The latest U.S. chip export controls: Q&A with Rhodium’s Reva Goujon

Last week, the U.S. Department of Commerce unveiled export controls aimed at China’s development of its advanced semiconductor industry, expanding on sweeping restrictions that were put into place on October 7, 2022. The new rules close loopholes in the previous regulations, changing thresholds for what kinds of chips and chipmaking equipment can be shipped to China, and adding several Chinese firms to the U.S. entity list, meaning American firms must apply for export licenses to sell to the firms.

The new regulations are packed with technical terminology, and run hundreds of pages long. We sat down with Reva Goujon, a director at Rhodium Group, who helped explain what the new rules will mean for companies in the semiconductor manufacturing ecosystem, and for China’s chip industry more generally. The conversation has been edited for length and clarity.

—Eduardo Jaramillo

So before we get into the really meaty questions, I’m going to ask you to oversimplify a very complex issue. If you had to sum up these updates with a topline takeaway, in a few sentences, how would you do it?

In the lead up to the new October 17 controls, this was framed as a patching up of the October 7, [2022] controls — updates, after a year of reflection. But this goes well beyond a patch-up exercise. There is a lot of punch packed in these 400-plus pages of new regs, and I think that reflects the perceived opportunity that U.S. officials view here; having an anniversary date to update rules also allows you to introduce novel provisions, extend your reach into certain areas, and steer partners in a desired direction. And so you see a lot packed in overall.

A little over a month before the new restrictions were published, Huawei released a new smartphone, the Mate 60 Pro, which included a 7-nm process chip, the Kirin 9000s, made by Semiconductor Manufacturing International Corporation (SMIC).

I’m curious to what extent you think this phone’s release pushed U.S. officials to beef up these controls, and whether the updated controls do a better job of keeping SMIC from advancing to those lower process nodes?

The timing of Huawei’s breakthrough in the Mate 60 Pro, and drawing attention to SMIC’s capabilities in manufacturing a 7-nm class system on a chip on a presumably commercial scale, given how many of these handsets Huawei is claiming to sell, that’s notable. Now, you don’t write 400 pages worth of regs overnight. Obviously, this is a long-running process and likely these regulations were nearly fully baked when that curveball revelation came through on Huawei, and so I wouldn’t see the October 17 controls as highly reactive to the Huawei Mate 60 reveal and the capabilities that underlie that, but you do see some interesting points in the rules that address the notion of SMIC advancing its manufacturing capabilities in spite of the controls. And so that’s where of course when you get into the equipment roles and where you see the capturing of a broader swath of older technologies to include extending into areas where ASML may have thought it still had a green light and now that’s in question. That is a way of potentially slowing down SMIC’s capabilities because when we learned of the Kirin 9000s chip, there was an immediate question of, whether this was achieved via stockpiling of the necessary equipment to produce those chips. Then what happens when controls tighten to the point where spare parts, and servicing of those machines becomes really difficult. And that then raises the question of what is “good enough,” so to speak, from a U.S. regulatory perspective and designing these controls.

I don’t think the U.S. is under any illusion that China will not be able to produce advanced chips under constraints. There’s a question of whether China can do so at commercial scale. But [it’s a question of if it’s] “good enough” from a U.S. regulatory perspective, to significantly slow down China’s capabilities and make it more inefficient through these rules.

Will SMIC be able to succeed in advancing to lower process nodes? Yes, with time, it’s going to find workarounds, and it’s obviously focusing its resources on trying to become as autonomous as it can…but [these new regulations] significantly [do] slow down China’s chipmaking capabilities and again, that may be the key objective from a U.S. national security point of view.

So in terms of the companies that are going to be affected I’m curious if you can get into who you think are going to lose out in both the U.S. and China because of these new updated controls.

Yeah, so when it comes to the U.S. players, immediate focus goes to those impacted by the new rule that they’ve broken out on advanced computing. And so when you’re looking at Nvidia, AMD, Intel — any company that has spent the past year trying to develop a “in China, for China” chip that would be compliant with U.S. regulations, at least the October 7 version, those are obviously the hardest hit. These rules are basically designed to say, look, there is a new threshold we’re setting for high performance compute chips, but there’s also a gray zone that we’re defining, of chips that fall just below that threshold that are now going to require notification. So the signal from that is really a message to these companies to not invest their resources, their energy, into trying to develop “in China, for China” advanced compute chips [to get around the new rules], and focus your efforts elsewhere, essentially. And so, yeah, I think that deterrence factor is a very strong feature in the controls.

How about the toolmakers — the U.S. companies that sell semiconductor manufacturing equipment to chip foundries located in China?

Now, U.S. toolmakers, they were already heavily restricted by the October 7 controls, so AMD, KLA, Lam Research on the toolmaking side, and then Cadence and Synopsys on the [chip design] software side. This is where you see sealing up restrictions in certain areas. But overall, I think that those affected companies have already been feeling it over the past year. There is some interesting nuance in terms of the semiconductor manufacturing equipment rule that puts the emphasis on front end manufacturing… there seems to be less of a concern on back end processes, so when it comes to outsourced semiconductor assembly and test, there are some carve-outs there, which is interesting.

