What are TSM's Plans for lower end chips?

[jms_embedded]

Pardon my skepticism but I don't see the auto chip shortage as over. I'll wait to hear what Kurt Sievers has to say in NXP's earnings call next week.

...and here we are, NXP's earnings call was today. Worth the read (or listen). They claim they're still supply-constrained.

NXP allegedly had 26% of the automotive MCU market at the beginning of 2021 (from an IHS Markit study) with Renesas at 30% and Infineon at 23%, so look at those three companies if you want a window on one part of the automotive supply situation. Two other major categories of automotive ICs would be analog ICs (including voltage regulators and PMIC) and power transistors, both of which are likely to be larger geometry and manufactured at in-house fabs of the auto/industrial semiconductor companies.

(After reading a number of these transcripts, I prefer hearing Sievers over the other European auto/industrial manufacturer CEOs: Jean-Marc Chery at ST, Jochen Hanebeck at Infineon. Sievers paints a clearer picture IMHO.)

Two highlights:

<--- begin excerpt --->
[Bill Betz, NXP CFO]
Now the good news is the inventory we have on hand is all long lived and has very low obsolescence risk. At the same time, I'd say, unfortunately, we're still constrained in several selective nodes. Remember, we buy about 60% externally. We do about 40% internally. That 40% really - more than two thirds of that internal capacity is linked to Auto and Industrial. So that's going to keep us at nice levels of utilization for the rest of the year.
<--- end excerpt --->

<--- begin excerpt --->
Joseph Moore (Morgan Stanley) -- Great. Thank you. I think you've talked about being for your Auto business having backlog coverage for the year. With the disruption in China, kind of would you still say that's the case? And as a follow-up, how are you guys thinking about NCNRs on your backlog this year? How flexible are you going to be if, for example, the China situation causes people to want to reschedule deliveries?

Kurt Sievers -- Yes. So Joe, first of all, the supply capability through the year. Clearly, the number of escalations has moderated. We still have a number of nastily short technologies. And I would call out 180 nanometers, 9055 gallium nitride [jms: I haven't heard the audio but I assume this should be transcribed "90 [nanometers], 55 [nanometers], gallium nitride"] and the high-voltage analog mixed signal, which is proprietary to NXP. This is, of course, in size less than it used to be, but it still leads to significant customer escalations and shortages, which we think will go through the year, but hopefully moderating towards the end of the year.

If we translate this back into supply capability, I think we said on the last call, we would be able to serve about 85% of kind of risk-adjusted backlog for the year. I'd say for this year, for '23, this is now more like 90% to 95%. So you see it's better. It's not yet on target. We are not yet in a position that we have visibility to serve everything we want, but we are coming closer.
<--- end excerpt --->

I don't know how much of the "nastily short technologies" Sievers mentioned are NXP's internal fabs, vs. foundry supply. NXP has in-house fabs down to 90nm (with 90nm from the Austin fab --- 200mm?! --- they inherited from Motorola) but outsources 90nm and below, so the 55nm is foundry and I would guess a chunk of the 90nm is also. From NXP's Q&A at Citi's Global Tech Conference in September 2022:

<--- begin excerpt --->
Chris Danely — Okay. Can you just maybe update everybody on your manufacturing strategy? How much is internal versus external and then where that’s trending in the future?

Kurt Sievers — Yeah. So, our manufacturing strategy for the back end, I stopped there because that’s simple and straightforward, is to do the majority in-house, so we are about 85%, we have our own factories. And we’ve also been able to scale them to the tune which was needed over this period. I mean, obviously, we also needed much more back-end manufacturing capacity, but that’s relatively quick to add. True, lead-times were long, but still that wasn’t really the source of shortage.

For the front end, we have about 60% from third parties and about 40% from our internal facilities, where the strategy is very straightforward. What we do internally is, is all in eight inch and at or greater than 90 nanometers. And it is very much geared towards proprietary special technologies. Think about gallium nitride, silicon-germanium, high-voltage mixed-signal technologies, which are proprietary and owned by NXP.

