Playing Catch-Up with Stalin, Mao, and Chairman Xi

It has been a long time coming, but we’re finally there.

Ninety-eight years after Stalin’s first Five-Year Plan and a mere 71 years after Mao’s first Five-Year Plan, the United States of America finally has its own Five-Year Plan. According to the White House’s just issued National Strategy on Microelectronic Research, “over the next five years” the White House and federal departments and agencies, “taking a whole-of-government approach,” will work together to:

• Enable and Accelerate Research Advances for Future Generations of Microelectronics
• Support, Build, and Bridge Microelectronics Infrastructure from Research to Manufacturing
• Grow and Sustain the Technical Workforce for the Microelectronics Research and Development to Manufacturing Ecosystem
• Create a Vibrant Microelectronics Innovation Ecosystem

It’s Holy Week, the culmination of Lent. As penance for our sins, we actually read the entire 60+ page report. You’re welcome.

The essential goal of Stalin’s and Mao’s plans was to build an industrial economy via slave labor. You could always see the blood flowing between the lines.

The White House Plan is rather in the new style pioneered by the Chinese Communist Party since the 1990s. Like the plans of Chairman Xi and his immediate predecessors, the top goal seems to be to grab credit for astonishing progress to which no politician has or will contribute in any way.

Both the Chinese and the U.S. plans proceed by the same method. Operating at the level of an unusually well-spoken high-school student, or perhaps ChatGPT on a good day, they recite the recent history of progress in a technology and then say “we will do more of that” as if they had done any of that in the first place.

Illustrative is the first section of the report, detailing the need to develop advanced semiconductor materials, including “wide-bandgap” materials such as silicon carbide (SiC), gallium nitride (GaN) and indium phosphide (InPh) as if they were new developments.

All three have been known and used by the industry for decades. SiC, for instance, is news not because it is a new development but because the SiC market has been surging as the material handles high voltages better than silicon and is therefore in demand for electric vehicles. The only contribution the government has made has been to artificially promote a market for those unpopular cars. That effort, costing hundreds of billions of dollars, and doing perhaps trillions of dollars in damage to U.S. automakers, looks increasingly unsuccessful, raising the distinct possibility that the industry has overinvested in SiC materials and devices.

Nowhere mentioned in the materials section is the discovery of the decade, announced in January of this year, of the first functioning semiconductor made from graphene. The most thermally conductive material known to man and with 3,000 times the “charge carrier mobility” of copper, graphene microchips have the potential to exceed the computational speed and power of silicon by orders of magnitude while generating orders of magnitude less heat.

Perhaps the White House omitted to mention this development because the graphene semiconductor was developed in a cooperative effort between a Georgia Tech professor from Holland and China’s Tianjin University. What a scandal!

The report dolefully notes that the U.S. share of backend chip manufacturing — protectively packaging the finished chips — had shrunk to 3%. That was true for as long as packaging was the most down-market, lowest-margin part of the business. This just in: new 3D chip architectures and multi-“chiplet” systems have made packaging important, hard, and profitable again.

Intel (INTC), arguably the technology leader in the field, is depending in part on its packaging in its challenge to Taiwan Semiconductor’s (TSM) overwhelming dominance of the foundry (contract manufacturing) sector. Nvidia has already announced that it is moving some of its packaging business from TSM to Intel.

The most obnoxious (as opposed to merely clueless) part of the report is its plan to “Create a Vibrant Microelectronics Innovation Ecosystem.”

We have, or had, for decades the most vibrant, global innovation ecosystem in history. It spans the United States and Europe (a complex supply chain centered in Holland makes the Extreme Ultraviolet machines for semiconductor manufacturing), with key contributions from Japan. Meanwhile, Taiwan remains the global center for actual chip fabrication, with expanding tentacles in mainland China.

For several decades, China’s chief contribution to the ecosystem was as a customer. Buying tens of billions of chips from U.S. companies, chiefly manufactured with Dutch equipment in Taiwan, and sometimes finished in China or other emerging Asian economies, the Chinese supported global progress on the learning curve. Without Chinese consumers and product assemblers, the magnificent seven of the U.S. stock market would fall far short of magnificence.

Lately, however, as the Georgia Tech-Tianjin collaboration shows, China has become a leading source of innovation at the highest level. Quite possibly the first company to release a commercial graphene microchip will be Huawei. If so, the first firm to manufacture the graphene chip in volume would be the Semiconductor Manufacturing International (SMIC) China’s leading foundry.

Flatly untrue is the White House claim that consolidation of chip manufacturing “has imposed limits on the associated R&D ecosystem” and that, as a result, “researchers in academia, government, and industry who do not require high-volume production have limited access to the capabilities needed for advancing the R&D frontier.”

Nearly opposite the truth, those sentences miss the greatest source of semiconductor innovation since 1987. Before Taiwan Semiconductor (TSMC) created the foundry model, in consultation with Caltech’s Carver Mead, wafer fabrication was unified in the same companies as design of new chips. To bring a new chip to market, the designer had to either own a factory costing billions of dollars or work for someone who did. Splitting design from manufacturing, TSMC liberated the world’s chip designers to create thousands of new application specific integrated circuits (ASICs) customized for particular purposes and often manufactured in small batches, like premium bourbon.

Today, TSM makes most of its revenue from its 10 top customers. But it still manufactures for more than 500 design shops, most of which would not exist had TSM not created the foundry model. Nor is it alone. A clutch of specialized foundries, from Israel’s Tower Semiconductor (TSEM) to Minnesota’s own Skywater Technologies, focus on bringing exotic and often even shorter-run designs to market.

If the magnificent seven should fall — already Huawei is eating Apple’s lunch in China — it will be because the consensus view in Washington, D.C., is that our highest national security priority is to rip apart this most bountiful global “ecosystem” and replace it with subsidized dinosaurs doing the government’s bidding.

P.S. I just finished a video interview with Gabriel René, the CEO of VERSES.AI and the executive director at the Spatial Web Foundation. We had an illuminating discussion on how it is helping to lead the artificial intelligence revolution. Click here to watch it! To become a member of my Moonshots trading service, which is solely dedicated to finding companies like VERSES, click here!