Is Anybody NOT Investing Gazillions in More Microchips?
TSMC retains dominance, if it can afford it.
Two significant cards were laid down lately in the Great Game for the future of semiconductors, the commanding height of the 21st-century economy. You can’t make computers, smartphones, cars, or most other post-industrial essentials without these minute electric circuits etched in silicon. A few companies dominate their manufacture, mostly in Asia.
China’s champion in the field, Semiconductor Manufacturing International Corp., or SMIC (ticker: 981. Hong Kong), lately announced it would invest close to $9 billion in a new Shanghai plant to produce chips with 28-nanometer “transistor gates.”
If that sounds like a lot of money, a week earlier Korean conglomerate Samsung (005930. Korea) set its 2030 investment target at $206 billion, a 30% increase from previous guidance. Three-quarters of that capital is supposed to pour into semiconductors, largely on the chaebol’s South Korean home turf. The strategic blueprint was unveiled days after Samsung’s controlling shareholder J.Y. Lee, got out of jail for the second time. (A long, only-in-Korea story that we won’t go into here.) So he had some time to think it through beforehand.
SMIC’s big move looks sensible and cool-headed amid much overheated geopolitical rhetoric. China wants to rid itself of foreign, specifically U.S., dependence for chips and chip-making equipment. Building 28-nm chips won’t exactly do that. To analogize from the old Space Race, it’s a bit like the USSR improving Sputnik while America was landing on the moon. Global industry leader Taiwan Semiconductor Manufacturing Co. (TSM) rolled out 28-nm 10 years ago. These days it’s down to 5-nm, and working on 3 and 2. The smaller the transistor gate, the more circuits and computing power can fit on a single chip.
But 28-nm is good enough for lots of applications. Lots and lots of applications. TSMC’s own web site offers this partial list: “Central Processing Units (CPUs), graphic processors (GPUs), high-speed networking chips, smart phones, application processors (APs), tablets, home entertainment, consumer electronics, automotive, and the Internet of Things.” China may be learning to take things a step at a time. Its last great semiconductor hope, Tsinghua Unigroup, promised global leadership and is headed for bankruptcy instead.
What stands out from Samsung’s investment plan, aside from sheer magnitude, is the Korea First orientation — striking from a company and country that have been avatars of globalization. Samsung isn’t worried about relying on America. The foreign dependency it wants to shake is Japan.
The two nations are core co-habitants these days beneath the U.S. security umbrella. But bad blood still simmers from Japan’s long, brutal occupation of Korea, 1910-1945. It boiled over most recently in 2018-19, when Korean courts approved seizure of shares in a Korean-Japanese joint venture to compensate four surviving ”comfort women” (i.e. sex slaves) for their suffering during World War II. Tokyo retaliated by removing South Korea from a free trade “white list.” That threatened vital imports of semiconductor equipment until politicians papered things over again.
Samsung, like SMIC, seems to be on the right track. The company is sitting on more than $100 billion in cash. Speculation swirled that a liberated Lee would splash out on a big acquisition to draw even with Taiwan Semi, for instance Netherlands-based NXP Semiconductors (NXPI). That risked running into a regulatory thicket, not to mention cultural conflicts as buttoned-down chaebol managers tried to wrangle authority-averse Western techies. Instead, Samsung is seeding a bunch of smaller domestic firms with names like Soulbrain and KCTech, aiming to build supply chains within its own borders.
These initiatives from its competitors spell both triumph and threat for TSMC, the reigning heavyweight semiconductor champion. Triumph because it remains pretty much unchallenged in the race for ever-more tiny and efficient circuitry. Threat because SMIC and others could bite off chunks of the reliable cash flow from older chips, and staying ahead by reducing ever-more nanometers is damned expensive.
Cutting-edge 5-nm chips cost 10 times as much as the 28-nm workhorse for “design and manufacture,” according to industry guru IBS, and 3-nm may cost 50% more than 5-nm. The performance improvement — TSMC estimates 10%-15% more computing power from 3-nm compared to 5-nm, and 30%-35% less power consumption — may or may not be worth it. “Even if [chip manufacturers] can find a way to build smaller transistors, their heroic efforts will ultimately become too expensive for their customers to afford,” Silicon Valley authority Linley Gwenapp wrote recently in Forbes.
The whole deglobalizing, national sovereign thing that Samsung is embracing is also bad news for TSMC. Economically, it makes sense for the company to supply the world from its home base in Taiwan, where it has molded generations of engineers and suppliers into an awesome technical cluster. That doesn’t look like it will work anymore, though. If Korea is pushing for a self-contained semiconductor network within its borders, so are the U.S., Europe and China. TSMC, and Samsung, have little choice but to oblige them, at substantial cost. The two between them have already committed to spending $29 billion on new U.S. plants driven by political, not business logic.
Then there’s the banal worry that chipmakers will overproduce during the current global shortage, turning it into a glut sometime next year. It’s happened before. No wonder TSMC’s stock has treaded water most of this year despite record demand for its products. The future looks … complicated.
Speaking of the space race, NASA spent $280 billion in today’s dollars getting to the moon in the 1960s, the Planetary Society calculates. TSMC, Samsung and SMIC should go way past that during the 2020s semiconductor race. NASA didn’t need to turn a profit, though.