The Acquired podcast on TSMC should be listened to, or the transcript read.
One thing I learned is that TSMC is a sort of public/private partnership with the Taiwanese government. They put up 50% of the initial capital.
> David: 50% by the government and the other 50% were owned by the investors. Morris got nothing.
> Ben: And just got to keep his salary.
> David: He was a government employee.
> Ben: Wow.
> David: There by the grace of the government.
> Ben: Oh my God.
> David: Isn't that unbelievable? This is so the opposite of Silicon Valley.
> Ben: How is he worth $3 billion today?
> David: Well, what he did—as TSMC started to work—he basically put all of his money into buying. He bought his own shares in the company. I don't know if it was privately. They went public on the Taiwan Stock Exchange in 1994, and then the New York Stock Exchange in 1997. But yeah, he put basically all of his excess cash flow into buying TSMC shares.
Thank you for suggesting this. One of the most interesting podcast episodes I've ever listened to. The background of Chang/TSMC is fascinating. Lots of lessons to be learned regardless of industry. While Chang is a technical genius, my main takeaway was that his smartest move might have been seeing the overall strategic picture / value chain of the industry and seeing the long term opportunity. A good reminder that it's always critical to deeply understand your current (or potential) customers and their needs and problems.
The most surprising thing was that during the late 90s, the government sold the absolute majority of its share. They really didn't think it was (going to be) an excellent cooperation.
Good point! Taiwan is a political anomaly, an accidental offshoot that came about as a result of the rise of the CCP in China (people moved to Taiwan for refuge), and Taiwan has become enormously successful in an organic, unplanned way.
This is probably the most commonly repeated myth about a potential westpac war, that the PRC wants to get TSMC and the US wants to preserve access to it, and that's why there's an incentive for war and why there's this constant escalation between the US and China. I don't know why it's so popular among people who don't know the subject matter, but it is.
If westpac goes spicy, TSMC as it exists today is toast. The US will bomb the shit out of it if they can't keep it, and probably target any personnel they can't extract from the island; there was recently a big kerfuffle in TW domestic politics over someone in the US saying so publicly. The PRC wants chip independence, but everyone involved in that discussion understands that if the PRC waved a magic wand and gained complete control of TSMC, it doesn't give them ASML, it doesn't give them Zeiss, and a lot of that technology and process won't necessarily be usable until there are domestic equivalents to those technologies, by which time it would probably be out of date. Reverse engineering manufacturing methods from a seized sample is incredibly, unbelievably difficult in the best of times, let alone for atomic precision mirrors and high-throughput miniature factories operated under high vacuum. More importantly, the PRC just isn't willing to go to war over chip independence; their bar for war is a lot higher than that.
The reason the PRC wants actual control of Taiwan in 2023 and could credibly go to war to achieve it is the same reason they wanted actual control of the island from the years 1949-2023 inclusive, the same reason that the Taiwan government wanted actual control of the PRC mainland from the years 1949-1990: because they view the territory as rightfully theirs, taken from them by people who they hated, and perceived that there was a decent chance they could get it by force. PRC control over Taiwan would be the final chapter in the civil war that quite literally birthed the PRC - it has immense political and national significance and has been essentially a major goal and organising principle of the PLA and the PRC as a nation ever since the KMT retreated to Taiwan. It is not about trying to get access to new GPUs.
ASML has 5,000 direct and indirect suppliers, some of these have no alternatives or competitors, and ASML is just one of TSMC's core suppliers. Advancing the cutting edge in processor design is a byproduct of global peace and stability at this point.
100%. I was not aware of any initiative to deny China access to TSMC products and services. Can you name some? NVidia is a different story as it is not owned by TSMC.
I do not follow this stuff, just a quick DDG search:
"The Trump administration’s campaign against Chinese telecommunications giant Huawei Technologies included strict restrictions on semiconductor sales to the company; those sanctions halved TSMC’s exports to China to 10 percent of its global sales. The company is now under increasing pressure to police the orders it receives from other Chinese companies, and as the Biden administration tightens export controls, the Taiwanese company has canceled more Chinese contracts."
Qianer Liu, Ryan McMorrow, Nian Liu, and Kathrin Hille, “Chinese Chip Designers Slow Down Processors to Dodge US Sanctions,” Financial Times, November 6, 2022; “Chinese Chip Startup Shows Key Gap in Biden Export Curbs,” Bloomberg, October 21, 2022, https://www.bloomberg.com/news/articles/2022-10-21/one-chine....
