New Chip Export Controls and the Sullivan Tech Doctrine with Kevin Wolf
How the US intends to outpace China in force-multiplying technologies
The US Commerce Department this past Friday dropped 100+ pages of new export control regulations that have the potential to reshape the future of the global semiconductor industry. Will these regulations stop China from getting below 14nm? Is that a goal even worth pursuing? And what are the trade-offs baked into taking a unilateral vs multilateral approach?
To explain what it all means, over the weekend I had on the podcast Kevin Wolf, partner at the law firm Akin Gump and former Bureau of Industry and Security (BIS) official with thirty years' experience in the field. He represents many of the firms impacted by this new regulation.
ChinaTalk is a reader-supported publication. To receive new posts and support my work, consider becoming a free or paid subscriber!
The Sullivan Tech Doctrine
Jordan Schneider: On September 16th, National Security Advisor Jake Sullivan announced a new policy on how the US should think about leadership and key technologies. He said:
Fundamentally, we believe that a select few technologies are set to play an outsized importance over the coming decade.
Computing-related technologies, biotech, and clean tech are truly “force multipliers” throughout the tech ecosystem. And leadership in each of these is a national security imperative.
On export controls, we have to revisit the longstanding premise of maintaining “relative” advantages over competitors in certain key technologies. We previously maintained a “sliding scale” approach that said we need to stay only a couple of generations ahead.
That is not the strategic environment we are in today.
Given the foundational nature of certain technologies, such as advanced logic and memory chips, we must maintain as large of a lead as possible.
Kevin, where were you when you heard the Jake Sullivan speech?
Kevin Wolf: That's actually a good way of putting it, because it is a transformational document in terms of the thinking and the philosophy of what national security means in the context of export controls. During the Cold War, the rules governing the movement of commodities, technology, software, and sometimes services between the US and the Soviet Union/Eastern Bloc had a larger strategic element to them. What changed, with the end of the Cold War, the rise of the current multilateral regime system, and the Wassenaar arrangement being created in 1996, was the focus of export controls on the inherent nature of commodity software and technology that were identifiably related to the development, production or use of either weapons of mass destruction or their means of delivery or conventional weapons.
After 9/11 some anti-terrorism aspects were added in, but they were not strategic controls. They were overtly not focused on specific countries. The laws of the allies are governed almost exclusively by the structure of lists and specific items that are identified for one of those WMD or conventional-weapons applications.
What changed, frankly, is China's technology acquisition policies, [including] acquiring commercial technologies in order to help modernize its military, the use of a state policy of achieving strategic economic dominance in key sectors for the economy that are also critical for the modernization of a military, [and] the use of commercial items to commit human rights abuses. There [needed] to be a fundamentally new way of thinking about what national security means with respect to China's strategic economic ambitions, civil-military fusion policies, human rights abuses, etc. With the Export Control Reform Act, which passed in August of 2018, there was a mandate to the agencies to start identifying the types of emerging and foundational technologies that didn't fit within the traditional concept of what export control should be for.
Congress didn't define what national security meant outside their traditional context. The Trump administration issued different rules and made some specific edits within this broader concept of “emerging and foundational”, but never really articulated a clear vision of what national security with respect to China meant when you get outside the traditional objectives.
The Biden administration hadn't either, until the Jake Sullivan speech. He did exactly what you just said: that we must think about force-multiplying technologies, high-end compute capabilities, advanced node semiconductor capabilities, quantum and AI applications… Not in the traditional sense of export controls, because each one of those items has inherent characteristics that are identifiable with respect to WMD or weapons in the classical sense of export controls. But we need to think, as he said quite clearly several times, about controls on a strategic basis, because those are force multipliers for other items of great significance: for purposes of competitive economy and advancing and modernizing a military.
Jordan Schneider: When you see a strategy like that, you are then waiting on pins and needles for how it's going to manifest. Kevin, what's in these regs?
