37 Comments

I'm going to give the benefit of the doubt that this analysis is largely impartial and accurate, but overall 'writing by a friend who has a financial interest in the industry he claims to be analysing' sets off a lot of red flags. I'm guessing when the company comes out of stealth mode the 'four recommendations' will align neatly with its profitability.

Expand full comment

Except the problem with the CHIPs act is all the requirements the administration grafted on to it. These range from childcare requirements to various labor agreements and diversity goals.

Whether or not you agree these are good things for our society to provide these are large expenses for those companies which directly trades off against the national security benefits of that money. Indeed, in my opinion, it undermines the intent of Congress in passing the law.

Expand full comment

That is why sanctions/bans and tariffs are likely to be more effective (certainly cheaper in terms of government spending). The idea that some partisan hack in the Biden admin or a self-appointed expert will design the “optimal” industrial strategy is insanity.

US investors don’t award high multiples to capital-intensive, commodity industries. Maximizing shareholder returns is what brought us the offshored, infrastructure light, high margin chip industry we have.

There are lots of steps and dependencies in fabrication- building a fab doesn’t solve the problem (no more than building an auto assembly line does much if you lack a parts industry, steel, aluminum, rubber, etc). I have a hard time seeing the US develop a vertically integrated chip manufacturing industry at scale and onshore, but I am no self-appointed expert so I could be wrong 😊.

I am OK with having some expertise and capacity domestically (a la Intel) but continuing to rely on countries like S. Korea, Thailand, Malaysia, India and perhaps even Mexico, Argentina and others for chips and intermediate inputs.

We have interlocking and co-dependent supply chains in almost everything these days, and the financial/competitive incentives seem likely to keep things this way. We need to change the financial incentives so that companies don’t use China as the first outsourcing option for offshore production and don’t use China as the dominant supplier.

End of the day, if China is competitive they are a going to have high market share and low costs. Subsiding our domestic industry based upon what some guy or gal things is “optimal” seems likely to result in a slow-moving, uncompetitive alternative. Let competition and cost drive decisions by incenting investment and trade with friendly countries and restricting it with China. Let Korea and Malaysia and Thailand and India compete against one another to keep things from ossifying (which is what will happen if we go all domestic).

Also, rather than using the administrative state to indirectly punish politically disfavored industries (eg disclosure of fossil fuel exposures, carbon emissions, etc), why not have the SEC require all companies to disclose their supply chain exposures and revenue reliance on China?

I think you might have to look at who the big tycoon donors are in politics to come up with that answer. The tech industry has a few China hawks, but most are China apologists. Not just in the US but also in Taiwan (where they support the KMT). Moreover, outsourced supply chains for commodity products is what has allowed them to focus on the highest value, highest margin activities and created billions in personal wealth and shareholder value.

Not an easy problem to solve and one certainly beyond my competence and knowledge base

Expand full comment

A lot of this is outdated thinking based on conclusions from fantasy academic models that never had much relation to the real world but have become embedded as common sense intuition. Many good bits too but spoiled by the intertwining.

Expand full comment

Ha! It's great that you are on this case.

This is an axe I've been grinding for an age: the problem is that (a) older chips are needed and (b) their price largely assumes manufacturing on a line that's already been depreciated so they can be made without any investment costs, and that (c) if you wanted to make a new line to make old chips, it would be really really hard to economically justify the investment required. I'm real glad to hear that someone's actually working on this. (Talking about it is important, too, of course.)

This isn't all that new a problem: Years ago, a friend asked me to chase down some old chips in the back streets of Akihabara, and I found him a few. (RFID radio chips, maybe. Something radio.) My memory was that I had failed him miserably, but when I ran into him recently, his memory of it was that I had found him enough for the project he was working on. Whew!

Expand full comment

Major Rookie Mistake: the author encourages us to pay close attention to a speech by Xi Jinping and doesn't give us a link. Serious Question I didn't find any clear reference in the speech to the issues raised by Gilnert. The speech was delivered in late 2020 and is mostly focused on China's response to the Covid 19 epidemic. It does discuss supply chain disruptions and developing plans to deal with such disruptions. Perhaps Gilnert can provide more recent statements by Xi that addresses more directly the issues he raises.

http://en.qstheory.cn/2021-01/14/c_581594.htm

https://interpret.csis.org/translations/major-issues-concerning-chinas-strategies-for-mid-to-long-term-economic-and-social-development/

Expand full comment

From the speech by Xi Jinping:

Rather, we must work hard to refashion them into new production chains, and increase the levels of S&T innovation and import substitution across the board. This is an important focus for deepening supply-side structural reform, and is key for high-quality development. First, we must build on our advantages, solidify and increase the leading international positions of strong industries, and forge some “assassin’s mace” technologies. We must sustain and enhance our superiority across the entire production chain in sectors such as high-speed rail, electric power equipment, new energy, and communications equipment, and improve industrial quality; and we must tighten international production chains’ dependence on China, forming powerful countermeasures and deterrent capabilities based on artificially cutting off supply to foreigners.

Expand full comment

Can you give me a link to this in an official Chinese source preferably?

Expand full comment

This is a translation with a link to the Chinese source:

https://cset.georgetown.edu/wp-content/uploads/t0235_Qiushi_Xi_economy_EN-1.pdf

Expand full comment

This is a good opening piece, but a lot more background reasoning is needed to support its conclusions (namely, the suggested remedy actions).

