Two ‘Pseudo-Facts’ About Semiconductors That Could Distort The CHIPS Act (Part 4)

Two ‘Pseudo-Facts’ About Semiconductors That Could Distort The CHIPS Act (Part 4)

  • “Demand for computer chips is already far outstripping supply, a problem that will intensify, [Commerce Secretary] Raimondo told journalists… ‘We are wasting time, precious time, every day that the Chips Act isn’t passed and appropriated in Congress.’”

The policy debate over what the federal government can do, or should do, about the “chip shortage” has crystallized into the “CHIPS for America” Act, passed in June by the Senate. Whatever the specifics – the draft bill is rather sketchy, both vague and over-specific – there may soon be $50 billion up for grabs. Let the lobbying begin. 

The challenges facing the semiconductor industry are exceedingly complex, highly technical, and subject to conflicting “expert” interpretations. Most people – and especially legislators, by nature – will instinctively look for some “bottom line,” a radical simplification that will bring into focus the “Key Facts” which will clarify the choices we face. These “Facts” become the headlines in the media, and the bullet-points in the advocacy briefs and Powerpoint slides – and, after a time, through the force of repetition, they turn into “Common Knowledge.” 

All such “Facts” are suspect, even the ones that are more or less correct. They are by definition over-simplifications, stripped of nuance and detail. But some of them are also simply wrong. “Pseudo-Facts” confuse public opinion, and mislead policy-makers – but they can be quite hard to root out of the discussion. 

Two prominent examples of Pseudo-Facts that baffle and bedevil the semiconductor policy deliberations are:

  1. The idea that China is an active threat to U.S. technological superiority in he chop business, working from a position of strength
  2. The idea that technological superiority — the “leading edge” – is defined by the so-called “process node” score – that is, the scale of the individual circuit elements etched into the silicon wafer — the player with the lowest number wins

These two Pseudo-Facts are often joined together, as here in testimony from April by the Secretary of Commerce Gina Raimondo:

  • “It is not an exaggeration to say at the moment that we have a crisis… America once led the world in making leading edge semiconductor chips. Today we produce 0% of those chips in America, 0%! That’s a national security risk and an economic security risk. The Department of Defense has been warning us for years…[We have to] protect ourselves. We are totally reliant on Taiwan and China for critical supply.”

This seems dire. Has the U.S. lost its leadership in this critical technology? Are we somehow we are “reliant” on China now? Does China have the upper hand? Secretary Raimondo’s analysis is darkly pessimistic: “We have seen an unbelievable decline in…semiconductors….” 

Wrong. Based on wrong assumptions and gross misunderstanding of the nature of the industry. Here is the counter-view.

China: The “Trailing Edge”

The idea that China has mutated from a supplier/partner, to a rival, to a hostile adversary, is the overarching theme of the past half decade or so pdf geopolitical commentary, and it may be accurate. The Chinese economy is very large and has grown very fast. China wields greater economic power than they did twenty years ago. But it is not accurate to assume that China enjoys a position of strength in semiconductor technology. It is rather the opposite: China is extraordinarily weak in this critical sector for the modern global economy. The weakness is both quantitate and qualitative, and it is not a problem that China has figured out how to fix.

The Quantity Deficit 

China imports today about $380 billion in semiconductors – 90% of its needs – , more than it spends on oil. Chinese customers for semiconductors (e.g., its huge electronics industry) are acutely vulnerable, as the Huawei episode has shown. (The world’s largest wireless company just a couple years ago has been crippled by the denial of U.S. chip technology.) 

Beijing has announced Big Plans to achieve self-sufficiency. Specifically, they want to be able to supply 70% or more of their needs by 2025. Success looks unlikely.

  • “Semiconductor market research company IC Insights forecast that semiconductors made in China would account for a bit under 20% of semiconductors sold in China in 2025, far below the Chinese government’s 70% target…. Chinese companies accounted for only $8.3 billion, or 37%, of China’s IC output, the rest being produced by the local silicon wafer fabs (semiconductor factories) of TSMC, Samsung, SK Hynix, Intel

    INTC
    and other foreign companies.” 
  • “[Indigenous Chinese firms accepted for] just 6% of China’s IC market and only 2% of the global market.The total value of semiconductor production by Chinese companies in 2020 was less than that of AMD, which ranked 15th in the industry with estimated sales of $9.5 billion. The sales of the top three companies – Intel, Samsung and TSMC – are estimated to have been $73.9 billion, $60.5 billion and $45.4 billion, respectively.”

The Quality Deficit

Even this overstates China’s position. Indigenous producers are technologically on the “trailing edge.” They cannot compete successfully even for the “free market” customer segment inside China:

  • “The Chinese military is the primary consumer of Chinese chips, because export controls prevent it from accessing leading non-Chinese chips. Non-Chinese chips, especially U.S. chips, dominate China’s civilian sector.” 

These firms stay in business mainly because of government support, and show little evidence of being able to win in open competition.

  • “Although China’s [firms] look strong, much of that capacity is aspirational (suffering from low yields and utilization) and at older nodes. Many of these fabs stay online with the help of state support, receiving subsidies far greater as a percentage of revenue than any leading fabs. Foreign chipmakers operate the most advanced and reliable fabs in China and generate more revenue than the country’s chipmakers.” 

