What are US Policy Goals for Semis?

We are a bit confused. What exactly is the US government’s goal for semis? We are not alone in this confusion. Many we speak with in the industry puzzle over this. For that matter, many of the people we speak with in policy circles and government in DC seem to wonder about this too.

To be fair, government can be seen as a giant machine for working out compromises and allocating scarce things. Nonetheless, we were speaking with a senior congressional staffer last month, and when he asked us what we thought the goal of US policy was, we realized that the array of choices is getting too wide.

Part of the confusion rests in the fact that the US government has not conducted explicit industrial policy for decades, at least not in technology. There is plenty of implicit industrial policy, largely a collection of inertia and interest politics, but the process of deliberately setting out a unified goal for a particular industry is not a well developed muscle within the US government.

And of course since this is politics, the reasons officials give for implementing these polices tend to resolve to a handful of overly-broad slogans – Jobs! National Security! Lower the Trade Deficit! While all those are valid goals, we should not confuse them with specific policy objectives. Instead, we tend to see those things as ways to balance the trade-offs of specific policies.

From what we have seen those policy goals tend to fall into one of five different categories:

  • Create self-sufficiency in American semis manufacturing
  • Grow the US semiconductor industry
  • Grow the US semiconductor industry’s manufacturing capabilities
  • Grow the US semiconductor industry’s advanced manufacturing capabilities
  • Advance US innovation

The problems start to emerge from the start. Is the goal to advance the whole US semis industry or is the goal to create self-sufficiency in manufacturing? As it stands today, the US semiconductor industry leads the world with most of the top 20 largest semis companies already based here. The government can enact all kinds of policies to encourage those companies to manufacture more onshore but that will likely mean an increase in production costs, which hurts the industry. TSMC, benefitting from a suppressed currency, has a 20%-30% cost advantage over US production. So just saying the government should advance US semis makes little sense. This seems to be a goal we have already achieved.

Another critical question is what kind of semis manufacturing capacity does the US need? Does it need to have leading edge process capacity or does it need raw, total capacity including trailing edge processes? Politically, it will be much easier to spread any subsidy dollars to a large number of geographically scattered trailing edge plants, but if the goal is to secure onshore advanced manufacturing then we can just write checks to Intel and TSMC, who are both building fabs in Arizona. Maybe throw some money at Samsung too for a plant in Texas. Those are very different outcomes politically.

Further complicating all of this is the role of companies and governments outside the US. One of the great geo-political advantages the US enjoys is its strong roster of allied countries. Pushing too hard on the domestic capacity angle will come at the expense of those allies, many of whom also have ambitions to build domestic fab capacity. One potential outcome of all the latest government initiatives is a massive amount of capacity additions at the trailing edge, just at a time when the giant wave of China capacity is coming on stream. Everybody wants to build semiconductor fabs, but how many of them are looking closely at just what types of capacity they are building? This is one of the big dangers of industrial policy, details matter, but they can get buried by political priorities.

Which leads to our chief concern with all of this – is the US’s industrial policy too backwards looking? The Soviet Union seems to have spent the 20th century building the world’s best 19th century economy – railroads, steel mills and hydro-electric plants. Are of all of today’s semis policies going to end up building the wrong things? We have argued that instead of trying to recapture a lead in old technologies, we should focus on what comes next. There is a big risk of the US aiming for the wrong target, albeit China seems to be making the same mistake. Of course, this approach. is much riskier – it requires betting on unknowns some of which could fail. And the failure of a government backed company makes for the worst kind of political headlines, a stain that lingers on political resumes, albeit not as bad as China where that kind. of failure leads to jail sentences.

Of course, if you ask anyone in China what the goal of US policies are you get a very different answer. There the policies tend to be viewed as punitive, an attempt by the US government to throttle China’s growing technology industry, if not its entire economy. We do not think this is the intended goal, but we also concede that there are definitely people in DC who would not mind if it was.

As outside observers, free of political constraints, we think US policy would probably be best doing less. The US government cannot will advanced fabs into existence, absent some acute national security crisis, but it does not need to. A little bit of help and encouragement (spelled dollars) will likely go a long way. A more active, intricate policy is both beyond the government’s current capabilities and likely to end up as counter productive. Intel and TSMC are building their fabs. True, the US lags behind China in trailing edge capacity, but Texas Instruments, ADI and Global Foundries seem to be doing a pretty good job supporting commercial demand. So instead of focusing too much on the past, we believe fairly strongly that the best path is to look forward by making it easier to develop new technologies, commercialize domestic research domestically and easing the burdens around new semis company formation. That does not make for an easy slogan (“Innovation!”) but it does look like the best path forward.

Leave a Reply