That’s what we what we use our own factories for, while all the more standard CMOS 90 nanometers all the way down to volume 16 nanometers, which we ship now in 14, is what we are getting from foundry partners. And again, that makes today about 60%. We think over time, it’s probably going to grow to 70% or even 75%.
<--- end excerpt --->

 

[Paul2]

Renesas is a very unique players in that they manufacture chips with effectively endless market lifetimes. We had a client which was buying a collection of Japanese MCUs from early nineties, if not eighties which were continuously for 30 years. Now it's all Renesas, but originally some of them were NEC, some Sharp, some Panasonic, and some from Japanese Co's you can't even google now.

They had a fire on the fab, and it apparently rocked hundreds of industries relying on "impossible to google chips" with extremely long support periods, from defence, to appliances, to medical devices.

 

[Daniel Nenni]

...and here we are, NXP's earnings call was today. Worth the read (or listen). They claim they're still supply-constrained.

NXP allegedly had 26% of the automotive MCU market at the beginning of 2021 (from an IHS Markit study) with Renesas at 30% and Infineon at 23%, so look at those three companies if you want a window on one part of the automotive supply situation. Two other major categories of automotive ICs would be analog ICs (including voltage regulators and PMIC) and power transistors, both of which are likely to be larger geometry and manufactured at in-house fabs of the auto/industrial semiconductor companies.

It will be interesting to see what the other automotive chip suppliers have to say. These companies are competing so there is overlap with customers.

I did not read that NXP will lose revenue in 2023 due to foundry wafer shortages nor have I read yet that car companies will not be able to sell vehicles due to chip shortages as in previous years. Given 2023 will be a down year and the foundries have been at 100% utilization but will soon be back to 80-90% I just don't see wafer shortages for 2023.

What I did read is that NXP is carefully managing inventory in anticipation of a possible slowdown in 2023. It was interesting to read that NXP has new 28nm chips and is also using FinFETs (16/14nm).

Here are TSMC's automotive numbers:

2022 +74%
2021 +51%
2020 -7%
2019 -7%

And if you compare these numbers to the automotive industry as a whole, even with more chips per car, you will see that the automotive chip shortages were self inflicted and should realign in 2023.

Bottom line: There is no reason for TSMC to build more fabs above 28nm, none whatsoever. Any company that is still dependent on chips from 200mm process technologies lack semiconductor vision and will again be punished in the not too distant future.

 

[nghanayem]

What I did read is that NXP is carefully managing inventory in anticipation of a possible slowdown in 2023. It was interesting to read that NXP has new 28nm chips and is also using FinFETs (16/14nm).

That is fascinating, although assuming the density is similar to foundry 28nm, this is not surprising. Planar devices were struggling real hard at the 30s and 20s and high-K metal gate was the only thing that even made them plausible in the first place. Now that finFETs are well understood it seems like it would make sense to move to that and HKMG ASAP once you get bellow ~45nm and leakage starts going through the roof.

 

[jms_embedded]

Renesas is a very unique players in that they manufacture chips with effectively endless market lifetimes. We had a client which was buying a collection of Japanese MCUs from early nineties, if not eighties which were continuously for 30 years. Now it's all Renesas, but originally some of them were NEC, some Sharp, some Panasonic, and some from Japanese Co's you can't even google now.

Microchip does that too, they call it "client-planned obsolescence". You can still buy PIC16C54A if you want, for goodness sake.

 

[jms_embedded]

Bottom line: There is no reason for TSMC to build more fabs above 28nm, none whatsoever. Any company that is still dependent on chips from 200mm process technologies lack semiconductor vision and will again be punished in the not too distant future.

That sounds a bit dogmatic to me... I guess we will just have to agree to disagree on some things.