Kevin Wolf: The short answer is: any of the inputs, tools, technology or software, whether US- or foreign-made, and any services in support of basically all the things that in Jake Sullivan's speech are now subject to some form of regulation. By regulation, I mean a licensed permission from the US government is required. It articulates, at the beginning of the rule, a much more focused and lengthy description of the general principles than were laid out in the Jake Sullivan speech.
The policy objective, [which is] the problem to be solved, is that China has mobilized vast resources to support its defense modernization and civil-military fusion strategy by targeting and affecting advanced computing and advanced node semiconductor manufacturing.
Strategic Motivations for Export Control
Jordan Schneider: Let's talk about the strategic thesis behind this, which is that 14nm and below is scary from a US national security perspective. The document makes the argument that this is because you can design missiles better and have awesome hypersonics that you couldn't if you weren't making or importing these chips.
I'm sure another reading of this is America just wants to be on the cutting technological edge for a long-term economic edge. What is the goal as BIS is defining it, and do you think it is justified?
Kevin Wolf: Between imposing controls on items that are critical to making advanced military items and WMD, etc., and a goal to have as a purely economic objective, there's nothing in this rule that indicates that the latter is the motivation. The overt articulation is that over time, you need advanced computing capabilities in order to modernize your military and make more advanced weapons. What's slightly different than what the traditional type of controls has been is the relationship between the specific item and the policy concern.
The classical view of export controls up until now is that there was much more of an identified relationship between WMD, the military item, and the actual item going into it. Is it radiation-hardened? Does it operate at a particular capability? Does it have a frequency that's used in military radar?
This basically goes back several layers before in the food chain: in order to get to the point for all sorts of other applications to make modern military items, you need to have an ecosystem that has advanced computing capabilities. For those things to function, you need advanced node semiconductors. For advanced node semiconductors to exist, you need certain types of production equipment and software. For those things to be put together, you need services of US persons behind it.
There's no doubt that the overwhelming number of applications for everything described in this rule are commercial. I want to emphasize: the controls that have been in place for a very long time already regulate, and in fact completely embargo for China, any type of item, whether it's sophisticated or simple, that's in any way modified, changed, altered, intended, or designed for a military application, satellite, or missile application.
[The items that these new regulations address] are items that don't have, by definition, any particular modifications, and are inherently commercial in their main application. But what Jake Sullivan said, and what the Commerce Department said, is that those capabilities are foundational to creating the military and other items of concern, even if these items themselves have not been modified for military applications.
Jordan Schneider: An exoscale computer can run meteorology and climate models. They can also run ballistic tests. I guess the argument is that if you sell these chips, you don't know where they're going, and because of the national security implications of leading node technology you end up not being able to play the game that export controls classically played.
What do you think are the good and bad arguments for going this route?
Kevin Wolf: I do think it's directly responsive to this need for a new way of thinking about national security with respect to the specific, unique, and significant threats that China presents. In my personal view, I think it's direct and responsive.
In [its] scope and the fact that the effective date was immediate for Friday, there was no opportunity for companies to think about what the implications, unintended consequences, or impacts on the supply chain would be. In that sense, it's awkward. But I also see the reason why they went forward without notice and comment. They overtly say this is not technically an emerging and foundational technology control even though it's all about emerging and foundational technologies, because that part of the Export Control Reform Act, requires public notice and comment for any rules that are under that new authority.
What they said is that the issue with China is an emergency.
There was probably also a concern about stockpiling. If they identified too loudly what their plans were, then companies would send ahead items that were thought to be controlled. That's a constant dilemma with export controls.
Jordan Schneider: Can you talk about how much work must have gone into writing these?
Kevin Wolf: This is an extraordinary amount of work. Although complex, it's really quite thoughtful in the way in which it, from an export control regulatory perspective, addresses the instructions given to them by the National Security Advisor.
Although comprehensive, it's really tailored to address the very specific high-performance computing and advanced notice semiconductor concerns. It's not a broad decoupling agenda.
