The chip industry and national security
This is the concluding article in a two-part review of Chip War, a book by Chris Miller. For Part 1, click here.
One significant point that Chris Miller (and, incidentally, many policymakers) seems to miss in his book Chip War is that discussions of semiconductors and national security are emphatically not solely, or even primarily, about the fabrication plants where chips are built.
One response to the China challenge is to persuade Taiwan Semiconductor Manufacturing Company (TSMC) or Samsung to build a modern fab in the US, and operate it as a foundry. But this would only ensure access to a specific process node (for example 5-nanometer or 7nm) that happens to be leading-edge today.
The US needs to ensure secure, asymmetric access to the world’s best chips across multiple generations. Building the world’s best chips is not a matter of a single factory or a single company: It’s a product of maintaining an entire semiconductor ecosystem, at home and within the borders of allied and partner regimes.
There can be no chokepoints – fabrication and metrology tools, expert knowledge, chemicals, packaging, testing, design tools, or anything else – located solely in adversarial regimes. Maintaining semiconductor supremacy is about maintaining US and allied dominance across the entire ecosystem.
The liberal-democratic/Western semiconductor ecosystem still has enormous geographic stickiness and, in Halford Mackinder’s parlance, “social momentum.” Moving factories is enormously disruptive and expensive; engineers and executives don’t want to move their households and families to distant and culturally incompatible places.
A critical question is how to leverage what remains of this social momentum in the US to enhance US dominance despite the fact that cutting-edge changes occur rapidly, across all areas of the semiconductor ecosystem.
The US retains enormous momentum in the electronic design automation tools, the design of chips, and the development of fabrication and metrology tools. Taiwan has accumulated enormous momentum for the fabrication and packaging sides of the ecosystem, especially for cutting-edge digital chips, and is moving into design.
China has very little momentum in most of these areas, but has made huge progress and constructed considerable momentum in trailing-edge and more-than-Moore chip fabrication technologies, as well as in integrating packaged chips on to boards and into finished products.
The Communist Party of China (CPC) is spending money at an enormous pace to try to jumpstart ecosystems in all of these other areas.
Today’s most advanced chips from TSMC are in the 3nm node; if the US government pays TSMC to build a 3nm plant in the US today, the United States will have ensured access to 3nm chips. It will not, however, have secured access to the intellectual and organizational machinery that is constantly innovating and building newer and better fabrication processes and plants.
Taiwan’s social structure and its education system undergirds TSMC and that constant innovation, rolling out a new process every couple of years, is the core of TSMC’s long-term value.
The tools to implement these fabrication processes change too, and at the same pace. Every generation of new chips needs new fabrication equipment, new chemicals, new software, new metrology equipment, new test equipment, new higher-density printed circuit boards, new package designs, new design IP, etc. And every one of these pieces is a potential chokepoint for an adversarial regime, should it manage to wrest control of it.
Consider the recent behavior of China with regard to the supply of rare-earth elements, a critical component in the development of many products relying on advanced technology: Upon establishing a monopoly on their supply through legal (and lethal) environmental externalities and government subsidies, the Chinese immediately made use of that leverage for political ends.
Some niches of the semiconductor ecosystem are actually natural monopolies. ASML has emerged as a monopoly because developing and building the EUV (extreme ultraviolet) steppers that are required for advanced node manufacturing is so ruinously expensive that ASML’s big customers were forced to directly subsidize their operations.
Demand for these machines, while large in dollar terms, is small in terms of number of machines; absent being allowed to sell them to China, there are only a handful of customers for ASML’s product. Setting up a competing source for these machines would be very expensive, so the users of these tools instead have made a choice to allow themselves to be sole-sourced, rather than paying to set up a competitor.
From a policy perspective, having this expertise exist only in one firm is quite dangerous, and Western governments should consider whether this is a situation that should be allowed to continue.
TSMC is also a natural monopoly. This is not because of the cost associated with bringing up a new process: Spending tens of billions is well within the capabilities of almost any large state, and certainly of many large companies.
Fabless semiconductor companies that use TSMC’s fabrication services get to tap into a very rich ecosystem of pre-designed IP blocks, from an entire network of different vendors. As a result, TSMC users can often license 90% or more of the design of their chip, while their engineers design only the pieces that generate significant technical differentiation.
