Array
(
    [content] => 
    [params] => Array
        (
            [0] => /forum/index.php?threads/tsmc-4th-qtr-2020-earnings-discussion.13609/
        )

    [addOns] => Array
        (
            [DL6/MLTP] => 13
            [Hampel/TimeZoneDebug] => 1000070
            [SV/ChangePostDate] => 2010200
            [SemiWiki/Newsletter] => 1000010
            [SemiWiki/WPMenu] => 1000010
            [SemiWiki/XPressExtend] => 1000010
            [ThemeHouse/XLink] => 1000970
            [ThemeHouse/XPress] => 1010570
            [XF] => 2021370
            [XFI] => 1050270
        )

    [wordpress] => /var/www/html
)

TSMC 4th QTR 2020 Earnings Discussion

Daniel Nenni

Admin
Staff member
Here is the Video:


The transcript:


(Presentation material is attached)

Press release:

Hsinchu, Taiwan, R.O.C., January 14, 2021
– TSMC (TWSE: 2330, NYSE: TSM) today announced consolidated revenue of NT$361.53 billion, net income of NT$142.77 billion, and diluted earnings per share of NT$5.51 (US$0.97 per ADR unit) for the fourth quarter ended December 31, 2020.

Year-over-year, fourth quarter revenue increased 14.0% while net income and diluted EPS both increased 23.0%. Compared to third quarter 2020, fourth quarter results represented a 1.4% increase in revenue and a 4.0% increase in net income. All figures were prepared in accordance with TIFRS on a consolidated basis.

In US dollars, fourth quarter revenue was $12.68 billion, which increased 22.0% year-over-year and increased 4.4% from the previous quarter.

Gross margin for the quarter was 54.0%, operating margin was 43.5%, and net profit margin was 39.5%.

In the fourth quarter, shipments of 5-nanometer accounted for 20% of total wafer revenue; 7-nanometer and 16-nanometer accounted for 29% and 13% respectively. Advanced technologies, defined as 16-nanometer and more advanced technologies, accounted for 62% of total wafer revenue.

“Our fourth quarter business was supported by strong demand for our industry-leading 5-nanometer technology, driven by 5G smartphone launches and HPC-related applications,” said Wendell Huang, VP and Chief Financial Officer of TSMC. “Moving into first quarter 2021, we expect our business to be supported by HPC-related demand, recovery in the automotive segment, and a milder smartphone seasonality than in recent years.”

Based on the Company’s current business outlook, management expects the overall performance for first quarter 2021 to be as follows:

  • Revenue is expected to be between US$12.7 billion and US$13.0 billion;
And, based on the exchange rate assumption of 1 US dollar to 27.95 NT dollars,

  • Gross profit margin is expected to be between 50.5% and 52.5%;
  • Operating profit margin is expected to be between 39.5% and 41.5%.
The management further expects the 2021 capital budget to be between US$25 billion and US$28 billion.
 

Attachments

  • 4Q20Presentation(E).pdf
    607.2 KB · Views: 382
Last edited:
$25-28b in Capex, and here we were all thinking it would be around $20b.
Yeah...that was the number that surprised the heck out of me. Tim Culpan (in Bloomberg Opinion) is stating TSMC "is making a massive move to obliterate rivals".

Culpan's key points:
1. TSMC may want to front-load spending because clients need the capacity now, not tomorrow.
2. Any customer (Apple, AMD, Qualcomm, Nvidia, carmakers) left waiting for supply is naturally inclined to shop around, so TSMC has the chance to use its resources to keep business away from rivals could be worth the money.
3. TSMC wants to win orders from Intel rather than letting them go to Samsung. If TSMC can prove its worth, Intel may shift to an "asset-light model".
4. By snapping up equipment supply from ASML and Lam Research, TSMC can slowly suck up the oxygen competitors need to breathe

Samsung is also growing & getting more competitive/customers...which TSMC sees as a long-term threat. From my limited view, I counted 3 customers for Samsung's 5nm node:
1. Ambarella's CV5 5nm 8K AI Vision Processor
2. Samsung Exynos 2100 mobile processor
3. Qualcomm's Snapdragon 888 mobile processor
 
Yeah...that was the number that surprised the heck out of me. Tim Culpan (in Bloomberg Opinion) is stating TSMC "is making a massive move to obliterate rivals".