When America evacuated the nationalists and placed their Pacific fleet in front of Taiwan Mao asked the Russians for help. He could defeat anyone on land but he didn't have a navy to get his troops across.
The Russians politely declined because it wasn't worth the risk of WW3 for them.
I think a really important point you make is that many people (especially techies) tend to overestimate "rational" reasons for war, and they tend to forget we're all basically monkeys wearing pants.
Just look at the war in Ukraine. Even if Russia completely "wins" the war, what do they "win"? A country where ~60-70% of the residents despise them? Some extra land (like Russia really has a problem with not enough land)? It's just a (very, very sad and tragic) dick-waving contest. A PRC attack on Taiwan would basically be the same thing.
> The US will bomb the shit out of it if they can't keep it
IMO no, this is where the US will bomb TSMC analysis falls short. Including US Army War College analysis that suggest bombing TSMC that started drama. Bluntly TLDR is during wartime, TSMC is a card PRC has, not US.
Incentive wise, it's in overwhelming US interest to keep TSMC tap open. If anything PRC has more incentive to crater TSMC to deny US access to significant % of high end chips that's already denied to PRC - US CHIPS reshoring only projected to capture like 10% of market share in next 10 years, and TBH TW will still retain many sole source semi supply chains that it's questionable if US high end fabs can produce without access to TW supply chain in short/medium term. Overall, US hi tech dependence and IP captures vast % of TW's semi value. Losing access to TW semi hurts US the most. Meanwhile, nothing in US strategic writing indicates attempt to completely rebuild indigenous semi supply chain, something only PRC is pursuing, because none of the CHIPS4 partner would be onboard if they knew US semi strategy is essentially to replace them long term. PRC semi independence is a technical problem, US semi independance is a political / security problem that could potentially undermine it's entire East Asian security posture.
Short/medium term keeping TSMC around is in almost everyones interest. PRC wouldn't mind control but it will still be dependant on western supply chains, US wants continued access because vast segments of their strategic industries depend on it, TW wants fabs around because however war ends, they'll want to retain profitable strategic industry to keep talent and help with recovery. Hence talking point that TW would rig their own fabs to explode even more stupid. Which all points to leaving TSMC alone, if not using it as moderating force in conflict, or more likely something PRC can leverage against US. The reality is US has leverage over export controls / sanctions during peace time, PRC has leverage over cratering US access to TW (really east asian) semi supply chain during war. If indopac goes spicy, one of PRC's trump card is basically resetting most of US/western semi by decades, which pending on how PRC semi industrial policy fares, could actually close the semi gap.
If conflict escalates to TSMC being gone, expect Samsung Semi and Tokyo Electron and most other western semi vendors in East Asia to be gone to. Expect CONUS fabs to stop operating unless US successfully reshores the entire semi supply chain on CONUS, which US partners will fight tooth and nail to stop. And even then there's escalation tier of homeland strikes in a few years that's going constrain US ability to intervene even more. While PRC wants TW for historic sake, preserving/destroying TW (and East Asia) semi decision matrix is going to play major factor in influencing scope of conflict. Which will likely boil down to US conceding TW to PRC without much of fight while PRC concedes to US strengthened security presence in East Asia maintained by unholy cooperation to keep semi spice flowing. Or total war that that covers mainland PRC and US partners in East Asia and CONUS.
This is nothing but propaganda. Machiavelli should be a good reminder to everyone putting logical reasoning aside and falling for this BS. The PRC cannot be allowed to have a monopoly on semiconductor production as they will use it to blackmail everbody else.
> TSMC pumps out close to 60 per cent of the semiconductor chips used around the world and makes 90 per cent of the most advanced technology used in phones and laptops all the way to missile systems.
I knew TSMC was dominant but not that dominant. Couple that with ASML in the Netherlands which makes the chip making machines, and I'm wondering if there has been any time in history when so much of our critical technology in our world has been concentrated in so few producers.
The tech industry tends to fixate on the cutting edge. Global Foundries has process tech that is 10 years out of date (14nm). 14nm is fine, a lot of good chips were produced 10 years ago. If we optimistically assume that there are usually a few players in the market that are 10 years behind, that means many sellers will be at 3nm soon.
TSMC is dominant, and we could imagine that in a wartime scenario maybe that'd matter (big maybe, why would a military need mass production of chips that didn't exist a few years ago?). But for anything else, capitalism will happen and 60% market share is impressive but not concerning. The current strategy of trying to restrict China's access to chips might easily spur some competition.
> I'm wondering if there has been any time in history when so much of our critical technology in our world has been concentrated in so few producers.
