TSMC to make advanced 2-nm chips in US sooner to meet AI demand

[hist78]

I'm honestly curious. What do you think the process to make this feasible looks like?

I’d like to hear what @MKWVentures thinks since he was there.

From my perspective as an outsider, I believe Intel could have implemented many cost cutting and productivity improvement initiatives long ago, similar to the measures Lip-Bu Tan is executing today. As an IDM, Intel should, in theory, enjoy a higher net profit margin than a pure play foundry like TSMC, because Intel can capture the added value from the final products. However, Intel’s overall costs have remained very high for a long time, especially compared with TSMC. This was true even when TSMC wasn't as dominant as it is today.

TSMC vs. Intel annual net profit margins (2004–2024)

 

[MKWVentures]

What are some of the biggest differences between how Micron operates vs Intel Foundry? Is one key difference the fact that Micron does most of its manufacturing outside the US?

manufacturing outside US impact is minimal. text or call me and I can discuss specifics

 

[MKWVentures]

It is my personal opinion that Intel Foundry is working to change their operational model so that they can show this degree of flexibility. The guy calling the shots was brought in from Micro, so he should have some idea about what needs to change to make Intel more flexible and nimble.

Intel has the leadership at executive level. But the other 40K people need to change their behavior. Its not easy if it is even possible.

 

[Artificer60]

Intel has the leadership at executive level. But the other 40K people need to change their behavior. Its not easy if it is even possible.

In my opinion the only really hard thing about changing people's behavior is finding the patience to wait for the change to take place. If leadership wants to see a change all they have to do is reward the behaviors they want to see. But they have to do this consistently over several years. People are generally pretty smart when it comes to figuring out how to get ahead. Once it becomes clear what is being rewarded the desired behaviors will follow.

I actually think this could be done more quickly, but an annual reward cycle (promos, raises, stock awards, etc.) will cause the pace of change to move more slowly, since many will not believe the change in expected behavior is going to last. In short you have to stick with it long enough to convert the skeptics to believers. And you have to be willing to suffer though a period of worse before better that isn't very palatable to most executives. I guess it remains to be see if Lip-Bu Tan has the stomach to tolerate that with Intel Foundry. His work at Cadence leads me to believe he does, but only time will tell.

And if you think that the scale of the problem prevents it from being possible, I would hold up Satya Nadella's work at Microsoft as a proof point that changes can be made even in very large corporations.

 

[BruceA]

Intel has the leadership at executive level. But the other 40K people need to change their behavior. Its not easy if it is even possible.

To change leadership needs to drive relentlessly, that has been the hallmark of any successfully company as well as any turnaround story

 

[Brady]

To change leadership needs to drive relentlessly, that has been the hallmark of any successfully company as well as any turnaround story

It also needs to be paired with a coherent strategy that is not shifting under people’s feet all the time. Pivoting is one thing, chaos is another.

Waiting to see if the steady hand that aims in the right direction is here with the new leadership.

 

[siliconbruh999]

You need to not keep employees in fear as well,consistent layoffs cuts affect people's morale.

 

[MKWVentures]

Side note: TSMC N2 is 3 layer GAA transistor. Intel is 4 layer. is samsung or anyone else planning 4? is this a big deal? @Scotten Jones

 

[tooLongInEDA]

I’d like to hear what @MKWVentures thinks since he was there.

From my perspective as an outsider, I believe Intel could have implemented many cost cutting and productivity improvement initiatives long ago, similar to the measures Lip-Bu Tan is executing today. As an IDM, Intel should, in theory, enjoy a higher net profit margin than a pure play foundry like TSMC, because Intel can capture the added value from the final products. However, Intel’s overall costs have remained very high for a long time, especially compared with TSMC. This was true even when TSMC wasn't as dominant as it is today.

TSMC vs. Intel annual net profit margins (2004–2024)

Thanks, that's very interesting data. Looking at TSMC in isolation, it seems to have quite stable net profit margins over the past 20 years. Does this give the lie to the periodic suggestion that TSMC is exploiting an effective monopoly to overcharge their customers ? If I read the data correctly (and it isn't more complex than it looks), they wouldn't appear to have been doing so just yet. Unless they've been doing it consistently for the last 20 years (surely not) !

But on the direct Intel-vs-TSMC comparison, my recollection was that 2000s Intel was extrordinarily profitable with the highest margins in the industry, so I wonder if these net margin numbers give the whole picture here.

 

[Artificer60]

To some degree Intel's margins are going to be affected by their business model. They do not run nodes for nearly as long as TSMC so they don't get the benefit of the long tail running an older process on a fully depreciated tool set. While not an accountant by any means I have to assume that the rapid process turnover is a drag on margins. That is part of the reason Intel needs to transition to a foundry model, to get better returns on their R&D investments over time.