4. By snapping up equipment supply from ASML and Lam Research, TSMC can slowly suck up the oxygen competitors need to breathe

My understanding is that this is a key consideration in the timing of Intel's decision to go fabless. If they want to remain an IDM, they need to make a move to secure equipment soon.

Supply is very tight right now for semis, and it's good to see TSMC committed to making sure they will have enough capacity available to it's current and future customers. Very happy to count myself among TSMCs shareholders today.
 
2. Any customer (Apple, AMD, Qualcomm, Nvidia, carmakers) left waiting for supply is naturally inclined to shop around, so TSMC has the chance to use its resources

Remember, TSMC is not Amazon with next day delivery. Wafer agreements are put in place before the design starts and it takes TSMC less time to build fab capacity than it takes to design a leading edge SoC/CPU/GPU.

If there is a wafer shortage TSMC customers are selling more chips than expected and it's not like they can get additional wafers from Samsung. Once a design is done to a fab it will stay at that fab.

TSMC's $28B CAPEX is a big tell as to what the future holds for the foundry landscape, absolutely.
 
That's a pretty big clue as to what is coming since TSMC builds fabs based on wafer agreements. My guess is that Intel is coming, absolutely.

This is pretty bullish news for Intel as well, if true (which I think it is). Assume $4-5b of that Capex is to build capacity for Intel. That's $4-5b that Intel doesn't need to spend. Of course there will be some up front costs and margins will go down slightly if they are paying TSMC for wafers instead of manufacturing in house, but probably not by that much. TSMC has already done the heavy lifting on process design so for them it's just incremental capacity. Intel probably ends up with at an extra $2b in free cash flow. They can pump that into architectural improvements.

Without in house fabs pulling it down, Intel is undervalued. A fabless Intel would be an extremely capital efficient cash flow machine.
 
This is pretty bullish news for Intel as well, if true (which I think it is). Assume $4-5b of that Capex is to build capacity for Intel. That's $4-5b that Intel doesn't need to spend. Of course there will be some up front costs and margins will go down slightly if they are paying TSMC for wafers instead of manufacturing in house, but probably not by that much. TSMC has already done the heavy lifting on process design so for them it's just incremental capacity. Intel probably ends up with at an extra $2b in free cash flow. They can pump that into architectural improvements. Without in house fabs pulling it down, Intel is undervalued. A fabless Intel would be an extremely capital efficient cash flow machine.

$20B TSMC CAPEX is what I heard late last year, it was raised to $17B in 2020 so a 15% bump is predictable. Adding another $8B on top of that is all Intel in my opinion. The new Intel CEO is a career Intel manufacturing guy so that could change. Interesting times for sure.
 
Internet rumor has it, TSMC will fab Intel core i3 H2 '21 on 5nm.

How likely might that be? If I understand Daniel, that's only possible if they designed it for TSMC N5 long time ago.

And didn't Apple already reserve ~90% of the N5(P) capa for 21?
 
Internet rumor has it, TSMC will fab Intel core i3 H2 '21 on 5nm.
How likely might that be? If I understand Daniel, that's only possible if they designed it for TSMC N5 long time ago.
And didn't Apple already reserve ~90% of the N5(P) capa for 21?

That is certainly possible. There are few secrets in the semiconductor ecosystem and I have not heard about Intel developing 5nm foundation IP, which is required before designing a chip. I only know about 3nm.

Apple uses a custom version of the TSMC processes that they do not share with other companies so no reservation is required.

TSMC clearly stated that 5nm usage will be mostly SoC and HPC which means AMD, NVIDIA, and others will also be getting 5nm wafers this year. TSMC also said the increased CAPEX will be used for both 5nm and 3nm capacity.
 