(I do understand and agree that 300mm fabrication is more cost-effective for high volumes and has much, much better supply assurance of equipment than 200mm. My understanding of the 200mm/300mm quandary is that there are a lot of local cost/risk minima for decision-making --- including at TSMC --- that produce a non-optimal global scenario.)

I did not read that NXP will lose revenue in 2023 due to foundry wafer shortages nor have I read yet that car companies will not be able to sell vehicles due to chip shortages as in previous years. Given 2023 will be a down year and the foundries have been at 100% utilization but will soon be back to 80-90% I just don't see wafer shortages for 2023.

I would be curious how Sievers would have answered a question of wafer shortages specifically from the foundries (as distinct from internal fabs). Anyway I'm not too worried about 2023, I'm more worried about 2024 --- or whenever the economy picks up again --- and beyond. Perhaps TI will gain market share since they have lots of 300mm fab capacity coming online.

 

[hskuo]

Yes, but think of the fraction of the profit. N7 didn’t have all the capacity it has now back then. So to dilute their corporate margin by 1-2% means that it might have only been sold at like a 20 or 30% margin. If you walk it back to the old days I thought I remembered seeing that in the very early days of a new node TSMC would even sell wafers at or below cost.

From my old memory, TSMC did not sell wafers at or below "cost"- a tricky word here. The wafer price should be settled upon some utilization rate and depreciation and to be profitable. It is only when UT rate is too low and wafer "cost" could be higher than the wafer price of the new nodes. Besides, TSMC CFO re-iterated several times in the earnings call that the new node typically will match "corporate gross margin" in 6-8 quarters due to continuous yield/efficiency improvement and cost reduction activities.

 

[nghanayem]

From my old memory, TSMC did not sell wafers at or below "cost"- a tricky word here. The wafer price should be settled upon some utilization rate and depreciation and to be profitable. It is only when UT rate is too low and wafer "cost" could be higher than the wafer price of the new nodes. Besides, TSMC CFO re-iterated several times in the earnings call that the new node typically will match "corporate gross margin" in 6-8 quarters due to continuous yield/efficiency improvement and cost reduction activities.

I could totally be wrong here since back in the 40 and 28nm days I was in JR high, so any knowledge I have from that period is from reading and conversations with others. Obviously modern TSMC new node wafers are sold at a nice markup (probably less so in the risk starts phase). My comments of at or below cost was more so when you'd have folks like Altera being the first customer for like 40 or 28nm during the risk starts phase/when fab utilization was low. Obviously they would (most of the time) quickly improve utilization and wafer costs to having their nice like 30-40% margin (to my understanding this was still industry leading for the time). If my understanding is wrong on this, feel free to call me out here, as I'm always down to to learn stuff from those 1st/2nd generation fabless guys. It is also cool learning more about such a dynamic period in the semi industry. Where everybody was doing their own thing, there were like 3 or 4 relevant pureplay foundries, and IDM galore.

 

[Tanj]

.. back to the old days I thought I remembered seeing that in the very early days of a new node TSMC would even sell wafers at or below cost.

Is that "old days" as in "when TSMC was still playing catch-up"? That behavior makes sense for a company trying to establish its rep at the leading edge. Clearly, that no longer is needed and the benefits of leading edge can command a premium, even if it is slimmed down by having the largest customers agree to contracts spread over time which lock in the revenue matching the investment.

 

[nghanayem]

Is that "old days" as in "when TSMC was still playing catch-up"? That behavior makes sense for a company trying to establish its rep at the leading edge. Clearly, that no longer is needed and the benefits of leading edge can command a premium, even if it is slimmed down by having the largest customers agree to contracts spread over time which lock in the revenue matching the investment.

By old days I mean the 2000s and early 2010s. Back then TSMC was still the most advanced foundry. Samsung just doing their own thing/serving Apple with excess capacity, and UMC/Charted (later GF) were fast followers. TSMC's leading technology and cut above foundry services still allowed them a premium, but not to the same degree of today given folks like UMC could swoop in a bit later and act as second source. I think a large part of the lower margins can also be traced to how poor fabless firms were at the time. NVIDIA is kind of the poster child for tiny little fabless company going to mega giant with huge margins over the course of like 20 years.