This is clearly intended to not try to affect global supply chains of mature node semiconductors and other items that are used for widely available commercial items. Thanks and kudos to the leadership and staff at BIS for getting out a really thoughtful, but clearly difficult, rule.
They acknowledge upfront that there will be mistakes. When you're moving on something this big, it's not just going through one person's brain. This also has to go around and get inter-agency clearance. Nothing publishes out of BIS unless at least the Departments of Commerce, Defense, State, and Energy have also reviewed, cleared, and commented as well. The inter-agency drafting process can sometimes lead to inconsistent outcomes as part of the clearance process.
A Tour of the Regs
Jordan Schneider: Fair enough. Kevin, what's in the regs?
Kevin Wolf: These are all unilateral controls: notwithstanding the normal structure of export controls being done by allies together at the same time with the same lists, these are US-only controls. They're extra-territorial in large part, in the sense that they apply outside the United States even to foreign-made items if they use US tools or technology.
The first thing is the creation of new controls on things previously uncontrolled: high-performance computing, integrated circuits, graphic processing units primarily, and the computers and other electronic systems that contain them. Also, you have to go through and determine whether your chip is above or below certain technical characteristics.
These are China-only controls. This is the other thing that's novel about it: most controls apply to the whole world, or minus allies, because it's been focused on the inherent nature of the item. These are unique in that there's a whole new reason for control: regional stability.
The second clump of controls are items destined for supercomputer/semiconductor development or production end-use in China. Specifically, [whether you are] sending an item of any sort to a semiconductor fabrication facility in China that fabricates integrated circuits that meet one of three criteria: logic, dynamic random-access memory (DRAM), and NAND flash memory. If you are sending something to a facility that fabricates logic that's either nonplanar or has a smallest feature size of 16 nanometers or smaller, that export is now controlled. If you are sending items to a semiconductor fabrication facility for use in producing a DRAM with a smallest feature size of 18 nanometers [or smaller], or if you’re exporting something for use in producing or developing NAND, that is now controlled as well.
If you, the US person, know your thing is going for one of these uses, you have an obligation. But if you don't know what the feature size or technology node of the thing being produced in China [is], then you also have obligations to either get a license or find out that it's not for one of these three types of semiconductors. There's also a similar semiconductor computer control if you're sending items for the production or development of certain types of supercomputers.
There is an expansion of the foreign direct product rule. This is a concept that was originally created in 1959, then I created two new foreign direct product rules dealing with satellites and military items. The Trump administration created the grandfather of these for the Huawei-specific controls in August of 2020, and the Biden team used the foreign direct product rule with respect to key parts of the Russia sanctions.
This creates three new foreign direct product rules. One, for foreign-made items that are used for the production of these types of supercomputers that have a technical threshold. [Two], for advanced GPUs and other advanced compute capability if you're using foreign-made items to produce them. [The third] expands the unverified list, which is not the imposition of new controls. It's actually a rule that I added to put pressure on China and Russia, in order to allow US government officials to do onsite verification and inspections of Chinese or Russian facilities. If they didn't, or if the company couldn't be determined to be a bona fide company, then they were put on [the unverified list]. It requires a lot more certifications in order to export, and you can't use license exceptions to ship. This rule added quite a few well-known companies to the unverified list. If the US government's concerns regarding the company, such as site visits or onsite inspections, are not satisfied in 60 days (that is, if the Chinese government doesn't allow access to these companies), then they're going to be moved to the entity list, which does have licensing obligations and does impose broad sweeping bands on the export of anything from the United States to those companies.
There are no new entities added to the entity list, but they have been determined to be involved in the supercomputing and high-performance computing industries of concern. Therefore, foreign-made items produced with US technology and software are now subject to licensing requirements, or will be subject to licensing requirements if destined to one of those identified identities.
