Are US Semis too expensive?

A few weeks back Morris Chang, founder of TSMC and industry legend, claimed that US’ efforts to onshore semiconductor manufacture were essentially doomed to failure. In particular he claimed that producing semis in the US cost 50% more than producing them in Taiwan. Chang pointed to a series of factors that caused this disparity, and in the past he is questioned the quality of the US workforce. He asserted that the US should focus on what it does well in semis – designing chips – an area where he acknowledged the US is better than anyone else. But he bemoaned the futility of US efforts to manufacture onshore.

His comments carry weight. His efforts at TSMC created the fabless semis model that is now used by almost everyone, and unleashed massive growth in the industry. According to legend, he tried to convince his employer, Texas Instruments, to create a foundry model, but TI rebuffed and ignored him, so he moved home to Taiwan to found TSMC. This is a man who deserves the moniker of visionary, because history proved him right as he built TSMC into the powerhouse it is today.

Is he right this time?

The answer is yes. And no.

He made these comments at a conference hosted by DC Think Tanks, where he argued that current US Government efforts to subsidize the chip industry are essentially wasteful and the sums being discussed are insufficient to accomplish their stated goal. We have pointed this out in the past, the US Chips Act proposes something like $50 billion over five years, by contrast TSMC is going to spend $44 billion in capex in 2022 alone. Like many, we worry that efforts like the Chips Act are essentially free money for a handful of companies that probably do not need hand outs.

That being said, there is a lot more to unpack from Chang’s comments. First, it is important to keep in mind that TSMC has been put under immense pressure to build a plant in the US. Chang is likely still the largest individual shareholder of TSMC, owning about 0.5% of the company. As a rational shareholder, he has reasons to object to how the company allocates its capital. In his comments, he pointed out that he was not involved in the decision to open a TSMC plant in Arizona.

Another sentiment, which we hear echoed by others in Taiwan, holds that the US cares a lot about the future of Taiwan in part because of the US economy’s reliance on TSMC. Will that interest be as strong, will the US commitment to Taiwan remain, once there is an advanced TSMC fab in the US? For this reason, we strongly suspect that TSMC will be very slow about adding capacity to that plant, which even when it opens in 2024 will already be at least one process node behind the mothership’s fabs in Taiwan.

But we mostly object to Chang’s characterization of the root of this problem. Every time a company from one country opens a factory in another, the home country managers lament the poor work methods of the new country’s workers. Such complaints are almost always caused by poor communication and poor understanding on both sides. That seems to be the case here. The US is absolutely capable of supplying qualified labor to produce leading edge semis. Despite its current travails, Intel has a fully capable fab workforce.

It is also important to remember that most economists would argue that Taiwan’s currency is greatly undervalued relative to the US dollar. The best author on this subject was Brad Setser who used to publish on this regularly, until he took a position with the Biden Administration. Many industry professionals had hoped that his presence at the US Trade Representative’s Office might usher in a new approach to Taiwan’s currency. Those hopes were dashed earlier this year when the USTR chose not to designate Taiwan a currency manipulator, a decision which speaks to the importance of geopolitics over individual industry preferences.

To really understand this, we turn to the only currency calculator that matters – the Economist’s Big Mac Index. According to its latest results, the New Taiwan Dollar is 52% undervalued relative to the US dollar. A figure which is suspiciously close to that 50% figure Chang cites. Of course, currencies move around a lot, so maybe that 52% value is just an artefact of the odd times we live in. But no, that undervaluation persists as far back as the Economist’s website shows data. The gap has actually widened from 9% in 2000. And not to put too fine a point on this, South Korea’s currency is currently 34% undervalued relative to the US dollar, which may speak to Samsung’s number two position in semis manufacture. Another great source for this can be found in this Odd Lots podcast episode which details the contortions of Taiwan’s central bank in manipulating maintaining its currency.

TSMC has benefited from this currency mismatch for essentially its entire history, and so it is reasonable to think that these benefits have compounded. All of this is to say that currency has been an important part of TSMC’s rise to dominance. We do not want to disparage the immense talent and knowledge that TSMC has, they truly are the best in the world, it’s just that this source of advantage does not rest in vague notions of the appropriateness of the Taiwanese workforce or other hazy notions.

Can the US overcome this? Again, the answer is mixed. If the US government and the semis industry choose to optimize for this, then absolutely, the US can bring leading edge semi manufacturing back. However, the ultimate cost of this is far larger than anyone in policy circles currently envisions, and thus seems unlikely to change any time soon.

6 responses to “Are US Semis too expensive?

  1. Morris Chang did not move “home” to set up TSMC. He was born in China, studied and worked in the US, and had never live in Taiwan prior to founding TSMC.

  2. There is a small typo: “but TSMC rebuffed and ignored him”

    Should be Texas not TSMC here

  3. I like your other articles but this one is wrong. Exchange rate has little impact to the cost of TSMC. Almost all TSMC CAPEX are USD cost. Labor cost is a rather small percentage of Foundries. Engineer cost between TSMC and Intel is rather small. Direct labor cost of Intel is probably 3 times more than TSMC direct labor. But it’s not important to the whole company and have very little to do with exchange rate. It’s the income level of Taiwan lower than that of USA.

    • Everybody in the industry has access to the same equipment priced in USD, but I would argue that TSMC’s expertise rests largely with its human capital – who are paid in NT$. They can pay their employees well, invest heavily in developing them, subsidize local academic research – all paid in NT$. Put another way, you say the “income level of Taiwan lower than that of USA” – which is largely possible because of the exchange rate.

Leave a Reply