customer centric approach to business. I do think, however, that competition is healthy and Intel or Samsung really need to step up. My personal bet is on Intel with Lip-Bu at the helm. Just give him some time. 14A might be the make or break node for Intel foundry. I really think the top semiconductor companies need to put there support behind Intel Foundry, just my opinion of course.

 

[user nl]

What specific added value is it that Intel (or Samsung) could bring to the leading-edge foundry clients by "just" competing with TSMC in Manufacturing? In what area of leading-edge Manufacturing TSMC needs competitive push?

My impression is that TSMC during these last 5-6 years has now mastered the EUV application in leading-edge Manufacturing to such a high level (see eg here: https://www.anandtech.com/show/2140...e-tools-more-wafers-best-pellicles-less-power), that their huge EUV-scale is simply providing an almost unsurmountable manufacturing advantage, relative to Intel Foundry. The inhouse EUV-manufacturing knowledge within TSMC must be quite intimidating for potential clients of Intel Foundry.

Furthermore, perhaps this huge client base of leading-edge customers of TSMC is far more important/relevant to push TSMC to keep improving their value to clients. In addition, it seems that half of the global semi-research community wants to collaborate with TSMC for further innovations (https://research.tsmc.com/english/collaborations/academic/academic-programs.html).

Indeed, as long as most clients simply want TSMC to succeed and TSMC provides value for their (leading edge) prices, it seems TSMC will enjoy many more years of "quasi monopoly" in leading-edge (foundry) manufacturing. Remember, TSMC originates from a rather small island in Asea, quite a different environment from a perhaps somewhat arrogant Intel culture based in superpower USA. Clients seem to simply like this (collaborative) TSMC culture in combination with TSMC's performance and scale.

In FY 2024 gross margin of TSMC was 56% and Intel 33%, below are the values from previous years. Hard to see how in this capital intensive business of foundry manufacturing clients would pick Intel over TSMC.

Culture, capital and customers may be the most important differences between TSMC and Intel foundry.......

 

[Daniel Nenni]

Competitive pricing is one thing, innovation is another. Intel brought us HKMG, FinFETs, BSPD, amongst many other technology firsts that TSMC now uses. Samsung Foundry not so much, just pricing.

 

[KevinK]

It seems that TSMC-Foundry manufacturing is truly something special. It is not just the Foundry eco-system, not just about the leading-edge R&D. It seems that Manufacturing (leading-edge) semi is something very very special. And TSMC over some 40+ years has entered now effective monopolist status.

There's a whole side of economics dedicated to "natural monopolies" where a natural monopoly occurs when a single firm can produce for an entire market at a lower average cost than multiple firms could. This happens due to significant economies of scale or economies of scope, meaning the cost per unit decreases as production increases or as related goods are produced. Examples include utilities like electricity, water, and telecommunications, where the high cost of initial infrastructure makes it more efficient to have one provider

Natural monopolies are only problematic if they use their sole source of supply to maximize profits and revenues. If that happens, governments will typically step in to either control pricing (utilities), or reduce barriers to entry for competitors (AMD vs Intel). Natural monopolies have a tendency to misallocate investments vs what the market actually wants if they do get caught up in revenue maximization. Natural monopolies are also prone to self-destruction due to failure failure to adapt to market substitutes and changing market conditions.

 

[DanX]

55-60% is a huge profit margin.
If Intel is 30% cost expensive than TSMC, It will be still 30% cheaper to manufacture the chip in-house.

 

[osnium]

Competitive pricing is one thing, innovation is another. Intel brought us HKMG, FinFETs, BSPD, amongst many other technology firsts that TSMC now uses. Samsung Foundry not so much, just pricing.

Samsung Foundry was first to GAA and they were also an early user of EUV. It’s not fair to say that they have only driven costs down

 

[user nl]

There's a whole side of economics dedicated to "natural monopolies" where a natural monopoly occurs when a single firm can produce for an entire market at a lower average cost than multiple firms could. This happens due to significant economies of scale or economies of scope, meaning the cost per unit decreases as production increases or as related goods are produced. Examples include utilities like electricity, water, and telecommunications, where the high cost of initial infrastructure makes it more efficient to have one provider

Natural monopolies are only problematic if they use their sole source of supply to maximize profits and revenues. If that happens, governments will typically step in to either control pricing (utilities), or reduce barriers to entry for competitors (AMD vs Intel). Natural monopolies have a tendency to misallocate investments vs what the market actually wants if they do get caught up in revenue maximization. Natural monopolies are also prone to self-destruction due to failure failure to adapt to market substitutes and changing market conditions.

My impression is that the global West (USA, Japan, Europe and others) has "decided" that TSMC may be that natural Foundry monopolist by (financially) stimulating the spread of TSMC's Manufacturing from the small island to those other regions in the world (not the gulf states). For Taiwan it may also be a way to build other global allies for its existential survival?

Who knows what will happen with Intel Foundry during the next 5 years........