There were also 11 specific types of tools (semiconductor production equipment, deposition equipment, I think generally) that historically had not been subject to export controls, but had been identified as being specific to producing advanced node semiconductors. Those specific items are now subject to export controls from the US. Also, there's a control now on exporting parts and components, or frankly any item, to China from the United States for use in developing or producing semiconductor production equipment of almost any sort in China. There's a temporary general license created with respect to the GPUs and other items that are subject to the new controls, allowing for another six months or so of continued production and transfer so long as the end use is outside of China. There was a lot of repair or warehousing work for GPUs going into commercial applications out of China that were caught up by this new rule. Commerce gave those companies time to figure out a different way of dealing with this because their customers and end uses were outside of China.
Jordan Schneider: Let's start with the last thing first. This 60-day unverified list: is there a potential that the impact of this could actually be a lot bigger than what we currently see, if some of these big names on the unverified entity list end up graduating to full entity-list status?
Kevin Wolf: It's really up to the Chinese government. If they allow access, give the information that the Commerce Department is seeking, and allow the two agencies in China that do inspections and site visits to do their jobs…
Jordan Schneider: Whoa, wait, two? There are, like, 30 firms on there!
Kevin Wolf: Yeah, I know, but no other country has any, and the Commerce Department is the only export control organization on the planet with its own enforcement agents who are subject matter experts in export controls.
It's a fascinating question: what's going to happen in the next couple of months with those, and will there be cooperation allowing visibility and access in the provision of information? The Commerce Department has clearly indicated [that] not cooperating with requests under the unverified list is a standard for being added to the entity list.
Jordan Schneider: Thoughts on the decision to do nationwide as opposed to company-specific regulations on these topics?
Kevin Wolf: Export controls have three legs of a stool. You've got lists of items that are controlled for a whole country. You've got regulations on specific end-users, just like the unverified and entity lists, which are tailored regardless of whether the item is listed or not. And then you have controls on end uses: even if the end user and the specific item isn't identified, there's a kind of end use that should be regulated. When an item is so widely available commercially that it can't practically be controlled, you add entities to the entity list or, as this rule does, take a page out of the Enhanced Proliferation Control Initiative book and create end-use controls, even for unlisted, uncontrolled items to unlisted entities if there are going to be for the development of production of missiles or chemical, biological, or nuclear weapons.
With respect to China as a country, it's clear that the policy objectives pertain to a state policy of using commercial items to help modernize its military — the civil-military fusion policy. In that sense it has to be China-wide, but to the extent that there are more detailed and specific controls that can be imposed on particular activities or end users, then that will be the approach.
Jordan Schneider: What am I, as a US person, not allowed to do now?
Kevin Wolf: Effective on October 12th is a requirement that US persons (whether a company, an individual, or basically anybody acting on their behalf), if they are providing support in terms of the shipment, transmission, facilitation (of shipment, transfer, or transmission), or servicing of foreign-made items at a facility that produces advanced-node NAND, logic, or memory, those acts are now controlled events. Whether it's an individual citizen working abroad for a foreign company, or a US company and its foreign agents working overseas, that will, on October 12th, be a controlled event requiring a license.
Enforcing the Regs
Jordan Schneider: Does this make it impossible for a US firm to make a little hive of non-Americans to sell this sort of stuff? To what extent is this new, novel, creative, and an escalation as well?
Kevin Wolf: It's really quite creative and novel, because it's a concept that, for dual-use purposes, had only been limited historically to controlling WMD applications. Even if the underlying technology is not controlled, or in this case explicitly foreign-origin, if a US person is providing such support, that's a controlled event. If a US company would somehow make arrangements to set up people from outside the US to do this, then that would be prohibited. Even though the people, activities, and technology are wholly foreign, it's the act of the US company or individual citizen in facilitating the transmission or the shipment of these items for one of these advanced node or supercomputer applications that would be prohibited.
It has no impact on non-US persons, or on companies outside the US that don't make their stuff with US technology and have foreign person employees. For the competitors of the US companies outside the US that have been able to design out US technology, citizens, or content, this has no effect.
Jordan Schneider: The theory, presumably, is that those firms with these technologies do not exist and that this is not something that Chinese firms can substitute out US technology for.
