The TSMC First Half 2021 Mega-Review
If you want to watch the video, you can watch it below
I added a few new comments in the text below. I feel like I covered most things pretty well here, but if there is anything more you’d be curious to see a deeper dive on, let me know.
If you would like to support the channel, consider subscribing to the Patreon. Members of the Early Access tier get to watch videos first before they released to the public. I just recently released a nice video about Nvidia’s rise in the graphics card industry.
We are already more than halfway into the year. Time really flies, doesn't it?
It has been a while since my video deep dive on TSMC's $28 billion capital expenditure plan at the start of 2021. Some time has passed since then and so I think we are due for another review.
So in this review, we are going to look back at the first half of TSMC's 2021 year. This is going to be a collection of individual thoughts on the company, its environment, and more. We will review financials, the Arizona fab, the outbreak, water, and more.
Thus, at the risk of looking too obsessed with TSMC, let us dive in!
The Financial Results
Let us start with the financials. The important metrics that TSMC financial analysts look at are revenue, gross margins, capital expenditures, and of course, net profits.
TSMC net revenues are up 28% year over year, and 2% quarter over quarter. Company revenues are seasonal, with the first half usually being a little soft. With a new set of iPhones coming in the second half of the year, additional revenue growth should be expected.
Gross margins are down. In the 2Q 2021 quarter, they are 50%. In the first quarter, the company pulled 52% and in the 2Q of 2020, the company made 53% gross margin.
2% might not seem like a lot, but it is enough to cause the stock to pull back. The company says that this impact is due to a stronger Taiwan dollar and "5nm dilution".
In other words, 5nm has lower margins than more mature nodes like 7nm, and the company is expecting more customers to use this node in the future. Nevertheless they say they can raise those 5nm margins as they work out all the kinks.
With regards to capital expenditures, TSMC has thus far spent about $14.8 billion this year.
But they also decided to slightly raise their 2021 budget to $30 billion. I remember last year that each capital budget raise would cause the stock to respond positively. That such a thing did not happen this year I think says something about investor sentiment about TSMC's spending. There's some concern there.
Net profits in the second quarter are up 20% from the previous year, but down from the first quarter of 2021. Net profit margin is down, and this is likely due to the lower gross margins and higher operating expenses. The company is spending more on selling, general, and administration.
The Next Node
The company is continuing its relentless grind towards the next node. TSMC's next generation leading edge full node after 5nm, which went online last year, is the 3nm process. In June, the company reported that they are on track for high volume manufacturing of 3nm chips by the second half of 2022.
TSMC 3nm process will be using the FinFet technology, same as with the 5nm process. This is in contrast with Samsung Foundry, which is using Gate All Around. It is expected that the TSMC 2nm will use GAA.
The interesting thing is that 3nm is going to miss the annual Apple cycle that defined TSMC's node releases for the past few years. TSMC started making and selling 7nm and 5nm in the year’s second quarter, with large scale production coming shortly thereafter.
This had been necessary so that Apple - TSMC's single most important customer - can begin assembling iPhones for a September release.
But 3nm will miss this cadence. This is mentioned in the 2Q earnings call:
About 3 to 4 months ... as compared with 5-nanometer. Yes, 3-nanometer technology actually is very complicated and in both processing technology and also the customers' product design. So we work with a customer, and finally, we decided to ramp up in the second half of next year.
This jives with a recent report in the Nikkei that says that while Apple will eventually use the 3nm process, the first 3nm chips will go into the iPad Pros rather than the iPhones.
The last iPad Pro came out in the second quarter of the year, so the schedules match with the 3-4 month delay.
The presumed iPhone 14 family will use the 4nm process, a third iteration of the 5nm process. The iPhone 13, FYI - the one that will come this year - will use the N5P, the second half step improvement of the 5nm process now in the iPhone 12.
COVID in Taiwan
The biggest news story in Taiwan over the past several months has been the island's virus outbreak. A (relatively) large outbreak that began some time in May has spread throughout the Taiwanese populace. Starting mostly in the Taiwanese societal underbelly and erupting from there.
The outbreak has resulted in widespread soft lockdowns across Taiwan. The cafes are no longer open, schools have gone to distance learning, and companies have gone remote. What the rest of the world underwent throughout 2020, Taiwan is now finally experiencing.
As of this writing, over 14,000 local cases have been reported and 755 people have died.
The majority of the cases have occurred in the two largest cities - New Taipei City and Taipei. TSMC is based in Hsinchu, with major facilities in Hsinchu, Tainan and Taichung. For the most part, these cities have been spared the brunt of the outbreak.
TSMC has done a lot here for Taiwan's people. Together with Foxconn, the company bought 10 million vaccine doses from the German company BioNTech. The company then donated the vaccines to the government for distribution and use.
In addition to the TSMC and Foxconn purchases, donations from Japan and the United States have helped to fill the gap. The vaccination rollout is ongoing.
Regardless of whatever politics, the Taiwan government has done a commendable job in bringing the outbreak under control. And you can’t say that for every government out there. Relentless testing, contact tracing, and surprisingly strict quarantines have drastically brought the case level down.