Doing so greatly reduces risk, and improves time-to-market. This network effect has not yet been replicated at Samsung or Intel; IP providers flock to the processes where they can capture the most users, and users flock to the processes where they can get access to the most robust basket of licensable IP.
Because TSMC is so dominant in terms of number of users, the IP vendors use scarce resources to develop for TSMC, instead of Samsung and Intel.
If the US wants to create an onshore alternative to TSMC to counter Chinese threats to Taiwan, building a factory is not enough: The entire ecosystem would need to be duplicated.
Creating redundancy for TSMC would not be just a matter of standing up processes and factories; it’s a matter of replicating this entire design ecosystem on top of those processes, plus the downstream offerings from TSMC’s partners in the packaging and test ecosystem. Network effects make building an onshore alternative to TSMC far more difficult and expensive than it might initially appear.
Having a parallel ecosystem in the US is both a national-security imperative, and, ironically, a great way to reduce the likelihood of an attack on Taiwan.
Right now, the US needs to defend Taiwan for multiple reasons: Taiwan is an ally that shares liberal democratic values, it is the most important source of advanced semiconductor chips, and it is the linchpin of the First Island Chain, helping to bottle up the PLA (People’s Liberation Army) Navy in home waters.
Removing the dramatic level of US dependence on Taiwanese chips would eliminate one of the incentives for an attack by the People’s Republic.
So long as TSMC remains a Taiwan-based company, the entire chip-manufacturing ecosystem is vulnerable to disruption – even if TSMC builds some factories in the US, as it is doing. The Communist Party of China may one day decide that reunification is a more important consideration than preserving the TSMC foundries.
And with China rapidly losing access to Taiwanese-manufactured chips because of US export controls and their effect on TSMC, it is entirely possible that the CPC will calculate that destroying TSMC in the short term is a net benefit for China, giving it a comparative advantage.
The China challenge
US semiconductor companies and those located in allied countries must never be allowed to fall into a situation where any critical part of their business – whether it’s things they need to buy, people they need to hire, capital they need to employ, or indispensable customers – flows through adversarial regimes.
Any dependency on adversaries – and the primary adversary is China, along with its allies Russia and Iran – needs to be ruthlessly and rapidly addressed through aggressive policy measures. This is true not just for chip companies, but for the companies making the inputs to the chip ecosystem, and the companies purchasing the chips.
Sadly, since the end of the Cold War, these measures have been woefully lacking, and China has been allowed to strong-arm US and allied companies into ever-more-close interactions. With every such interaction, more value and technical knowledge is transferred from the West to China.
The CPC has deployed a full panoply of strategies to advance their domestic semiconductor industry. The CPC has repeatedly demonstrated a willingness to flout treaty obligations, human-rights standards, environmental standards, and all the aspects of the international rules-based order in their pursuit of power.
One strategy that has been used again and again, and which Miller highlights, is the creation of what appear from the outside to be corporate entities, such as Huawei and ZTE. These entities are, in reality, extensions of the Chinese government.
Their agenda is indistinguishable from that of the state, their ownership is obscure at best, and they operate with massive subsidies from the government which allow them to break into key and strategic industries where they would be unable to compete on economic terms. They leverage the sharp power of the state for espionage, negotiation, and gray-zone warfare against Western companies and their personnel.
Furthermore, these “companies” benefit from an asymmetric regulatory regime. Foreign companies are permitted to sell into the China market only to the extent that they team with local firms in ways that involve mass transfers of intellectual property, while the PRC national champions have, by default, exclusive access to the local market in China.
These national champions are treated by Western governments, by and large, as economic competitors in Western markets, but they get all the advantages of being arms of the state for purposes of developing strategic and key goods for both export and domestic use.
Because of the enormous size of the China market, this asymmetric access offers decisive advantages, and has allowed Chinese companies systematically and successfully to attack a number of markets – including telecom equipment, electronics assembly, the large-scale manufacture of circuit boards, display technology, and many more.
The CPC is deeply committed to replicating this model for advanced semiconductors, and has already had considerable success in doing so for the less advanced variety.