Kevin Wolf: Well, I don't know. In the short and medium term, it's really quite effective because US tools, software, and equipment are quite commonly and widely used throughout the world in producing the types of semiconductors and tools.
Say you're a non-US company that didn't want to have US law apply: depending upon your product, it’ll take many years for other items or be done overnight.
Some things are very hard to design out and other things can be swapped out immediately. There are foreign competitors for most of the companies that are affected. Unless there are multilateral or plurilateral controls that their governments agree to, [they] will be able to engage in all of exactly the same acts described in this rule without the need for any authorization.
Jordan Schneider: I guess the theory of the case is that tooling is an incredibly tight market, and by doing this sort of work you'll be cutting yourself off from the rest of the global technological ecosystem. The sort of investment that it will take for a firm in Japan or the Netherlands to be able to capture the Chinese market share would just not be worth the squeeze.
Kevin Wolf: I've been doing export controls for 30+ years now and started doing it as a young attorney even before the Wassenaar was created. I've got a pretty long history in my head about the ebbs and flows of export control policy debates.
There's one constant that's reflected in the Export Control Act quite overtly, [which] is that multilateral controls tend to be more effective and unilateral controls tend to not be. Now, I'm not saying that unilateral controls are always bad, always wrong, or always ineffective. Sometimes unilateral controls are critical, even if less effective, to project American disapproval of a certain behavior; that's why human rights controls historically have been unilateral. I see these controls and locking-in structures over decades. That's why I'm a huge advocate for multilateral controls, and in particular a whole new regime of a smaller group of close allies, not to replace the four existing regimes but to be in addition to the four regimes, to regulate the types of items that are in this rule for non-traditional national security reasons.
In the short term, [these regulations] will absolutely be effective and have the desired effect of the US government, but over a much longer period of time non-US companies — if they're not subject to the same controls in the producer nations — will fill the gap.
I think it's really going to be a function of how persuasively, and [with] how much evidence, the US government presents at very senior levels to the allied governments about the non-classical threat that is China, with respect to the need to regulate inherently commercial items. Other countries, frankly, are more exposed than the US is with respect to their dependency on sales to China for their economic livelihood. But I do think that, and this I laid out in an article that I published earlier in the year, if there's an appeal to core common principles of techno-democracies [like] freedom of speech, the advancement of democracy, and protection of human rights, I'm confident that the US government can be successful. If they provide the evidence, like they did at the beginning of the preamble, about how all of these things are not themselves directly used in military items, but are force multipliers to the development of sensitive things, I do believe that a small group of allies eventually will come on board, notwithstanding what is no doubt intense lobbying by companies in their countries.
Jordan Schneider: I'm curious if the unilateral vs. multilateral trade-off is different when you're playing on these global emerging technologies, which already are trillion-dollar industries, as opposed to the more niche missile design, engine turbines, satellites, or what have you.
Kevin Wolf: One of the virtues of working both with industry and government [is] I get to see both the government and the industry’s concerns. These high-tech industries have very high percentages of their income devoted to R&D — sometimes 20-40% — whereas in the military space it's 1-3%. You can only get that massive amount of income for R&D if you have lots of international sales. Those international sales are, then, critical to fund the R&D to out-innovate your foreign competition, a significant percentage of which comes from sales to the very large market that is China historically.
What the government has decided, in this case, is not to try to affect that which is for that trailing edge. This rule, although comprehensive and big, is really focused at the high-end chip. The intent is for advanced node semiconductors of different types, supercomputers, and advanced AI applications. They deliberately are going out of the way to not affect design, development, production, or use of things at the trailing edge or mature note semiconductors, where frankly that's where a lot of money still is for all the consumer applications like cell phones.
Getting the Allies Onboard
Jordan Schneider: It seems pretty impossible for a US firm to sell semiconductor manufacturing equipments to China nowadays. But chips are chips, right? Is the golden triangle now going to be this underground highway to smuggle in tens of thousands of Nvidia GPUs? Is controlling the import of commercially available chips that are sold in the hundreds of thousands possible?
