A different take on our Govts $500B mal-investment in Chip Mfg

[surfrebel]

Guideposts: CHIPing Away at U.S. Technology Leadership​

by George Gilder and Richard Vigilante 10/25/2023​

The CHIPs Act, which allocates some $500 billion to subsidize manufacturing of microchips in the United States is a catastrophic example of the “damage multiplier” of government spending. The biggest damage comes not from waste of the government’s money but the resulting massive misallocation of private capital. This is not a new thing. Consider government subsidies for electric cars and trucks. The butcher’s bill only starts with the subsidies themselves. Yes, those multi-billion government payouts are nearly irrelevant to their stated task of thwarting global warming; EVs only trivially reduce CO2. The subsidies are burnt money. Far worse, however, is the damage to the U.S. auto industry and its workers. Money-losing electrics that cost more to make and which consumers reject shred profits and are destroying jobs. And although EVs cost more to make, they require less labor and will lead to layoffs. The UAW strike is a symptom of the subsidies. The CHIPs act is driving massive over-investment in semiconductor manufacturing around the globe. Last November, the industry pub Semiconductor Engineering assembled what is a partial list of planned investments by the industry into semi-conductor manufacturing capacity, covering only announcements made in 2021 and 2022. These totaled up somewhere between $300 billion and $450 billion, and were announced even before the European Chips act was passed last summer, which will add $47 billion to the subsidy pile. Even as his article was being published, Ed Sperling, editor in chief of the journal, estimated that at least another $200 billion of chip infrastructure investment was in the planning stage, though not yet announced. Sperling noted that the spending splurge risk creating “talent shortages, duplicative inefficiencies, and” yet another chip glut “that will spawn price wars and inventory write-downs.” Selected New Investments in Semiconductor Manufacturing Capacity By Country 2021-2022 Source: Semiconductor Engineering

The problem the subsidies seek to address is not unreal: The pandemic revealed how breaks in the semiconductor supply chain can be devastating across the entire economy, as chips fly out of the home and office and pervade every human activity our in the wide world. Unfortunately, the shortages reinvigorated the foolish notion that we need “semiconductor independence.” This parallels the wrongheaded 1970s argument that we needed “energy independence” because of an Arab Oil embargo that happened exactly one time, lasted for only six months and was entirely the result of Nixon breaking the link between gold and the dollar in 1972. Arab hostility to Israel and the Yom Kippur war provided a convenient excuse but OPEC’s real motive was to break dollar-denominated contracts at a time the dollar had lost two-thirds of its value, compared to gold, and would go on the decline by 90% by the end of the decade. The Arabs can’t drink the oil; they must sell it to someone. And it is as fungible as any internationally trade commodity. It is a fine thing that the United States, pre-Biden, was making money as an energy exporter. But it is absurd to suggest U.S. power or prosperity depends on whether we live on our own oil. Of the 10 largest economies in the world, only two, the United Kingdom and Russia, are net energy exporters. The proposition that we need to manufacture our own chips to avoid a supply crunch is even more foolish. Yes, there is the Taiwan question. It is risky that one Taiwanese company, TSMC, has a 50% global share of contract chip manufacturing. Worsening the risk, U.S. foreign policy makes a seizure of Taiwan ever more attractive to the CCP. The diversification of chip manufacture to other nations, especially in Asia, is a good thing for that reason. We should not, however, be deluded that chip manufacturing itself is especially attractive economically. Nor is a strong chip manufacturing sector an indicator of technology leadership.

Global GDP now stands at nearly $90 billion annually. Some enormous portion of that—half? two thirds?—would disappear if there were no such things as microchips. Yet the microchip industry, which enables all that wealth, is itself rather small. Global semiconductor sales last year were roughly $570 billion and are not projected to exceed $1 trillion for another few years. Moreover, of that $570 billion, the lion’s share went not to the firms that manufacture chips, but to the firms that design them, such as Nvidia (NASDAQ: NVDA), Advanced Micro Devices (NASDAQ: AMD), Qualcomm Inc. (NASDAQ: QCOM), Arm (NASDAQ: ARM) and more. Even the declining number of large, integrated design and manufacturing (IDM) semiconductor firms such as Intel (NASDAQ: INTC), Texas Instruments (NASDAQ: TSN), Micron Technology Inc. (NASDAQ: MU), and Samsung Electronics Co. (KRX: 005930), make most of their profits from design. The $40, $50 or $60 trillion in global GDP enabled by this tiny industry is a miracle of what economists call “consumer surplus,” the value a product creates above and beyond what the product makers will ever put in their pockets. A big part of what we do at the GTR is analyze the competitive advantage of semiconductor firms. We have no interest in trivializing their importance. Yet, if not a single chip were ever again manufactured in the United States, it is far from obvious that this would be a negative for the U.S. economy, as long as we could import them via a secure supply line. Spending half a trillion dollars to obviate the need for a secure supply line is about as foolish a waste of money as our government has ever perpetrated. Every dime of capital investment over and above what U.S. chip companies would have invested without the subsidies is misallocated capital. And this is an especially bad time to be wasting capital in this industry. The end of Moore’s Law as conventionally understood—shrinking circuit sizes as the way to increase computational power per dollar—is not a dead end for the industry. But it does place a huge burden on the design guys to come up with new ways to sustain a new version of Moore’s Law in which innovative chip architectures and materials, rather than shrinking line-widths, continue to increase value. So far, the industry has been doing a great job: chiplets, embedded memory, advances in packaging that add 10X more value than the old “stick it (in an insulator) and ship it” approach used to, are powering the industry forward. When we get the graphene microchip, progress will accelerate 10X or more. To continue progressing, however, requires efficient deployment of human capital, especially engineers. The subsidy programs threaten to massively divert engineering talent away from design and into manufacturing. Some sense of this distorting demand can already be seen in TSMC’s decision to delay (and probably cancel, though they don’t want to say it out loud) the company’s plans to build a foundry in Arizona. The company just could not find enough skilled labor to go ahead. That scenario is going to repeat over and over as the CHIPs act distorts investment. “The number one choke point going forward is talent,” offers Ondrej Burkacky, senior partner at McKinsey & Co. ”What is going to limit the growth of the silicon industry isn’t going to be lithium or neon,” Burkacky said. “It’s going to be people.” The silliness of “semiconductor independence” will only make getting the right people into the right place harder.