Nuclear bombs? And I'd assume this happens all the time. Any cutting edge market, in fact, seems like it should often start with one big player that has a better understanding of new research that is then slowly worn down by competition as the knowledge spreads.
I'm still using an intel i7-4790 (22nm process) on my primary small form factor desktop and it's perfectly fine for day-to-day computing. It can even run most games just fine after I stick the fastest slot-powered gpu I can find (RTX A2000).
To make these EUV mirrors, each 2 to 4 nanometer layer of molybdenum and silicon has to be deposited one by one. Fifty times, without fatal defects.
After that, each mirror has to be painstakingly polished. The acceptable surface deviation metric is 50 picometers, or 50 trillionths of a meter. That is a staggering number.
To put that into perspective. 50 picometers on a mirror that is 450 millimeters wide. If that 450 millimeter mirror was blown up to the size of the United States, 4,500 kilometers or 2,800 miles wide, then 50 picometers would be just 0.5 mm tall.
Altho there's a couple ways to define an atomic radius, molybdenum's and silicon's empirically measured covalent atomic radii are 145 and 110 picometers respectively [1]. So 50pm is a very small bump indeed.
ASML is the only provider of EUV lithographic machines. TSMC is just slightly ahead of Samsung and Intel at utilizing that technology. I honestly can't see how anymore puts TSMC above ASML.
ASML is concentrated in a country that has been politically stable and uncontested for 80 years and doesn't seek the media spotlight.
It is easy to overlook. Nobody really pays much attention to the Netherlands or it's non existing foreign ambitions and there is no state involvement in ASML.
Basically no drama means no journalism.
Yeah this story doesn't make much sense. Intel and Samsung are only a few years behind TSMC, recreating ASML and their sole suppliers even within 10 years would be quite a stretch.
If TSMC is the most advanced and other players are 2-3 years behind, by definition 100% of the most advanced would be produced by TSMC. And I think they are only talking about CPU or microcontroller, not all silicon.
Here is an interview with a former vice president of research at TSMC (and later SMIC).
I encourage you to read the interview in its entirety. It is illuminating.
On the question of of bringing up a new process node:
"We all take two years to develop one generation, how come you guys can do it in one or one-and-a-half year?" And they asked if some of your customer transfer technology to you or what not? And I told him, "No," I told him that, "That's not true." I think he probably implied we steal technology from customer, the way he talk.
And I say, "I'll tell you why." I said that, "When we develop one node, basically you have some learning cycles. First, you do some simulation. And you have some idea, then you run wafers to prove that. So, you run a group of wafers according to simulation and you have some splits. The wafer runs through the fab, they come out and you measure them, you analyze them, and you try to improve and you run this again. This again, you run. So, this is learning cycle." At that time, "It takes about six learning cycle, roughly, to complete one generation." Of course, you had some short loops and not just one. I said that, "My R&D wafer in the fab run much faster than yours, because my R&D engineer works three shifts and you only work one shift. So, your R&D wafer move eight hours a day, my work/move 24-hours a day. So, my wafers go three times faster, even if you are twice smarter than me, I still beat you up." <laughter>
His engineers work three shifts. They aren't (necessarily) working longer hours. But they're willing to work second and third shift at times, which R&D engineers in bascially every other first-world country refuse to do.
If you had a bunch of R&D engineers willing to work 16-hour days every day of the week, as long as they could sleep from midnight to 8am, you still wouldn't be able to keep up with TSMC.
Considering how much fabs cost, or so all publications say, it's actually surprising for any R&D departments to not run three shifts 24/7 to maximize utilization. When 1 EUV machine setup and ready is a quarter billion dollars...
> But they're willing to work second and third shift at times, which R&D engineers in bascially every other first-world country refuse to do.
What do you mean by "refuse"? If the price is right plenty of people would agree. Although I suspect the value I would bring during night shifts would be net negative...
There’s a point at which money is no longer the deciding factor in people’s decision making. The amount of money required to make it a deciding factor again is uneconomical, and is of the “I’ll do this for a few years, then retire” variety, which is a net negative due to brain drain. So, no, just paying people more isn’t the answer.
From what I heard salaries of HW engineers are not anywhere close to FAANG SW engineers. I don't think paying double for night shifts would put them anywhere close to “I’ll do this for a few years, then retire”.
A few decades ago, I have worked for some years in an European semiconductor fab. There the R&D engineers also worked in 3 shifts, to follow experimental wafer batches all day.
However, most of the time the R&D engineers worked in the normal shift. By rotation, a few worked in the evening shift and even fewer worked in the night shift, typically only when it was expected that after finishing some process step it will be necessary to measure test structures on the experimental silicon wafers and make decisions about process parameters for the following process steps.