 

[Barnsley]

Thanks, that's very interesting data. Looking at TSMC in isolation, it seems to have quite stable net profit margins over the past 20 years. Does this give the lie to the periodic suggestion that TSMC is exploiting an effective monopoly to overcharge their customers ? If I read the data correctly (and it isn't more complex than it looks), they wouldn't appear to have been doing so just yet. Unless they've been doing it consistently for the last 20 years (surely not) !

But on the direct Intel-vs-TSMC comparison, my recollection was that 2000s Intel was extrordinarily profitable with the highest margins in the industry, so I wonder if these net margin numbers give the whole picture here.

Hopefully its their customers who get squeezed and not their suppliers.

Tiz a different story further down the food chain , whereby lesser Foundries love nothing more than to shaft their suppliers to try and prop up their own inefficiences for creating margin on their products.

 

[hist78]

Thanks, that's very interesting data. Looking at TSMC in isolation, it seems to have quite stable net profit margins over the past 20 years. Does this give the lie to the periodic suggestion that TSMC is exploiting an effective monopoly to overcharge their customers ? If I read the data correctly (and it isn't more complex than it looks), they wouldn't appear to have been doing so just yet. Unless they've been doing it consistently for the last 20 years (surely not) !

But on the direct Intel-vs-TSMC comparison, my recollection was that 2000s Intel was extrordinarily profitable with the highest margins in the industry, so I wonder if these net margin numbers give the whole picture here.

I believe Intel’s high profit margin was a common perception throughout the 2000s, but it wasn’t the reality. Intel’s gross profit margin was indeed high, yet its operating and net profit margins dropped significantly once all expenses and costs were taken into account.

The problem is that, despite being an IDM and theoretically able to capture more value and profit than a pure play foundry, Intel’s high costs have eroded its net profit margin.

A company’s net profit ultimately determines its ability to reinvest and stay competitive. In semiconductor manufacturing, heavy and continuous capital investment is essential for maintaining an edge. After more than 20 years, it’s not surprising that Intel is now facing so many challenges.

Different long term management visions, execution capabilities, and financial disciplines have created significant and different results between TSMC and Intel.


Note:

Gross Profit = Revenue - the direct costs of producing goods or services

Operating Profit = Gross Profit - Operating Expenses (e.g., rent, wages, utilities, R&D, marketing, advertising) - Depreciation/Amortization

Net Profit = Operating Profit - interest, taxes

 

[Arthur Hanson]

How is ASML going to meet this high demand with HN-NA EUV? ASML will ship a total 60-70 production level EXE:5200 systems in the next four years. TSMC will need more than that for full HNA-EUV implementation at a given process node? Intel and Samsung will also need them for logic and then comes the memory makers.

Dan, will alternatives ever be developed for the ASML equipment or ways of getting more production out of existing equipment including the equipment used to build the ASML machines?

 

[Xebec]

Intel has the leadership at executive level. But the other 40K people need to change their behavior. Its not easy if it is even possible.

Would the Tower acquisition have helped this a lot? a little?

 

[Xebec]

Gross Profit = Revenue - the direct costs of producing goods or services

If TSMC's gross margin starts to match/exceed Intel's margin at their heights (62-63%), then that's a sign TSMC is getting (very?*) expensive on newer nodes:

- Intel was capturing both "fab" and "wholesale" chip profits in that 62-63%.
- GM %'s are super-linear: a 66.6% margin is twice the profit ($$) of a 50% margin, i.e. twice the cost to customers.

*TSMC's margin includes substantial amounts of legacy nodes that have/had competition, and are known to be lower margin than 7nm and below (per TSMC's own financial reports/history) - TSMC's latest nodes have to be *well* above that 56% average last year.

 

[MKWVentures]

Would the Tower acquisition have helped this a lot? a little?di

I am a huge fan of Tower and loved the acquisition plan. But that plan was for how to utilize older technologies. Tower doesnt know how to develop leading edge cost effective technologies either.

TSMC is unrivaled in developing leading technologies that are manufacturable. I have talked with people who went from Intel to TSMC and the development processes are quite different. TSMC has a development machine running. Intel did years ago.

 

[blueone]

Otelnni later admitted that he was wrong

Yeah, but he wasn't. PSO admitting he was wrong was just like Marc Benioff saying he was sorry for supporting federal troops in SF. Politically correct apologies are not real apologies. @Artificer60 got the reasoning right. And remember, PSO figured he'd have to build fabs for projected Apple volume. Would anyone here really recommend building new fabs to lose more money?

 

[Xebec]

Yeah, but he wasn't. PSO admitting he was wrong was just like Marc Benioff saying he was sorry for supporting federal troops in SF. Politically correct apologies are not real apologies. @Artificer60 got the reasoning right. And remember, PSO figured he'd have to build fabs for projected Apple volume. Would anyone here really recommend building new fabs to lose more money?

Sounds like a scenario even if he said yes -- the Intel culture may have made it a financial disaster because they wouldn't have been able to attract other Foundry customers for N-1, N-2 nodes as time marched on? (leaving idle fabs)