DIEing to stop CHIPS

And other government roadblocks

A Micron chip plant in Hiroshima (source)

Via powerlineblog I found an article in the Hill about how the USA’s CHIPS act to fund semiconductor making has been subverted by DEI.

DEI — the identity-obsessed dogma that goes by “diversity, equity, and inclusion” — has now trained Google’s new AI to refuse to draw white people. What’s even more alarming is that it’s also infected the supply chain that makes the chips powering everything from AI to missiles, endangering national security.

The Biden administration recently promised it will finally loosen the purse strings on $39 billion of CHIPS Act grants to encourage semiconductor fabrication in the U.S. But less than a week later, Intel announced that it’s putting the brakes on its Columbus factory. The Taiwan Semiconductor Manufacturing Company (TSMC) has pushed back production at its second Arizona foundry. The remaining major chipmaker, Samsung, just delayed its first Texas fab.

The TL;DR summary is that the CHIPS act requires recipients to hire the usual females and minorities even when it makes zero sense and is probably impossible (e.g. female construction workers and childcare for them). Hence progress has been slower than desired and companies are looking elsewhere for additional future fabs because they see no point in taking US subsidies if the subsidies force them to make poor business decisions

By comparison, in Japan, TSMC has just opened it’s new Kumamoto fab with actual production expected to start in a few more months.

The opening marks a significant milestone for JASM, which was established in 2021 and started construction in April 2022, with production targeted to begin by the end of 2024. TSMC and minority investors Sony Semiconductor Solutions (SSS), DENSO and Toyota also recently announced further investment into JASM to build a second fab, which is scheduled to commence construction by the end of 2024 and begin operation by the end of the 2027 calendar year.

The site has been built in two (2) years. From initial contract signing (late 2021) to actual chips out the door is going to take three years. A second fab is starting construction anytime now and should be done in 2027. The second one will be larger and more advanced/complex than the first.

The chipmaker is expected to invest around 2 trillion yen ($13.5 billion) in the new plant, raising hopes that it will boost the economy in the southwestern Japanese prefecture. However, it could lead to complaints about groundwater pollution and traffic jams, locals said.

TSMC's first plant, a wafer factory, cost $8.6 billion, with mass production scheduled to begin later this year.

Tom’s hardware had a recent article on the (slow) speed of US fab construction that makes the contrast very clear.

The CSET study looked at fab construction between 1990 and 2020, and concluded that for the roughly 635 fabs built in that timeframe, the average time between the start of construction and production was 682 days. Three countries beat that benchmark: Taiwan at 654 days on average, Korea at 620, and Japan at a staggeringly fast 584 days. Meanwhile, Europe and the Middle East were about on par at 690 days, as was China at 701 days.

However, the U.S. clocked in at 736 days, well above the worldwide average and second only to Southeast Asia at 781. Things look even worse when you look at specific decades. In the 90s and 2000s, the U.S. was pretty fast and saw average construction times of about 675 days. In the 10s, that number dramatically increased to 918 days. Meanwhile, China and Taiwan were going at a much faster pace that decade, with an average completion time of 675 and 642 days, respectively.

And the slowdown looks to have coincided more or less with Obama and his pandering to lefty pressure groups. In this case environmental ones to begin with, though the DIE stuff of the last few years certainly doesn’t help and neither does the duplication of regulators at the federal, state and local levels

Regulation in the U.S. can be incredibly complicated thanks to its federal structure, which has created a hierarchy of one national government, 50 state governments, and countless local governments. Fab construction needs to comply with three rulebooks, which is significantly more complicated than the situation in countries like Taiwan. Solving these regulatory issues is likely to be challenging at best.

America's environmental protection policies have also hindered fab construction. Three government bodies at the federal level alone have their own policies, with the Environmental Protection Agency (EPA) being the largest, and then state- or local-level agencies like the Texas Commission on Environmental Quality.

However one does wonder why “[t]hree government bodies at the federal level alone have their own policies”. A sane government would only have a single bureaucratic hierarchy regulating things at a single level of government. And a really sane subsidiary level of government would ensure that meeting its requirements also met the requirements of the upper level so things only needed to be submitted and approved once. Of course all this would reduce the demand for bureaucrats so we know why that doesn’t happen

It is notable that as a result of the 2020 supply chain issues and the general covidiocy both the USA and Japan decided they really needed to have more critical bits made in places other than the PRC. In the US this resulted in the CHIPS act, with lots of hot air, much posturing and limited success. In Japan, as I’ve noted above, words have turned to actual money and thus into actions. Moreover unlike the US where the focus seems to be on the high visibility parts of the process - the latest generation of semiconductors - in Japan the focus has been broader. Yes they have TSMC building fairly high-end fabs and planning to build even more advanced ones, but they also have paid some attention to other parts of the supply chain.

Companies like Murata are building new factories to make electronic components at a rapid pace. The Murata subsidiary local to me built another factory that opened last year for ceramic capacitors which was built in about a year, and has just started construction on yet another, much bigger one that will be finished in a couple of years.

[Update: in contrast, US efforts to induce suppliers to build local seems to be also hitting issues with construction costs and labor availability (archive link)]

As the Nikkei points out in a recent spotlight article, Japan may have lost its 1980s leadership in chip-making, but it still has leadership in much of the supply chain (e.g. Ebara, Tokyo Electron and Shin-Etsu as manufacturers of the machines that help turn lumps of silicon into chips). Further more companies like Sony and Toyota are major consumers of the end product. Many Japanese electronics companies outsourced their product building to the PRC in the 1990s and 2000s but now want it back, and also want the entire supply chain back and predictable even if it costs a little more.

The Nikkei article does note one niggling problemette - skilled labor. But at least the Japanese government isn’t trying to force companies to hire minorities or under-educated technicians. It seems likely that the labor issue is solvable by a combination of new graduates, guest-workers and robots.

I’m generally skeptical of government industrial policies; Japan has blown a lot of money on failed ones, as has the US (solar power etc.), and many other countries (see France or Germany) have also wasted money promoting “national champions” that have generally failed to provide much benefit, frequently even failing to help the re-election efforts of the politicians that subsidized them. But to a degree they seem to be required to counteract short-sighted corporate behaviors and to handle geo-political threats. If have to have them they need to be more like Japan, broad support across the whole supply chain, and not like the US where they come with strings attached for irrelevant side-goals that make them of questionable value.