This agenda is no secret. President Xi Jinping, in addressing the importance of core technologies – specifically semiconductors – is quoted in Chip War (page 248)as saying before an elite audience of Chinese entrepreneurs and technologists:
“We must promote strong alliances and attack strategic passes in a coordinated manner. We must assault the fortifications of core technology research and development…. We must not only call forth the assault, we must sound the call for assembly, which means that we must concentrate the most powerful forces to act together, compose shock brigades and special forces to storm the passes.”
According to Chris Miller, Xi’s audience “counted on” the use of non-economic means when seeking a competitive commercial advantage.
One interesting point that Miller makes is the high cost of building advanced fabs, and (as a separate example) of replicating the EUV machines being built by ASML. As such, he is somewhat sanguine about the danger of China duplicating these advanced capabilities; he seems to believe that the high cost will prevent them from catching up.
But Miller misses a crucial point: Building a cutting-edge semiconductor foundry process costs several tens of billions of dollars only under a very particular set of circumstances. Access to the Western ecosystem of suppliers is required. For it to cost tens of billions, the entity setting up the foundry needs to be able to buy the tools and chemicals and wafers and masks and design software at market prices.
If, on the other hand, access to all of those items is denied, then the cost to build a cutting-edge foundry would be in the trillions of dollars, since it would be necessary to replicate the entire ecosystem.
Tens of billions of dollars is a check that the CPC would be delighted to cut for advanced foundry capabilities. Trillions is not feasible. This difference is critical. Allowing China access to the Western semiconductor ecosystem cuts the cost of China standing up an independent ecosystem by a factor of perhaps 100, and makes doing so feasible.
With the China advancing autocracy and threatening its peaceful, democratic neighbors, there is no obvious policy justification for the liberal democratic regimes of the world continuing to allow China any access whatsoever to this ecosystem.
The Western response
How should the West respond to the China challenge? Miller’s book largely eschews policy recommendations, but it is easy to derive a number of lessons from the stories in his book and from our experiences in the worlds of semiconductors and geopolitics.
In the US, the ongoing debate about how to help the semiconductor industry tends toward two orthogonal schools of thought.
The first are free-traders, who believe that the most important policy questions revolve around how we can make the semiconductor and high-tech industries in the United States maximally competitive and efficient. They tend to believe that government intervention in the semiconductor industry will only create structural inefficiencies, because the people who know how to pick economic winners tend not to go into government.
They believe that less regulation will make the industry more efficient and more competitive. And in a world where chips are primarily an economic good, and our companies can compete primarily on an economic basis, this argument is sound and powerful.
The second group are those who see semiconductors primarily as a “key” or a strategic good, and who believe that strong government intervention is needed in order to ensure the long-term competitiveness of our on-shore semiconductor industry. They tend to focus on the massive subsidies and the use of state power to bolster Chinese competitors vis-à-vis Western firms. They argue that similar measures are needed in the US.
There is a third path, which is more nuanced and gets less attention, that preserves the distinctive advantages of America’s capitalist, open system for producing and fostering innovation, and allows for competition and specialization not just among firms but also among allied states.
This third path is to resurrect the Western entente of the Cold War, not just as a military alliance but also as an exclusive trade zone for high-technology goods. Free trade of high-tech goods with free people, those who share US values and interests, ought to inform the economic policies of the liberal-democratic alliance.
China, Russia, Iran, North Korea and their allies and clients should no longer be allowed access to America’s high-tech goods and services, and should be comprehensively frozen out of access to the entire semiconductor ecosystem. This most especially should include the tools to build and design semiconductor chips, the raw materials and fine chemicals needed as inputs to the factories, and the spares required for their existing factories.
The price of participation in the Western ecosystem for US allies will necessarily be a strong commitment to preventing their citizens and companies from working with PRC-controlled and allied entities in the semiconductor space.
This same regime will need to be expanded over time to other industries producing “key” and strategic goods; only by doing this will it be possible to preserve the distinct advantages of an open market among the free peoples of the world.
