Array
(
    [content] => 
    [params] => Array
        (
            [0] => /forum/index.php?threads/tsmc-earnings-solid-hpc-leads-growth.11428/
        )

    [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 Earnings solid, HPC leads growth

count

Well-known member
TSMC posted solid earnings, but what caught my attention was the eye popping 23% growth in HPC. This will be a bigger market for TSMC than mobile in a few years.
 
Here are the slides:


And here is the trabnscript:


12
 
Last edited:
Other key notes from the transcript are:

-N5 is in risk production and volume production first half of next year. So TSMC will be shipping 5nm in volume while Intel will still be in limited 10nm production.
-Utilization on 7nm is improving, which will help margins in the second half.
-They are expecting 5G to accelerate based on big players. "And in the 5-nanometer, the big players get into 5G smartphone very aggressively. And it doesn't take many tape-outs, just a few tape-outs. The demand out is there of much bigger capacity than any other products." I think this another confirmation of rumors that Apple will use a Qualcomm 5G modem in 2020 iPhones.
 
Per the slide on revenue per platform: 45% of revenue is smartphones (5% growth). Wow, the gift that keeps on giving! HPC is 32% (23% growth), IoT is 8%, (15% growth), Automotive is 5% (3% growth), DCE 6% (-3%), others 4% (-2%).

Per the slide on revenue per technology: 20nm 1%, 16nm is 23%, 10nm 3%, 7nm 21%.

Comments:

I was in Hisinchu this week and noticed a bounce in the step of both TSMC and UMC workers so I expect to hear some good news out of UMC as well. The lunch rooms were buzzing. I mentioned that to a friend and he said it means they are not at their desks working so I guess it could go either way.

Apple gets first pick at 5nm so I doubt very many others will have 5nm system level products on the shelves in 2020 other than Apple in the second half.

It will be interesting to see if TSMC breaks out 6nm or just lumps it in with 7nm like 6nm and 12nm. And I'm wondering if 5nm will keep in trend with 20nm and 10nm as half step nodes that only Apple ship in high volumes. Thoughts?

CC Wei's opening "don't trust Samsung" salvo:

Now let me talk about TSMC's competitiveness. The foundry business model has proven to be the most efficient model in the semiconductor industry. As a pure-play dedicated foundry, we collaborate and work closely together with our customers to unleash their innovation to the market and enable their success. We do not have any internal products, and we do not compete with customers.

Q&A

Asked about 2H 2019:

Dr. C. C. Wei
Actually, I don't want to specifically pinpoint one customer only. All I can say is the second half, the new smartphone launches, especially the premium grade, has been the seasonality phenomena for us, always the second half. And the acceleration of the 5G actually enhance this kind of increase. That's all I can say.

That is Apple and Huawei. If the trade problem with China goes away in the second half, which I believe it will, it will be a big upgrade cycle.

Asked about Samsung 5 and 3:

Dr. Mark Liu
Let me add that, actually, our 5-nanometer is a full-node stride from our 7-nanometer. And our 3-nanometer is another full node stride from our 5-nanometer. This is very different than our competitors' road map. So if you compare their numerical, three is probably closer to the five. Secondly, on the 5-nanometer capacity build -- the second question is 5-nanometer capacity.

Scott Jones was at SEMICON West so hopefully he can chime in here or include it in his next landscape blog. TSMC really is focused manufacturability (yield/cost) so 3nm GAA might be early for them.

Funniest response:

Bruce Lu
It seems to me that the microenvironment seems so dynamic, but we have -- as analyst, we have a lot of difficulties, right? We can't just follow the tweets. So can you help us how to do the -- how do you predict that kind of customer behavior in terms of the production planning moving forward?


Dr. Mark Liu
So don't follow the tweet. As the fabless inventory, I think Lora just reported is coming down.

I wish Morris Chang was there for that one. Tweet? Why do you listen to birds? That is stupid. Hahahahaha

Sebastian Hou
Okay. My second question is can TSMC talk about the possibility of building the fab or acquiring a fab company in the United States? And would this is -- would this be out of the geopolitical concerns? Or any other consideration?

Daniel Nenni: No, that way the US will be sure and protect Taiwan from China.