During the Q&A session, every analyst tried hard to find out if the huge Capex budget has something to do with the Intel's outsourcing deal. They asked directly and indirectly through several seemly unrelated topics.

Although CC Wei kept denying to comment on any questions specific to a particular customer, at the end it seems those analysts understand what the true answer is. And we will find it out next week from Intel's earnings conference call.
 
We had expected $20B in spend or something north of that as the additive spend of 3NM and Arizona would likely cost a bunch more money but $20B to $28B is a huge uptick that reeks of a high level of confidence in both customer demand as well as confidence in being able to implement the technology.

TSMC has previously shown that it is more than capable of modulating its spend during the year either up or down depending upon near term trends.

Starting the year with such a huge uptick says that much of that caution is thrown to the wind because the circumstances are just that good.

Customer Demand- TSMC is the fab for the entire industry

TSMC has Apple, AMD, Nvidia, Intel, Qualcomm and on and on, plus a host of foreign manufacturers. Its an endless list. Samsung obviously also supplies foundry services but doesn't hold much of a candle to TSMC.

On an incremental basis , TSMC now has to start supplying all the Apple laptop and desktop chips that Intel no longer supplies.

Even though Intel has a new CEO who will likely try to fix Intel's manufacturing problems, Intel will still have to outsource a significant amount of chips to TSMC including some new leading edge devices recently announced. Given how long it will take to fix Intel's issue, assuming that is the decision, it could be several years of increased demand from Intel on TSMC. We would be overly surprised to see Intel become TSMC's number two customer after Apple. We have heard that Intel is in the top five of TSMC's customers a while ago.

Then add a sprinkling of 5G and continued work/school from home and overall demand is high and not slowing any time soon.

High confidence in 3NM

With TSMC talking about "risk production" of 3NM in the second half of 2021 it seems pretty clear that they haven't hit any known roadblocks or issues that would throw their Moore's Law march off kilter. This would seem to match pretty well to getting 3NM up to full speed in time for Apple 2022 Iphone launch as well as further refinements of the hugely successful M1 processor.

In our view this seems to suggest that while Moore's Law gets harder and harder to keep up with (read that as more and more expensive) that much of the recent concern seemed to be around the adoption of EUV which was certainly gut wrenching for the industry but now that it is behind us we seem more confident of the cadence of the industry.

This bodes very well for TSMC versus both Intel and Samsung who have been having more growing pains with EUV and are clearly not out of the woods yet. We would imagine that ASML will be quite happy as lot of TSMC's increased spend will be on EUV, both the first and second generation.


Arizona obviously bricks and mortar for now

Part of the spend will be to build a fab in Arizona but it will be a while before equipment can even be ordered. This means that the percentage of capex that is for bricks and mortar is likely a little higher this year.

Both a big blessing and curse for equipment makers

The huge increase in equipment spend will unleash a tidal wave of spend with the equipment companies which is obviously great.
However this makes the industry even more dependent upon TSMC who is already a tough customer and likely even tougher than Intel was when Intel was dominant.

At this point, if you don't have either TSMC or Samsung as customers, you might as well fold the tent and go home. If you have both, things are great.

DejaVu of Samsung's Stupid Spending Spree?

We are always nervous when the industry gets into the "euphoria" phase of the cyclicality. This is the point where industry analysts start to convince industry management that the industry is no longer cyclical and this outsize spend will go on forever....just as the car goes over the cliff without skid marks.....

While we think its healthy to remain paranoid we think that there is a significant difference between Samsung's Spending blip of a few years ago and today's TSMC spending increase. Memory is obviously a commodity product which has substantially more cyclicality than foundry production. Memory easily gets over and under supplied and pricing is a second order derivative of over and under supply. While Samsung is dominant in memory there are other players that keep the game somewhat fair.

In the foundry business TSMC is way far away the leader in a non-commodity business. Its hard for fabless houses to switch horses in mid stream. So while keeping a watchful eye on demand, we don't think we'll see the horrible hangover that we had from Samsung's party.

The Stocks...