So work at least partially in shifts is really expected in such an environment, because it is inefficient and costly to stop the processing of a wafer batch and wait until next morning, when an R&D engineer would come and would decide what to do with the wafers next.
Funny thing I learned is that Taiwan actually needs Typhoons every year.
The west coast of the island has many huge water dams (for instance, Baoshan dam), that are critical to sustain life and industry throughout the year.
Each typhoon (tropical cyclones are called "typhoons" in West Pacific, and "hurricanes" in the US) that passes in the area brings about torrential rains, thus refilling the dams. And a year without typhoon means the freshwater reserves will go dry.
They have droughts every 3-5 years near Taipei and during those droughts they regularly consider shuttering parts of Hsinchu Science Park where many of the largest fabs are are major users of water. They likely wouldn't turn off the newest fabs and might just run at lower capacity, but it's competing with farming and drinking water allocation. The dams don't hold that much capacity.
If you were surprised about that, note that Samsung is responsible for 20% of the South Korean GDP… (and Samsung Electronics taking up most of the revenue)
In Taiwan it's one of the controversial things that comes up every year when there are electricity shortages or water shortages; whether TSMC should get special treatment, or not.
Most places have occasional electricity shortages when some power plant has some short term issue, but have a large connected grid that let's them get power from elsewhere.
Ignoring all political difficulties that Taiwan might have doing that, they have a huge population on a tiny island 100 miles from anything else.
Importantly very little water is actually consumed by the process of making semiconductors compared to how much water goes in, almost all of the water can be recycled or released back into the water cycle.
> Oddly the US state winning the beauty contest has .. A water shortage!
Just amazing.
But on
> Water shortages in Taiwan. That impacts production
that I now re-read: to the best of my understanding, water is not simply an impacter on production: it is critical for the survival of the factories, its machinery will break in its absence. It is brought to the factories through trucks, not just the rivers.
I think there's a lot of truth to this, but from that angle, TSMC really isn't all that important.
Much easier to strong-arm/guilt nvidia into licensing software and designs to intel to fab than it is to go to war with china over taiwan/tsmc for example.
Presumably, whichever side was losing would destroy it, it they were able. I wouldn't be surprised if the US also has a plan to evacuate skilled staff.
There's an actual measure of how important companies are and how many want control of it and how much they want, and TSMC is #12: https://companiesmarketcap.com/
But recoverable quite easily. TSMC? No. Massive shortages. iCloud going bye bye sucks, but you can cut your losses and move. Samsung can't just triple capacity over night, and thus the entire world crashes and burns for a while.
For the first month or two. Then we start running out of the good chips.
Though I'd say a clearer scenario is them disappearing with a couple months of warning. In that case the impact of Apple or Google disappearing would be greatly lessened, while TSMC's impact would be just as enormous.
Did Buffet actually dump his stake as the article says? He must be completely convinced it will be seized because that seems like a silly investment to walk away from.
> He must be completely convinced it will be seized
If you've listened to Buffett talk about investing, probably not. He got to be the world's greatest investor by understanding risk and probabilities very well. He doesn't need to be "completely convinced", just convinced that the expected value of his shares won't be positive. Suppose he thinks there is, say, 25% chance of a takeover (which is definitely scary but far from 100%) - in that case his shares go to 0. He could easily think that the potential upside just doesn't outweigh that significant chance of a total wipeout.
It can be something more than TSMC being seized. It could be that he sees TSMCs share of the market going down over the long term. Which it probably will. Many governments are pretty unhappy with dominance of TSMC. The US in particular is dumping some money out there to onshore chip manufacturing. It reasons that this will probably eventually result in some shifting in the market.
One thing I learned is that TSMC is a sort of public/private partnership with the Taiwanese government. They put up 50% of the initial capital.
> David: 50% by the government and the other 50% were owned by the investors. Morris got nothing.
> Ben: And just got to keep his salary.
> David: He was a government employee.
> Ben: Wow.
> David: There by the grace of the government.
> Ben: Oh my God.
> David: Isn't that unbelievable? This is so the opposite of Silicon Valley.
> Ben: How is he worth $3 billion today?
> David: Well, what he did—as TSMC started to work—he basically put all of his money into buying. He bought his own shares in the company. I don't know if it was privately. They went public on the Taiwan Stock Exchange in 1994, and then the New York Stock Exchange in 1997. But yeah, he put basically all of his excess cash flow into buying TSMC shares.
https://www.acquired.fm/episodes/tsmc