Expand full comment

Nothing wrong with the reasoning, but yes, more information, discussion and analysis of these proposals is warranted.

Expand full comment

Excellent and thorough analysis, couple with useful suggestions for ways forward. Thank you for posting this.

Expand full comment

If it’s gonna be that hard to build new fabs for bigger chips in the US, how about building some in Mexico? Potentially a cheaper way to do it. Would it be palatable to China hawks?

Expand full comment

I'm torn on this. I want to believe the author's analysis, but I'm just not sure about his objectivity. He's a chip-maker CEO who says we ought to be subsidizing the thing he makes. In that sense, this is a great example of the difficulty of relying on "industry experts" in policymaking. Maybe he's correct, but the incentives make it hard to know.

Expand full comment

I work on the leading-edge side of chips (which would not benefit from these policy recommendations -- in fact, we would be hurt by it because we are customers of trailing-edge chips when we make motherboards for our leading-edge chips). My perception is that this policy recommendation is good for the US, modulo details like whether industrial policy is better than tariffs.

Expand full comment

What about near shoeing the more mature nodes to countries like Mexico?

Expand full comment

In the scheme of things, it is much more important to win on the leading nodes. The trailing node problem has many existing solutions. These include: 1) incremental redesigns ... while inconvenient, this happens all the time. 2) build your own supply....the US is too expensive...look to Mexico/Central America. 3) control skew counts and minimize dependencies on those skews. 4) Flexible low-volume manufacturing: DoD already does this to some degree today, but with some innovation, one can build low-volume manufacturing which can handle multiple process nodes.

Going forward, the strategic way to solve this problem is to Design for Supply Chain (EDA): On the lagging node side, one can build system architectures in EDA for easy shift.. this leverages a field known as reconfigurable computing. This can easily be done for processors and digital computing today. Analog/Sensors is emerging with the combination of AI/digital nose techniques.

Expand full comment

Not to mention that today's leading node sizes will be the trailing node sizes 5 years from now.

Expand full comment

The process nodes mentioned in the article were

180 nm (1999)

90 nm (2003)

65 nm (2005)

So the time scale is a little longer but the point stands.

Expand full comment

Alas, this post rapidly devolved into "Charlie Brown adult-speak" for me. Basically all I got from it was that less-sophisticated chips are important too; the CHiPs Act doesn't sufficiently cover those yet; and we need to expand chip-related tariffs. I think that could have been stated in 2-3 paragraphs.

Expand full comment

“Not to be overly cynical, but the fact that Intel and GlobalFoundries definitely contributed to the contents of the bill should make it rather unsurprising that this was the outcome.” INTL is structured and making the same mistake it made more than a decade ago: now that INTL was pushed into the chip-fab business by market forces, it still is status-seeking and wants to be in the sexy end of the chip business: fabricating 3nm chips. Cripes, that machine costs $380 million, to say nothing of the maintenance/repair costs. If anything will put constant pressure on margins, this will do it. INTL is starting out on the wrong foot in the wrong location. Only government officials incentives can force INTL to quit making the same mistakes.

AMD is building its largest-ever fab in the north. What are the government incentives for that fab to produce trailing-edge chips.

“Rust never sleeps.” The way to reawaken the Rust Belt is to put trailing-edge chip fabs in an area that has decades of industrialization experience. Existing infrastructure can be upgraded. It has the greatest fresh-water resources. Numerous vendor businesses serving a chip fab will spring up like mushrooms across the Great Lakes Region. It’s colder in these climes, which will reduce the cost to heat/cool fabs. Springsteen wrote: “those jobs are gone and they ain’t coming back.” The right policy will bring jobs back. Just how hard is it for policy makers to understand how to steer government incentives. Replace politics with geography.

As for amortizing old equipment: it should be cost-effective when you consider 90% of ASML machines are still working. I think this expert is right when he writes about the mud-structure of semiconductor corporations. If they aren’t willing to restructure, don’t give them government funding. As we know, Silicon Valley was built with U.S. Defense Department money. Before that, the railroads were financed out of Silicon Valley with government money. This is doable.

Expand full comment

Incredibly informative. Thank you.

Expand full comment

Really good guest post. Kudos to the author and Noah both.

To my mind, we started down this path (of giving up on manufacturing commodity chips) in the 80s with RAM. until that point one could source almost every type of semiconductor and integrated circuit from a US supplier.

Expand full comment

Strong focus on mature nodes and supply chain security is critical. I particularly appreciate the analysis of the economic and geopolitical ramifications. While tariffs and sanctions may be unpalatable, they could be necessary to curb China's rapid advancement. Ultimately, a multifaceted strategy that prioritizes both mature node production and robust supply chain monitoring is essential for long-term semiconductor security.

- marketeer @ https://www.cientra.com/

Expand full comment

“there are suspicions that the relentless waves of negative stories about TSMC in Arizona may be misinformation from China; there certainly is a motive, and there have definitely been real challenges, but the fact remains the first fab is on pace to open next year, and TSMC has actually increased its commitment to the site.”

-- Stratechery

ASML employs people from 90 countries. It makes a conscious effort to encourage subcultures/social groups within its workforce to make employee transition less fractious.

Maybe it’s harder to blend two cultures (U.S./Taiwan than 90 cultures.

Expand full comment