And this is despite more people than ever on the streets and shops. With numbers continuing to stay low, the government is looking to slowly open things up again. It has consumed the daily news, but in terms of TSMC’s economic results it is a non-issue.
Author’s note: We are at single digits now and things are opening up again.
The Drought
As I write this, Taiwan is undergoing its worst drought in years. It might be weird to think about, but the island is highly dependent on typhoons and other large storms for water. Typhoons might cause their own damage, but they also fill the reservoirs.
The lack of rain has made the global news, because of the chip shortage as well as the fact that the semiconductor fabrication process is about as thirsty as a sixteen year old boy with an Instagram account.
For all those who read those articles at the time and clutched their pearls close to their chest, I would like to update you on the situation.
The plum rains rainy season or 梅雨, méiyǔ in Taiwan lasts for two months - usually May and June - and brings good rain.
2021's plum rains were quite excellent, bringing water to both the north and south.
I have the photographs, video, and wet socks to prove it.
As a result, many of Taiwan's critical reservoirs have refilled to a less pearl-clutchy level.
The Sun Moon Lake reservoir in central Taiwan is at 98% capacity. The Baoshan Reservoir No. 2 in Northern Taiwan, which supplies water to companies in Hsinchu, is now at 100% capacity. It had been at 2.8% in late May.
And all of the reservoirs that provide for southern Taiwan's water needs are all at 50% or higher capacity. Including the ultra-critical Zengwen Reservoir in Chiayi.
Author’s note: It is still raining and all the reservoirs are full now.
The rains are over for now, and of course there remain many problems with Taiwan's water supply infrastructure. The drought articles are likely to come back at the start of next year again as the reservoir levels drop low once more. But for now, I do think this issue is less of a concern.
The Arizona Fab
In June 2021, TSMC reported that they have begun construction on their Arizona fab. Equipment will begin moving into the clean rooms in the second half of 2022. They will join Samsung and Intel in adding capacity in the American continent. Right now, $12 billion is committed to the campus, and it will employ 1,600-1,900 new workers.
There are media reports that the fab will eventually cost 3 times that, and support six full factories. But the only official thing that TSMC is announcing is their Phase 1. In Phase 1, the company is targeting 20,000 5nm wafers by the first quarter of 2024.
The city of Phoenix estimates that the total economic impact is estimated to be $38 billion over the next 20 years.
> "This innovative growing company will bring one of the largest single foreign investments in Arizona history and provide thousands of quality jobs and will facilitate thousands of additional jobs to support its ecosystem and operation"
In December 2020, the company paid $89 million for 1.7 square miles or 1,129 acres of undeveloped land off Interstate 17. Another 450 acres will be rezoned for industrial use, to support suppliers working near TSMC.
For instance, LCY Chemical, a Taiwanese petrochemical company that makes chemicals for cleaning wafers, announced that they too will build an American factory to support the Arizona fab.
The Phoenix City Council not only sold the land in "auction" with no other bidders, but will also commit $200 million to build road, sidewalk, and water infrastructure to connect the land to the rest of the city.
The Arizona fab will require a lot of water. Arizona, as my commenters like to remind me, is a desert. I have done a video about the semiconductor water issue, and you can watch that if you want a deeper dive, but it is likely that the water allocation for the fab will come out of farmers and agriculture in the area. It sucks for the farmers, their families, and their livelihood but it is what it is.
Phoenix Mayor Kate Gallego and economic development director Chris Mackay had traveled to Taiwan in October 2020 to pitch TSMC and close the deal.
Chips do not spring willy nilly from wafers. They have to be turned into such. These services are referred to as the back-end. While there are reports of backend services facilities to be included in the Arizona fab location, that has not been confirmed as of yet. Right now, wafers made at Arizona fab will have to be shipped back to Taiwan to be packaged. Sounds a little weird, but it is normal for the industry.
In summation, this is nothing like the Foxconn vaporware that Terry and Trump dug up for a photo-op long ago. This is real and this is moving.
A Japan Fab?
In June 2021, the Nikkei Asia Times reported that TSMC is mulling a proposal for building a plant in the western Japanese area of Kumamoto. The plant is not leading edge. It would be working in the 28nm and 16nm specialty nodes - and would service large customers like Sony.
Author’s note: A later news item confirms the Sony involvement
Sony also has a plant in Kumamoto. The company is the global leader in image sensor chips, the stuff that goes into your mobile phone and iPad cameras - though I have no idea who would ever use an iPad to take pictures of things.
TSMC later confirmed in its 2Q 2021 call that a Japan fab is being discussed, subject to economic viability issues and other things.
TSMC has also agreed to build a $190 million R&D site near Tokyo. To first commit to an R&D site and then a plant, that is interesting.
TSMC has built plants before in other countries. They did a joint venture to create SSMC in Singapore. They also currently run WaferTech LLC out in the American northwest, and that started out as a joint venture too.
So maybe doing a joint venture fab with Sony, likely the biggest direct customer, would be the most logical outcome. TSMC gets to contribute something, but leaves the heavy lifting to the local partner. Sony has also recently started working on expanding their capacity in 2021, in Nagasaki. So a match made in heaven? We shall see.