 

[Daniel Nenni]

I could totally be wrong here since back in the 40 and 28nm days I was in JR high, so any knowledge I have from that period is from reading and conversations with others. Obviously modern TSMC new node wafers are sold at a nice markup (probably less so in the risk starts phase). My comments of at or below cost was more so when you'd have folks like Altera being the first customer for like 40 or 28nm during the risk starts phase/when fab utilization was low. Obviously they would (most of the time) quickly improve utilization and wafer costs to having their nice like 30-40% margin (to my understanding this was still industry leading for the time). If my understanding is wrong on this, feel free to call me out here, as I'm always down to to learn stuff from those 1st/2nd generation fabless guys. It is also cool learning more about such a dynamic period in the semi industry. Where everybody was doing their own thing, there were like 3 or 4 relevant pureplay foundries, and IDM galore.

TSMC was always the first foundry to a node so they got a wafer premium. At 28nm the other foundries botched it and didn't yield so TSMC had a monopoly for a brief period. This was the gate-first gate-last HKMG controversy. TSMC followed Intel with gate-first while the others (IBM, UMC, Samsung, and Charter) banded together and went gate-last.

At 40nm there were yield problems. This was back when they had recommended and required design rules. Companies like Nvidia thought they knew better and skipped the recommended rules, redundant vias being one of them. I worked with a competing GPU company and they yielded but Nvidia did not. Of course Nvidia implied it was TSMC and TSMC kept quiet. I did not and that was when I started SemiWiki.

 

[hist78]

TSMC was always the first foundry to a node so they got a wafer premium. At 28nm the other foundries botched it and didn't yield so TSMC had a monopoly for a brief period. This was the gate-first gate-last HKMG controversy. TSMC followed Intel with gate-first while the others (IBM, UMC, Samsung, and Charter) banded together and went gate-last.

At 40nm there were yield problems. This was back when they had recommended and required design rules. Companies like Nvidia thought they knew better and skipped the recommended rules, redundant vias being one of them. I worked with a competing GPU company and they yielded but Nvidia did not. Of course Nvidia implied it was TSMC and TSMC kept quiet. I did not and that was when I started SemiWiki.

"TSMC kept quiet."

TSMC seldom attacks or comments on its competitors let alone its customers, partners, or suppliers. It is a stark contrast to Intel's practice, such as:

Or this:

I think Intel's logic is that the more attacking on Intel's customers (or past customers) the more chance they will fall in love with Intel.

 

[nghanayem]

For what it is worth intel is B2B and B2C, whereas TSMC is only B2B. Other companies that have to sell to customers and build a desirable brand often do so by hyping up their own products and pointing out deficiencies with competitors products. This video was not "Hey Tim Cook look at us aren't you so sad you left us.", but rather "Hey PC and intel products have x y z capabilities go guy an intel PC not a Mac.". Not all that different from AMD talking about how their cores are more efficient than M1 or ADL.

As for the frenemies thing, as Dan mentioned I think this might just be from Pat's era/learning under Andy Grove. From what I can gather back then everybody used to be co-workers or at the very least well antiquated at some point. In the office we might be fierce competitors, but off hours we can grab a beer and reminisce about Fairchild/talk as engineers sort of things. It certainly sounds nice, but it doesn't seem like that is how things work anymore. And obviously intel marketing hasn't gotten this from Pat, since they've been doing their own thing for years.

Fortunately for TSMC, now that Morris is retired, TSMC has him to engage in shall we say "colorful exchanges" with Pat and others on their behalf .

 

[Daniel Nenni]

"TSMC kept quiet."

TSMC seldom attacks or comments on its competitors let alone its customers, partners, or suppliers. It is a stark contrast to Intel's practice, such as:
I think Intel's logic is that the more attacking on Intel's customers (or past customers) the more chance they will fall in love with Intel.

