When Donald Trump launched the trade war with China in 2019., it inevitably led to a tech one. At the time, we witnessed the introduction of an enormous set of sanctions against Chinese companies, but that wasn’t all. The main goal of Trump’s administration was to move as much chip production to the West as possible.
Following this plan, the US government asked Taiwanese TSMC, the largest chip foundry in the world to build a plant on US soil. The initial investment was a little above 10 billion US dollars, which was acceptable for TSMC itself when the construction in Arizona started in 2021.
It’s also worth noting that the initial idea for TSMC’s Arizona plant was to produce the most advanced semiconductor chips. Meanwhile, many problems have occurred. The first one, and probably the most important was the fact that the US and Taiwan don’t have the double taxation avoidance agreement. This means that TSMC was, and is still obligated to pay taxes both in Taiwan and the US.
Such a situation inevitably led to rising costs. In the next few years, the cost of constructing the plant rose to 40 billion USD. The additional problem was the fact that the US government refused to pay the requested subsidies to TSMC, even though they had initiated the construction of the plant in the first place. Worst of all, they haven’t allowed TSMC even to take advantage of tax credits.
Of course, TSMC, obviously under the political pressure of its own government, continued building the plant, adding more of its own money to this black hole.
When they finally thought that all the problems had been solved, a new one came up. There is literally not enough qualified workforce in the US to even complete the plant construction, and we can leave to the imagination what will happen when the production eventually takes place in the future.
The story continued when TSMC tried to overcome newly occurring issues by bringing their own workforce from Taiwan to complete the construction of the plant. To be precise, the workforce they were looking for was the one to finalize machinery implementation. When applying for work permits, the US workers’ unions came out of the blue, arguing against importing foreign workers at a time of rising unemployment in the US.
TSMC somehow overcame this problem, but the result was another delay in production start. As reported by Asia Times, this can’t happen before 2025., in the best-case scenario. The first chips that will come from the Arizona plant production line won’t be available in the market before 2026, but the worst part of this story is just about to come. Those chips will be produced in 5nm and possibly in 4nm nodes. In the year 2026. TSMC, (and Samsung Foundry) will already start production of 2-nanometer chips, but on their own soil.
As Hong Kong’s South China Morning Post and Korea Times report, TSMC is investing tens of billions of dollars in new plants on the Taiwan island, where the new plants will be located. These will be the ones that will produce 2nm chips in 2026. and 1,4nm in 2027. Do the math, and you’ll quickly come to the conclusion that TSMC’s Arizona plant will no longer matter, at least not in the manner it was supposed to when the decision to build it was made.
There’s a similar story about Samsung Electronics Foundry. Korea Times reports that Samsung is also building new plants in the Seoul area for 2nm and 1.4nm chips, leaving its US investment worth $17B in the cold.
Both TSMC’s and Samsung’s plants in the US will probably produce chips at some point in time, and even make a profit to some extent, but they definitely won’t fulfill the intention of the US government to move chip production to the West. What’s worse, according to many analysts, the chips produced in the US will be 30% more expensive than those made in the Asia region due to supply chain reasons. Namely, almost all supporting industries, and ecosystems are located in that region, so the move of US pressure to build factories in the West was doomed in the very beginning. That’s what happens when you’re led by political and not economic logic.