Kevin Wolf: Enforcing violations involving foreign-made commercial items that aren't subject to any other country's laws, that are going into exclusively commercial items, that are extraordinarily hard to identify, and that are sub-components going into other things is no doubt really, really difficult, but just because something is hard to enforce doesn't mean you don't do it.
The solution lies in cooperation with allies. Just like the sanctions and comprehensive export controls against Russia after the invasion of Ukraine demonstrated, when allies work together and share resources on enforcement, tips, leads, and intel, enforcement of these controls can be far more effective.
Goal number one for the enforcement side is to get other allies on board as quickly as possible. Goal number two, which I think the Commerce Department is already going down, is aggressive, overt leadership about the seriousness of export patrols and getting the word out that if you violate these rules, we will punish you quite severely. Goal number three is to keep funding and training a large cadre of enforcement agents in order to do investigations and track down violations on the front lines of companies and their compliance programs. This is what I do for a living: what compliance attorneys do is help companies set up systems in order to be able to identify and spot potential violations, and make it so that shipments or transfers that require license don’t occur without permission.
You need massive engagement of industry, resources, compliance and licensing officials, systems, and procedures in order to be able to identify the controlled items and activities and stop them. It's hard, but you don't let the difficulty of enforcement stop something that warrants being done for broader national security reasons.
There are those that say you shouldn't act until the allies are on board, because if you go first and you go alone, then the allied companies are going to say, “Woohoo, they've just opened up a gap in a market for us to fill.” There's absolutely truth to that; I see firsthand in my practice non-US companies legally doing things that US companies are not allowed to do. But I do see from a broader diplomatic perspective the need to signal how serious the US government is, which this rule does in spades, in order to get a really serious discussion going with the allies about the need for them to change their rules to achieve common national security objectives.
We're all in uncharted territory here and I don't know what the right answer is, but I do know that over the long term, if the effort to get the allies on board fails, the incentive for non-US companies to produce, develop, sell, and do R&D outside the United States will just accelerate. Over a very long period of time, they will achieve advantages that they wouldn't have but for the unilateral controls.
Jordan Schneider: I guess the bet, then, is that America will be running faster, and you'll have a gap that will be expanding, not contracting, over the long run.
Kevin Wolf: Absolutely fair point. The other big difference with what's happened in the last couple of years is that there's the keep-away strategy, which is the export control strategy, and then there's the run-faster strategy, which is the industrial policy: the CHIPS Act, subsidization of STEM education, and subsidization of production domestically to counter Chinese production. The Jake Sullivan speech laid out quite clearly that we can't just do a keep-away strategy. It absolutely requires the run-faster strategy as well.
You can't just export-control yourself into a healthy economy, given the fungibility of technology and the capability of smart people outside the United States.
Jordan Schneider: My hot take is that I don't think this is going to be taken lying down. There have just been so many different small cuts from a micro-electronics perspective over the past five years, and I found it really remarkable that the Chinese government’s biggest response was after the Huawei CFO was arrested [when] two Canadians were deprived of their liberty for a number of years. Really, nothing directly impacting American business has fallen out in a dramatic way.
The Human Rights Factor
Human rights, Kevin. This was kind of missing here, wasn't it?
Kevin Wolf: Well, yes and no. There is a reference at the beginning that part of the motive for these controls is that the advanced node items can be used for human rights violations.
Now remember, all these are going to be, by definition, unilateral, since there is no multilateral export control arrangement that has human rights as part of its mandate. Another interesting aspect of the standards for what could cause an entity to be added to the entity list: they didn't overtly say [that] engaging in violations of human rights is a per se rule for being added to the list.
Now, they can still do that — there's the legal authority to do it — but I thought it might be called out more. That could be coming later in other rules; it just wasn't highlighted in either of the two rules that were published on Friday.
ChinaTalk is a reader-supported publication. To receive new posts and support my work, consider becoming a free or paid subscriber!