That was Morris Chang's TSMC. Today's TSMC is different. TSMC never forgets. Apple is the same. For anyone to think Apple will go back to Intel that is just nuts. Semiconductor people can forgive but semiconductor people do not forget.

 

[jms_embedded]

Infineon and Microchip had earnings calls today, very interesting.

Infineon highlights:

- "Semiconductor markets are bifurcating. Those driven by structural content like automotive and industrial
applications show robust dynamics. Others driven more by cyclically fluctuating demands, like consumer
computing and communications continue to be weak."
- "Automotive is fully booked for the 2023 fiscal year."
- "... we have some cancellations of long term orders but we see a healthy take in of new orders. Three quarters of the backlog consists of orders up to 12 months and the big majority is still in automotive."

Two excerpts worth noting in more detail:

<--- begin excerpt --->
Jochen Hanebeck (Infineon CEO)
Automotive recorded revenues of EUR1.872 billion, a slight sequential decline influenced by a less favorable U.S. dollar exchange rate. On an annual basis revenue was up by a staggering 35%. Here the U.S. dollar had a positive impact. But even assuming a constant exchange rate, year-over-year growth amounts to 26%, showing unabated strength of demand across our automotive product groups. In line with this, we were
once again able to bring up our profitability.

The segment result for the December quarter improved to EUR532 million equivalent to a segment result margin of 28.4%, up by more than 200 basis points compared to the previous quarter. The increase was mainly driven by price and mix effects. Overall supply disruptions, material shortages, as well as macro uncertainties still weigh on automotive production volumes, but to a lesser extent. According to S&P Global the number of cars produced worldwide in 2022 was 82 million, still below pre-pandemic levels. For 2023, S&P is predicting around 85 million units, which we deemed somewhat optimistic as a part of pent-up demand might be affected by the lack of consumer confidence.

Having said this, structural content growth remains by far more important driver for our business than unit evolution. From current interactions with our customers we learned a couple of things. First, the two automotive megatrends, e-mobility and ADAS are intact and irreversible. Second, there is a clear willingness to increase visibility and predictability of buying behavior and to secure semiconductor supply by engaging into schemes like capacity reservation agreements or longer dated committed orders.

Third, there is a strong tendency for OEMs to source strategic parts such as xEV specific components and MCUs either directly or as so called directed buys, and to strive for higher stock -- targets stock levels. In this regard, our AURIX MCU remains a critical bottleneck part, or bolt and screw [jms: mis-transcription, should be "golden screw"] as demand continues to outstrip supply despite the fact that this year we are ramping our selling rate to close to 1 million pieces per day.
<--- end excerpt --->

And then in Q&A, touching on which nodes Infineon has the most scarcity, Hanebeck called out 65nm and 40nm several times, including this exchange:

<--- begin excerpt --->
Janardan Menon (Jefferies) Just want to dive a little bit into your automotive business. Obviously the numbers you're reporting a very strong. Are you seeing any differences between the different parts of automotive? Or is it all quite uniformly strong? I'm saying different products like
MOSFETs, microcontroller, sensors, IGBTs, et cetera as well as between ICE and EVs. And there's a small follow-up. On the microcontroller side, are you saying that you're already getting as many wafers from TSMC, Global Foundries and you're now building up that inventory? Or is it that you expect the shortage to ease and availability to improve? And following that you're saying you will continue to build up? I'm just referring to TSMC is common during their conference call and I think they supply you the 65 nanometer that they would expect the microcontroller shortage to finish reasonably quickly.

Jochen Hanebeck Yeah, thanks for your question. So I said that we would like to take more wafers on in order to prepare for further ramp. AURIX is in full ramp mode. Now focus will shift slowly from 65 to 40 nanometers. We have gotten more wafers and therefore the -- let's say burning allocation will get better in the second half. But again, AURIX is ramping and therefore, we are securing capacities.