The U.S. Semiconductor Industry Association (SIA) summarizes China’s standing as follows:

  • “Chinese chip firms are notably absent in the market for high-end logic, advanced analog, and leading-edge memory products. China’s indigenous semiconductor supply chain is even less developed. It lags significantly in advanced logic foundry production, EDA tools, chip design IP, semiconductor manufacturing equipment, and semiconductor materials. Chinese foundries currently focus on more mature nodes, and China’s supply chain capabilities at the equipment and materials level are presently limited to older technologies.” 

No Presence In The High-Value Industry Segments

As described in several previous columns, the “semiconductor industry” must be understood as a segmented eco-system, in which certain segments (Chip design, Core IP, Semiconductor Manufacturing Equipment) generate most of the value-added. China has essentially no share in any of these high-value segments. 

Inadequacy of Chinese Government Efforts

Those who maintain the validity of the China Threat are often reduced to the argument that even if China is not strong now, Beijing could mobilize massive investments to catch up. As again from Secretary Raimondo:

  • “China [is] not waiting. They are incentivizing and subsidizing the production of chips right now, and they have been for a long time.”

Pseudo-Fact. In truth, Beijing’s efforts to create an indigenous chip industry have failed for two decades. Announced plans for the coming years do not match up against the much larger investments by private sector players in the U.S. and its Asian allies (Korea, Taiwan). Also, keep in mind that Intel, TSMC, Samsung et al have a crucial advantage:They know what they are doing. The same cannot be said for the CCP bureaucrats whose key idea seems to be to headhunt Taiwanese executives to run their chip enterprises.

China’s plans are dwarfed even by Korea alone. Between private and public sector, the Koreans will spend almost half a trillion dollars to build up their chip industry in the next decade.

  • “South Korea is going all out to bolster its critical semiconductor industry, with the government on Thursday announcing a plan by companies to invest 510 trillion won ($451 billion) to boost chipmakers’ competitiveness amid a critical global shortage of the key components.The corporate outlays through 2030, including 171 trillion won announced by Samsung Electronics alone, come as the country is facing severe challenges from competitors including Taiwan Semiconductor Manufacturing Co., which leads the world in the output of foundry chips — those made for other companies. ‘Our government will unite with companies to form a semiconductor powerhouse,’ said President Moon Jae-in. ‘We will support companies concretely.’”

Korea’s initiative is three times the size of China’s program. As the headline of one recent article put it – “Get Real About the Chinese Semiconductor Industry.”

The Process Node Yardstick

A 2nd Pseudo-Fact is the assumption that technology leadership in the chip business is all about driving down the “node size” in the chip fabrication process. Measured in nanometers, the scale of the individual circuits etched into the silicon has come to be the unidimensional metric of success for the industry. And so, when Intel (once the leader in driving down the node size) “stumbled” a couple years ago, and delayed its next downward step, it was said to have “fallen behind” its main Asian rival (Taiwan Semiconductor Manufacturing Corporation – TSMC) which would field a smaller “node” in production sooner. 

Cue Secretary Raimondo’s panicky pronouncement to Congress: 

  • “America once led the world in making leading edge semiconductor chips. Today we produce 0% of those chips in America, 0%!”

Again, however, this concern is largely misplaced. Node size is not all-important.

Relabeling

First of all, the “node” metric characterizes only one dimension affecting performance — the purely physical. IC design is just as important. Two different nodes, with different designs, may functionally equivalent.

  • “It initially seems like Intel is two chip generations behind TSMC. Yet Intel’s 10-nanometer chips are just as dense as TSMC’s 7-nanometer chips, with roughly 100 million transistors per millimeter squared. In other words, Intel’s 10-nanometer chips are technically comparable to TSMC’s 7-nanometer chips… Intel is renaming its 10+ node, also known as the 10nm Super Fin node, as its “new” 7-nanometer node. These new chips, which will launch by the end of 2021, should offer better performance than TSMC’s 7-nanometer chips, but fall short of matching TSMC’s 5-nanometer chips.”

Next-Gen Tech

“3 nm” – the current leading-edge process node – is today’s headline must-have. It is baked into the CHIPS Act — which is a mistake. The Strategic Fact about semiconductors today is that Moore’s Law is coming to its end, reaching its hard physical limit. The next few years will see some further progress, but with diminishing returns. It’s call an “asymptote.”

New paradigms are emerging, like neuromorphic computing (in which Intel has taken a strong interest), and quantum computing, along with other design-based approaches that side-step (rather than assaulting frontally) the physical limits of traditional processor designs.

“Process node” competition is hardware-intensive, physics-intensive, materials-intensive technology. It is not the only way to drive progress. Design-driven solutions (software) can sometimes overleap hardware-intensive technologies, and with far superior economics. ARM (a core IP supplier) has achieved near-monopolistic dominance of some applications (95%+ of the smartphone market – the largest processor application today), not by brute-force hardware innovation, but by the opposite – subtle-force software innovation.

These new approaches will eventually displace today’s “process node” fixation.

Clear Thinking

A thorough deconstruction of these two Pseudo-Facts – the Overblown China Semiconductor Threat, and the Process Node fixation – would involve more complicated arguments than can be presented here. Still, this oversimplified assessment of these oversimplified (and wrong) notions can help to illuminate some of the misleading propositions that have been baked into some of the current proposals, including CHIPS. Perhaps in time to make adjustments.