Lots of 5G discussion which I can summarize here:

11
 
"We expect our customers' end products using N7+ will be in the market in high volume this quarter. We expect strong demand to continue into next year." -Dr. C. C. Wei

TSMC said in the Jan 2019 call that 7+ would generate less than NT $1B, or ~$32 million USD in 2019 (EDIT nvm this obviously must've been a typo. An analyst said that number and C. C. Wei confirmed it, but I think he must've thought $1B USD). If it's now going into high volume production, does that mean that this node was pulled forward a couple of quarters ahead of its original schedule, and 6nm was added as a lower cost version 7nm? I'm curious as to what the customer split between 7+ and 6nm will be. I'm guessing customers who are willing to pay extra for an improvement over 7nm sooner than than waiting until 2021 for 6nm will be will pick 7+ as a stop-gap before 5nm. Apple and Huawai this year, next year AMD for Zen 3, maybe Nvidia too in 2H2020 if they release a 7nm card later this year.

I thought AMD might wait for Q4 2020 to launch Zen 3 on 6nm but according to the following article Milan is a mid-2020 product, so too soon for 6nm right? https://www.anandtech.com/show/14568/an-interview-with-amds-forrest-norrod-naples-rome-milan-genoa
 
Last edited:
TSMC posted solid earnings, but what caught my attention was the eye popping 23% growth in HPC. This will be a bigger market for TSMC than mobile in a few years.

HPC includes cryptocurrency which I do not see as a long term market such as mobile.
 
Other key notes from the transcript are:

-N5 is in risk production and volume production first half of next year. So TSMC will be shipping 5nm in volume while Intel will still be in limited 10nm production.
-Utilization on 7nm is improving, which will help margins in the second half.
-They are expecting 5G to accelerate based on big players. "And in the 5-nanometer, the big players get into 5G smartphone very aggressively. And it doesn't take many tape-outs, just a few tape-outs. The demand out is there of much bigger capacity than any other products." I think this another confirmation of rumors that Apple will use a Qualcomm 5G modem in 2020 iPhones.

Apple’s iPad schedule has always been all over the place, BUT they could repeat something they’ve done before and fab the A13X on 5nm, releasing new iPads in April 2020 or so...

IF 2020 is the year of the ever-mythical ARM Mac, that schedule might make even more sense, allowing the A13X on 5nm to test the waters not just for the 5nm A14, but also for the manufacturing of the ARM Mac chips/chiplets.

Point is, we may see 5nm in reasonable volumes as early as April or so next year, not have to wait till Sept.
 
Apple’s iPad schedule has always been all over the place, BUT they could repeat something they’ve done before and fab the A13X on 5nm, releasing new iPads in April 2020 or so...

IF 2020 is the year of the ever-mythical ARM Mac, that schedule might make even more sense, allowing the A13X on 5nm to test the waters not just for the 5nm A14, but also for the manufacturing of the ARM Mac chips/chiplets.

Point is, we may see 5nm in reasonable volumes as early as April or so next year, not have to wait till Sept.
5nm won't be ready for a product launch for April. TSMC started risk production in the second quarter of this year and it is a full node shrink so no less than a year for volume production and then few months more for an actual product launch. The earliest I expect 5nm is June as Apple did with 10nm in the iPad Pro in 2017. And I get the feeling that 5nm will get to volume production later than 10/7nm did.
 
5nm won't be ready for a product launch for April. TSMC started risk production in the second quarter of this year and it is a full node shrink so no less than a year for volume production and then few months more for an actual product launch. The earliest I expect 5nm is June as Apple did with 10nm in the iPad Pro in 2017. And I get the feeling that 5nm will get to volume production later than 10/7nm did.

From what I learned in Hsinchu last week 5nm is exceeding expectations so a 1H 2020 lunch is quite possible for low volume iPad production. Remember, Apple uses a customized version of 5nm which can be frozen for production at the request of Apple to better time it with product launches.
 
5nm won't be ready for a product launch for April. TSMC started risk production in the second quarter of this year and it is a full node shrink so no less than a year for volume production and then few months more for an actual product launch. The earliest I expect 5nm is June as Apple did with 10nm in the iPad Pro in 2017. And I get the feeling that 5nm will get to volume production later than 10/7nm did.

The schedule I suggested matches the 7nm schedule.
TSMC 7nm started risk production April 2017, and HVM April 2018.
TSMC announced 5nm risk production April 2019.

It's really just a question of how many chips you need to stockpile given the expected volume of sales right at the point of launch. iPhone needs massive stockpiles. But lower volume devices with less publicity...
 
Back
Top