I think the 65 nanometer node at -- is still not really opening up to the extent we would like it to be. But I do expect customer escalations to ease for the microcontroller -- automotive microcontroller in the second half of '23. But we need more capacity to fuel the design wins we have achieved in the past. In general across the automotive business, of course, there is some hotspots in terms of strong demand being e-mobility, ADAS, also some analog mixed signal parts. Other parts are normalizing. And yeah, that's the way I would describe it.
<--- end excerpt --->

I'll post a summary from the Microchip earnings call when I have a chance; I've finished reading a transcript but it's late and I'm signing off for the evening.

 

[hist78]

Infineon and Microchip had earnings calls today, very interesting.

Infineon highlights:

- "Semiconductor markets are bifurcating. Those driven by structural content like automotive and industrial
applications show robust dynamics. Others driven more by cyclically fluctuating demands, like consumer
computing and communications continue to be weak."
- "Automotive is fully booked for the 2023 fiscal year."
- "... we have some cancellations of long term orders but we see a healthy take in of new orders. Three quarters of the backlog consists of orders up to 12 months and the big majority is still in automotive."

Two excerpts worth noting in more detail:

<--- begin excerpt --->
Jochen Hanebeck (Infineon CEO)
Automotive recorded revenues of EUR1.872 billion, a slight sequential decline influenced by a less favorable U.S. dollar exchange rate. On an annual basis revenue was up by a staggering 35%. Here the U.S. dollar had a positive impact. But even assuming a constant exchange rate, year-over-year growth amounts to 26%, showing unabated strength of demand across our automotive product groups. In line with this, we were
once again able to bring up our profitability.

The segment result for the December quarter improved to EUR532 million equivalent to a segment result margin of 28.4%, up by more than 200 basis points compared to the previous quarter. The increase was mainly driven by price and mix effects. Overall supply disruptions, material shortages, as well as macro uncertainties still weigh on automotive production volumes, but to a lesser extent. According to S&P Global the number of cars produced worldwide in 2022 was 82 million, still below pre-pandemic levels. For 2023, S&P is predicting around 85 million units, which we deemed somewhat optimistic as a part of pent-up demand might be affected by the lack of consumer confidence.

Having said this, structural content growth remains by far more important driver for our business than unit evolution. From current interactions with our customers we learned a couple of things. First, the two automotive megatrends, e-mobility and ADAS are intact and irreversible. Second, there is a clear willingness to increase visibility and predictability of buying behavior and to secure semiconductor supply by engaging into schemes like capacity reservation agreements or longer dated committed orders.

Third, there is a strong tendency for OEMs to source strategic parts such as xEV specific components and MCUs either directly or as so called directed buys, and to strive for higher stock -- targets stock levels. In this regard, our AURIX MCU remains a critical bottleneck part, or bolt and screw [jms: mis-transcription, should be "golden screw"] as demand continues to outstrip supply despite the fact that this year we are ramping our selling rate to close to 1 million pieces per day.
<--- end excerpt --->

And then in Q&A, touching on which nodes Infineon has the most scarcity, Hanebeck called out 65nm and 40nm several times, including this exchange:

<--- begin excerpt --->
Janardan Menon (Jefferies) Just want to dive a little bit into your automotive business. Obviously the numbers you're reporting a very strong. Are you seeing any differences between the different parts of automotive? Or is it all quite uniformly strong? I'm saying different products like
MOSFETs, microcontroller, sensors, IGBTs, et cetera as well as between ICE and EVs. And there's a small follow-up. On the microcontroller side, are you saying that you're already getting as many wafers from TSMC, Global Foundries and you're now building up that inventory? Or is it that you expect the shortage to ease and availability to improve? And following that you're saying you will continue to build up? I'm just referring to TSMC is common during their conference call and I think they supply you the 65 nanometer that they would expect the microcontroller shortage to finish reasonably quickly.