Companies – especially smaller ones – are generally not in any position to hedge unforeseen geopolitical risks. They go where it’s cheap to design, build and package chips, within the constraints of law – both local laws, and those of the states where they have their headquarters and primary operations.
When geopolitical realities change, it can and does create enormous economic disruption. Requiring Western semiconductor companies to bear the costs of decoupling, and endure the sudden loss of a large fraction of the market for their goods, would be enormously damaging.
The cost of keeping these companies in business and allowing them time to restructure their businesses is part of the cost of decoupling from China, and is not corporate welfare.
This is especially important in a semiconductor downturn, like the one that is starting now. Semiconductors are a highly cyclical industry, and making sanctions against China stick is much easier when there is strong demand for capacity, in a hot market, than in a downturn.
In down-cycles, contracts from firms in the PRC may be the only thing keeping a supplier of tools or chips afloat. Trying to freeze PRC firms out of the Western ecosystem at the beginning of a major semiconductor down cycle introduces the risk of irreparable short term harm to crucial Western firms.
Structuring short-term subsidies to buffer the lost revenue from rapid regulatory changes is necessary to make sure that the US doesn’t lose these firms. To preserve the competitive nature of the Western semiconductor ecosystem, any subsidies to support the cost of decoupling must be a temporary bridge, and not a long-term crutch.
Without these kinds of policy changes, there will be enormous political and lobbying pressure from semiconductor companies that sell to China, demanding exceptions and waivers from any action that drives rapid decoupling. With current regulatory strategies, companies are already responding both with political pressure and by developing products that skirt the edges of the newly introduced policies.
What exceptions should the United States allow within such a policy? Only the ones that create long-term dependency and insecurity for its adversaries. Only chips that are several generations behind the cutting edge should be sold to adversary regimes and their aligned companies, and only in the form of finished systems, rather than as bare chips.
The US and its allies should, on security grounds alone, entirely forbid the import of chips or associated systems that are made, packaged or designed in China. It should sell only end-user systems that are generations back from the state of the art, and which will cause the CPC leadership to question whether the systems on which they depend can be relied upon in a crisis.
Conclusion
China is at war with the West. This war is being fought today primarily through economic means – not because of any qualms among the leadership of the CPC about using kinetic measures, but because they believe that economic warfare is better suited to their ends at this time.
The goal of the CPC is for China to become the essential power, and to exert hegemony over the rest of the world. Any liberal democratic regime that can deny key or strategic goods to China is a threat to that agenda.
Right now, China is engaged in a long-term effort to subsume the independence of the liberal-democratic states of the world by acting outside the strictures of the rules and free-trade based international order that the United States championed in the aftermath of the Cold War.
Therefore, the US should move expeditiously to reinforce its relationships with liberal-democratic allies; allowing them to compete for government subsidies, as part of an exchange for eschewing work with PRC entities, is very desirable and was an essential part of reinforcing America’s Cold War alliances.
The recognition and reinforcement of this geopolitical reality will reinforce a liberal-democratic alliance of free peoples, while reducing authoritarian access to this entire ecosystem.
As Adam Smith noted in The Wealth of Nations, some goods and sectors enhance security, not just opulence. The US and its allies need to take a careful measure of which goods are purely economic, which are strategic, and which are key.
The US needs to construct a system where it can engage in open, capitalist competition with its allies and their corporations. And it needs to stop letting its adversaries mine America’s key and strategic industries for their own benefit, which has been to the detriment of the free peoples of the world.
Michael Hochberg
Michael Hochberg earned his PhD in applied physics from the California Institute of Technology and is president at Luminous Computing, a company building supercomputers for machine learning. He co-authored a widely used textbook on silicon photonics, and has published work in Science, Nature, National Review, The Hill, American Spectator, RealClearDefense, Fast Company, and others. More by Michael Hochberg
Leonard Hochberg
Leonard Hochberg taught at Stanford University and other institutions, was appointed a Fellow at the Hoover Institution, and co-founded Strategic Forecasting Inc (Stratfor). He has published work in Social Science History, The Journal of Interdisciplinary History, Orbis, National Review, The Hill, American Spectator, RealClearDefense, and others. He serves as the coordinator of the Mackinder Forum-US. More by Leonard Hochberg