While this is obviously a hugely positive development for TSMC, especially in the longer run, it will also put pressure on earnings in the short term as TSMC is spending with a fire hose. Even with their dominant position and great demand and ability to control pricing, it will be hard for profits to keep up with the outsized spend.

This could somewhat dampen the stock price in the near term. For equipment makers, this is all happy days (assuming TSMC is your customer).
We would caution that much of the positive news has already been baked in given that the stocks have been on fire. This seems to be more of supporting evidence that the positive tone has been justified for the past months. We could see a less than anticipated positive tone as investors view it as already in the stocks.
 
According to Trendforce the decision has already been made:


I highly doubt UMC is getting any of this. I think I will wait for the Intel investor call before I dig more into it.
As big as Intel is, I think Bob Swan got Intel board of directors' blessings/approval/encouragement/demand to proceed such important outsourcing deal.

It's less likely a new Intel CEO will make 180 degree change and get the same board's blessings/approval/encouragement/demand.

If that's the case, then Intel's board is either crazy or stupid or both.
 
What I saw in the past week were:
1. intel was stumbling in moving to next step. Even changed the chief.
2. tsmc was ambitious and marching bravely with surprising increase in CapEx.
3. Samsung was muting and waiting high level decision.
4. All equipment vendors were cheering to have bright future expected.
5. China's impact was forgotten temporarily.
 
High confidence in 3NM

With TSMC talking about "risk production" of 3NM in the second half of 2021 it seems pretty clear that they haven't hit any known roadblocks or issues that would throw their Moore's Law march off kilter. This would seem to match pretty well to getting 3NM up to full speed in time for Apple 2022 Iphone launch as well as further refinements of the hugely successful M1 processor.

In our view this seems to suggest that while Moore's Law gets harder and harder to keep up with (read that as more and more expensive) that much of the recent concern seemed to be around the adoption of EUV which was certainly gut wrenching for the industry but now that it is behind us we seem more confident of the cadence of the industry.

This bodes very well for TSMC versus both Intel and Samsung who have been having more growing pains with EUV and are clearly not out of the woods yet. We would imagine that ASML will be quite happy as lot of TSMC's increased spend will be on EUV, both the first and second generation.

Per C. C. Wei: "We continue to improve the EUV's productivity because we are working closely with suppliers. And so far, we -- the improvement is obvious but still not up to our expectation yet. "
 
Per C. C. Wei: "We continue to improve the EUV's productivity because we are working closely with suppliers. And so far, we -- the improvement is obvious but still not up to our expectation yet. "
If we made a simple calculation based upon tsmc owns 50% tool sets and output 60% EUV wafers, we can see tsmc provide +50% output/tool comparing to the average of all others. It is not enough upon CC Wei's expectation. Besides, what this imply? interesting.
 
I think it means they want to maintain productivity at higher dose, due to stochastic (photon shot noise) issues:

EUV shot noise vs pitch.png

So either the source power has to go up, or they need more tools to compensate.
 
That's a pretty big clue as to what is coming since TSMC builds fabs based on wafer agreements. My guess is that Intel is coming, absolutely.

Iam not sure, if intel is honest they haven't decided if they outsource yet (also they want to wait on new CEO).
I think its more likely that AMD made a big order and also Apple increased quite a bit...
 
Iam not sure, if intel is honest they haven't decided if they outsource yet (also they want to wait on new CEO).
I think its more likely that AMD made a big order and also Apple increased quite a bit...
The number is too big for AMD and Apple increasing their orders. We are talking about a 50% jump in TSMC capex over 2020, and a 25-40% jump from expected capex just a few months ago. I'm sure some of it is Apple asking for more wafers for M1, and AMD trying to get more capacity to fill demand, but I think it's likely part (if not most) of it is Intel. Don't forget Intel already oursources a lot to TSMC, so it's not a matter of "to outsource or not" it's a matter of "which products and how much".
 
Last edited:
Announcing CAPEX and spending it are two different things, right? The extra CAPEX is Intel for sure. Tomorrow could be a big day for TSM but not so much for INTC.
 
Back
Top