Jochen Hanebeck Yeah, thanks for your question. So I said that we would like to take more wafers on in order to prepare for further ramp. AURIX is in full ramp mode. Now focus will shift slowly from 65 to 40 nanometers. We have gotten more wafers and therefore the -- let's say burning allocation will get better in the second half. But again, AURIX is ramping and therefore, we are securing capacities.

I think the 65 nanometer node at -- is still not really opening up to the extent we would like it to be. But I do expect customer escalations to ease for the microcontroller -- automotive microcontroller in the second half of '23. But we need more capacity to fuel the design wins we have achieved in the past. In general across the automotive business, of course, there is some hotspots in terms of strong demand being e-mobility, ADAS, also some analog mixed signal parts. Other parts are normalizing. And yeah, that's the way I would describe it.
<--- end excerpt --->

I'll post a summary from the Microchip earnings call when I have a chance; I've finished reading a transcript but it's late and I'm signing off for the evening.

It looks like Infineon and TSMC will still have a wonderful 2023 in the automotive market.

 

[Daniel Nenni]

It looks like Infineon and TSMC will still have a wonderful 2023 in the automotive market.

It sounds to me that the automotive supply chain is still in disruption and end customers are still moving to "Just in Case" inventory versus "Just in Time". The next thing you will hear, as we did with TSMC, is that customer inventories are at a record high then comes the drop as customers burn excess inventory. Deja vu all over again.

One thing he did not mention is that car companies are starting to build their own chips so the supply chain will keep evolving. The other thing that I did not hear is that he will not make his revenue numbers in 2023 due to wafer shortages. That is very good news but I doubt TSMC's automotive revenue will jump 75% again this year. It will definitely be an up automotive year but 75% is a big increase. Even so, more wafers will go to automotive but there will still be plenty to go around.

Ford had an interesting call. That was one big piece of humble pie:

And while we generated record cash flow, we left about $2 billion of profit on the table due to cost and especially continued supply chain issues. These are the simple facts and to say I'm frustrated is an understatement because the year could have been so much more for us at Ford.

Ford Motor Company (F) Q4 2022 Earnings Call Transcript

Ford Motor Company (NYSE:NYSE:F) Q4 2022 Results Conference Call February 2, 2023 5:00 PM ETCompany ParticipantsLynn Antipas Tyson - Executive Director, Investor RelationsJim Farley -...

seekingalpha.com

 

[Daniel Nenni]

Fortunately for TSMC, now that Morris is retired, TSMC has him to engage in shall we say "colorful exchanges" with Pat and others on their behalf .

First of all TSMC would never do that, it is 100% against company culture. Second Morris would never do that, he speaks for himself. The man is 91 Years old. That is 637 in dog years.

 

[nghanayem]

First of all TSMC would never do that, it is 100% against company culture. Second Morris would never do that, he speaks for himself. The man is 91 Years old. That is 637 in dog years.

Of course. He is his own man and has been out of the saddle for years now and can do what he wants. My statement was merely in jest about some of the more "interesting" statements that both men have made. Regardless of my personal thoughts on some of the statements that were more "out there", the love and pride that these two men have for their past/current companies shows on their sleeves.

 

[hskuo]

TSMC was always the first foundry to a node so they got a wafer premium. At 28nm the other foundries botched it and didn't yield so TSMC had a monopoly for a brief period. This was the gate-first gate-last HKMG controversy. TSMC followed Intel with gate-first while the others (IBM, UMC, Samsung, and Charter) banded together and went gate-last.

At 40nm there were yield problems. This was back when they had recommended and required design rules. Companies like Nvidia thought they knew better and skipped the recommended rules, redundant vias being one of them. I worked with a competing GPU company and they yielded but Nvidia did not. Of course Nvidia implied it was TSMC and TSMC kept quiet. I did not and that was when I started SemiWiki.

Daniel: tsmc did follow intel with gate-"last" not gate-first. IBM/Samsung/GF/UMC went gate-first by claiming less process masks and layers.