The 5-Year Plan for Semis

Semiconductors are, oddly, on everyone’s minds lately. This mature, arcane industry that powers the whole global economy had been an obscure corner of technology un-scrutinized by most for years, has suddenly become the focus of a lot of attention. In particular, semis have become the target for US industrial policy and government action. We touched on this last year, but were prompted to write about it today after listening to this episode of the always solid Odd Lots podcast. The guests on the show work for an economic think and have recently been publishing an interesting series on the prospects for US Semis Industrial Policy. The series is not complete, but we felt the issues they raised on the podcast highlight the difficulty formulating such a policy.

In general, we ae not big fans of industrial policy. These kinds of things can all too easily go astray, squandering tax dollars and distorting markets. In particular, the US does not have a great history on this front, with past policies often leading to… let’s call it highly inefficient capital allocations. That being said, the US is now facing some serious concerns around semis. So before any one uses those concerns as an excuse to implement a useless or even harmless policy, we think the issue needs some exploration.

Our key problem with much of the talk of semis policy lately and of this episode of the podcast is that semis is a complex industry and many issues are being conflated into one giant problem. So the risk is that any proposed solution or policy addresses the wrong issue and likely exacerbates the other problems.

Right now there are really two issues facing the US semiconductor industry: the utter dependence on TSMC and Samsung (sortof) for advanced semis manufacturing; and the perceived decline of US manufacturing capabilities. These are the real problems, but then people often throw in the entirely transitory supply chain shortages and Intel’s entirely self-inflicted crippling and create a narrative of a decline in the US semis industry. The opposite is true. The US semiconductor industry is the best in the world, by far, but the details matter.

The truth is that the US does have a critical, strategic shortage of advanced semiconductor manufacturing capabilities. There is really only one company capable of producing chips with the leading edge process. That company is located 100 miles from a rising economic rival and potential military threat. The only other company even close to TSMC’s process, Samsung, is not too much further away. So much of the US economy and its military are dependent on these chips that the US has to develop alternatives.

A secondary concern is that many seem to think that the US has lost its ability to manufacture things. This is really only true for leading edge semis. The US can still produce plenty of chips using older processes (which is what most military applications really need). And this is to say nothing of other things we seem to manufacture pretty well – vaccines for one. True, the electronics manufacturing industry has largely moved to Asia, but there are some pretty valid commercial reasons for this. Electronics manufacturing (as opposed to semis manufacturing) is generally a low-margin, labor intensive, polluting industry.

You can tell when people confuse these two issues whenever they talk about TSMC and then start extoling the virtues of Shenzhen. Shenzhen has nothing to do with TSMC, it is just currently the best place to take chips from TSMC and solder them into circuit boards. Shenzhen is a fabulous place that does have a concentration that gives it a big advantage in electronics manufacture, but the US is not clearly any worse for lacking this center. Or put another way, the solution to the lack of a US equivalent to TSMC cannot be solved in the same way that one would approach the lack of a US equivalent to Shenzhen. Shenzhen’s position is by no means certain. It has already become too expensive for a lot of categories of production. And the history is not good, over the past 70 years, the center of electronics production has shifted form the US West Coast, to Japan, to Korea, then to Taiwan and only lately settled in South China. Electronics manufacture should not be a focus of US industrial policy, at least not for Strategic, National Security reasons.

By contrast, the lack of advanced US semis capacity could be a concern. But the trouble is that this has no easy solution. As we noted in the past, handing $10 billion or $20 billion to Intel to help them build a fab is no guarantee of a solution. Intel has plenty of money. Yes, TSMC is going to build a mini-fab in the Arizona desert, but that will offer limited capacity for the foreseeable future and the parent company/government is unlikely to let it grow too big for fear of losing their own strategic leverage. No easy answers here. And let’s not forget that any advanced plant would cost something like $20 billion, probably $40 billion if the government is in charge.

The podcast does make a valid point that the US does have a weakness when it comes to commercializing the output of its world-best academic complex. There is truth to that, we could do a much better job, but this same argument was made in the 1980’s when faced with the “threat” from a rising Japan. Yet none of the panic from back then played out the way the scaremongers feared. Japan’s electronics industry hardly matters any more (with a few exceptions). And maybe that is the point. Industrial policy can lead to bad outcomes, not just a waste of money but hobbling of an economy’s potential to move to more promising arenas. Europe is a great example of this, an economic bloc that is vastly underperforming in technology. And this should also be noted by those worried that China’s semis industrial policy will eventually mean an eclipse of the US’s. China does have some interesting policies in place, some of which might make sense here, but it is not guaranteed to succeed. After ten years, they have not caught up in semis manufacture, semicap equipment or design tools. So maybe industrial policy is harder than it looks….

The best path forward for the US is to focus on what it does best and to continue to out-innovate the world.

Leave a Reply