A Europe Fab?
I recently came across some remarks by TSMC chairman Mark Liu. In his capacity as the head of the Taiwan Semiconductor Industry Association, he made some comments on the current chip shortage.
I cannot quite find the exact words as news reports keep paraphrasing what he said. But here is the closest of what I have been able to locate:
If the entire supply chain were moved to the US and Europe, or the US and Europe were bent on expanding their own fab capacity and technological prowess, it would create large amounts of "non-profitable capacity,"
That last part is kind of interesting to me since it is the only one put into quotes. Europe for whatever reason lags behind in terms of customer volume for the semiconductor industry.
The entire EMEA area accounts for less than 10% of TSMC's revenue - just barely more than Japan. And I have not seen mention of fabs in Europe in any of the recent transcripts. So I guess they are sticking to that.
Since then, authorities in Europe have brought in company representatives from Samsung, Intel, and TSMC to talk about the potential of a fab in Europe. Intel has a fab in Dublin, and it seems likely that they are going to expand it. But so far, it looks like TSMC and Samsung are not ready to build a European facility yet.
China
Something I noticed. TSMC revenue contribution from China grew from the first quarter, essentially doubling from the previous quarter. This is way down from the previous year, but that comparison period had Huawei revenue in it. China might be trying to build out native capacity and SMIC seems to be doing well right now, but TSMC is still winning share there. Likely from drawing on China’s burgeoning fabless industry.
The company announced that they will invest $2.87 billion to expand their Nanjing facility's 16nm and 28nm capacity. Their 28nm output in particular would be quite substantial, reaching 40,000 wafers a month by 2023.
This has caused some minor concern on the mainland, with one analyst Xiang Ligang calling for the Chinese government to stop TSMC from doing this expansion.
> "TSMC is such a sophisticated player with great power, and if it is in the arena, it can easily defeat all the China players ... It’s a very serious issue and I certainly hope that there will be a future for China’s domestic chip industry"
To me, these concerns seem unfounded. By itself, TSMC Nanjing is only the third largest foundry after SMIC and Grace Semiconductor. SMIC is far ahead of everyone else, with some $3.9 billion in revenue in 2020. To compare, Grace only had $961 million that same year.
And the Nanjing fab’s processes are several steps behind what is available on Taiwan. The expansion might help it win additional customers from those players, but SMIC and Grace are not going away.
The Competition
The semiconductor foundry space is heating up. There is more money pouring into the space than ever before.
Intel is moving forward with renewed emphasis on their Intel Foundry business division. The Wall Street Journal recently reported that they are interested in buying third or fourth place player GlobalFoundries. The mooted price would be $30 billion.
Intel generated $20 billion of free cash flow in 2020. They can afford to spend $30 billion.
Consolidating foundries seems like the natural response to counterbalancing TSMC's dominance in the space. It would be a good way to kickstart Intel Foundry and skip many years of work.
That being said, it is likely that a transaction like this is going to undergo pretty heavy regulatory scrutiny. I cannot help but think of the American telecom space. AT&T, one of the two large players, tried to buy T-Mobile and the US did not approve of the purchase.
Eventually they allowed the third and fourth largest players, T-Mobile and Sprint, to merge. I guess it depends on whether or not the regulators are going to see Intel in the independent foundry space as a lagging player or not.
This whole Intel thing is going to get super weird because it has been reported that Intel is also going to be a flagship 3nm customer alongside Apple. Intel is competing with AMD, Apple, and TSMC itself. Something about that ... not sure what to make of it.
And of course, let us not forget about Samsung. For all of Intel's recent huffery, the Korean giant is still TSMC's fiercest competitor.
The company has been able to deliver competitive nodes at 10nm and 7nm in terms of density. But the company has struggled with the 5nm and 3nm generations. The density metrics compare unfavorably with TSMC.
Furthermore, the yields at 5nm lag behind. Some sources at Korean news agencies cite yields at under 50%.
Their 3nm process, which utilizes the new Gate All-around technology, is looking to be delayed, moving to high volume production in 2023. As I mentioned, TSMC for their part is looking at the second half of 2022.
Conclusion
TSMC is living up to its promise to spend immense amounts of money and continues to hold substantial advantages over its competitors. Their next generation nodes have significant customer interest and, more importantly, are on schedule.
With that being said, the industry is heating up and Intel Foundry is showing signs of being a formidable competitor. TSMC is going to have to work even harder to maintain its edge. Economic concerns also hang over the industry and its rampant expansion as we go into 2022.
As of this writing, TSMC's market capitalization is around $600 billion. This corresponds to about 10th or 11th place, on the same tier with companies like Visa, Tesla and Alibaba. I don't know if TSMC should be valued on the same level as those companies. That is for you to decide.
For what it is worth, Morgan Stanley recently put out a research note making it clear that they think the stock is overvalued - saying that the stock is "potentially dead money in the next 12-18 months". They don't like the stock.
Regardless, it remains by far Taiwan's most valuable company. And it holds a strong lead over the rest of the industry.