Also to bear in mind [are] downstream effects on players like cloud service providers [and] data centers, and then lots of questions that emerge from these rules for other technology providers that are in the leading edge space. And those that are just reliant on advanced chips, even for consumer applications, and rely on the China market for a good portion of their revenue. So even though there is some language in the rules around a licensing process and exceptions for consumer applications, that language is still very vague, and I think there’s still quite a bit of concern of how those rules could tighten up over time.

So that’s on the U.S. side. How do these regulations look for Chinese firms?

On the Chinese side of course it’s the usual suspects in terms of the chip makers themselves [who are going to take a hit]. We’re looking at SMIC and [Hua Hong Semiconductor Limited], YMTC [Yangtze Memory Technologies Co.], CXMT [ChangXin Memory Technologies], etc.

On the entity listings, that’s where clearly you see [Biren Technology] and [Moore Threads Technology] being singled out and [I’m] not necessarily surprised given that these are China’s premier [graphics processing units] makers and the strongest competitors to U.S. tech champions in this space like Nvidia. This is where, as the U.S. is trying to manage expectations with industry and try to get buy-in from the private sector on these regulations…the U.S. can argue that it is “leveling the playing field” as well in targeting companies like Biren and those that are already are likely to receive substantial subsidies from the Chinese government.

Dutch photolithography equipment maker ASML Holdings has been thrown around as a potential loser here, but there seem to be some loopholes. Can you sketch out how you think ASML and its business might be affected here?

I think it will be greatly impacted by this update. In the lead-up to the new rules, there was a lot of framing of this as harmonization with U.S. partner controls, following the Japanese and Dutch export controls over the past year. This goes beyond harmonization from my read, where when you kind of get down into the nitty gritty of it and the technical specs of how the Dutch wrote their export controls and where the U.S. has adjusted its own, the U.S. regulations appear to capture older DUV technology that earlier would not have been under restriction. So this I think goes beyond what the Netherlands and certainly these companies, ASML in particular, [were] expecting…So there’s some tough partner discussions ahead, but there’s some interesting language and the regulations that effectively say that licensing will be provided on presumption of approval, and it will be a streamlined process. But there has to be an equivalence of export control regulations by other countries in order to have the benefits of a streamlined U.S. regulatory approval process. And so that’s a way of the U.S. providing a gesture to partners, that we’re not trying to make this really difficult for you but at the same time, the U.S. is setting the standard of what partner controls need to mirror.

One detail you mentioned was ASML’s servicing of things like a deep ultraviolet (DUV) lithography machine. I’ve also read in some places that China is developing its own capabilities to service those machines without foreign help. Do you have any thoughts on the servicing question? If the U.S. or the Netherlands cut off the servicing, would that be really important? Can China service those itself if it really needs to?

Again, over time, I’m sure China will find ways to get by, but it will likely be very difficult just given the sheer complexity of these machines. There’s so many layers to this, where you need engineers in person to work these machines, you need software updates… I think this is an area where China’s trying to really build up its capabilities through Shanghai Microelectronics Equipment [SMEE], NAURA, and other equipment providers. But the servicing aspect of this, I think is quite critical.

There’s also new updates on restrictions on what U.S. nationals can do under these regulations. Can you explain if there are major changes to what U.S. persons are allowed to do with respect to China?

This one did seem more of a cleanup exercise in clarifying things for people involved in administrative or clerical activities, it doesn’t capture that role, and it shouldn’t capture that.

At the same time you see…there’s a strong defense of the person’s activities rule…I’m paraphrasing here, but it was basically along the lines of saying that if there is a concern of people changing citizenship in order to avoid getting ensnared in these rules, the [Commerce Department] is going to bet on the idea that it’s a high price to sacrifice U.S. citizenship for the sake of trying to provide semiconductor manufacturing capabilities.

[There’s] not just the defense of applying the U.S. persons rule, but acknowledgement that this can encourage some companies to exclude U.S. persons from certain activities and from certain commercial and leadership opportunities. And so I think that’s still a reality of where companies are going to be looking at U.S. persons and U.S. origin technologies as a vulnerability. But at the same time, it’s going to be really hard to design out U.S. origin technologies as well as U.S. persons linkages to those technologies.

Any final thoughts, or anything else you’d like to mention that we haven’t touched on already?

I mean, there’s a lot packed in here. The other thing just to bear in mind is on the question of whether these rules would encompass infrastructure as a service in covering chips used by cloud service providers. There’s a seedling of a rule contained in these regulations, where the focus for now is on Chinese cloud service providers and data centers….[and] the scope right now is very focused on Chinese ownership. But then for public comment, there’s a question of how to develop rules around infrastructure as a service to address the perceived risk of Chinese entities accessing [U.S. cloud service providers’] high performance compute capacity from overseas. And that is something that would obviously ensnare leading U.S.cloud service providers potentially.

The concern that I’ve raised around this is that once you have regulations that creep into the cloud space, that could set off a chain of retribution, where reciprocal Chinese response could also affect and limit U.S. cloud service providers that are already operating under pretty tight conditions in China. But when you think about the development of technology platforms by major companies that need interoperability in China and outside of China, choosing a cloud service provider that can do that for you without getting ensnared in these regulatory entanglements is getting a lot harder.