Clean Investment at the Community Level

The Inflation Reduction Act (IRA) contains multiple provisions that seek to incentivize clean energy investment in specific communities, such as low-income communities or communities that have traditionally produced fossil fuels (i.e., energy communities). These incentives take the form of additional tax credits and priority access to loans and grants. The Clean Investment Monitor’s (CIM) comprehensive facility-level investment data is uniquely suited to assess the extent to which actual—not just announced—investment is indeed flowing into these communities as legislators intended. In this note, we provide an overview of actual community-level investment trends in the first year following the IRA’s passage.

We find that these communities are successfully attracting an outsized share of national clean investment (Figure 1). The share of clean investment occurring in energy communities is nearly double the share of the national population living in these communities (36.8% versus 18.6%). The share of clean investment occurring in disadvantaged communities is considerably higher than the share of the national population residing there as well (44.5% versus 32.7%). Actual investment in low-income communities was also positively disproportionate to the portion of the national population residing therein (40.5% versus 38.5%).


It’s still too early to robustly evaluate the extent to which community-focused incentives in the IRA are shaping geographic investment trends, given that the implementing regulations for many of these incentives were only recently completed. Noting that 34.7% of actual investment in the year following the IRA came from projects announced after the IRA, we expect these trends will likely extend as more post-IRA project announcements turn to actual investment. We will continue to track community-level investment trends and publish our findings and underlying data on

Read the full reportRead the full report Explore the Clean Investment MonitorExplore the Clean Investment Monitor

All In: US Places a Big Bet with October 17 Controls

The United States is asserting its role as the de facto regulator of the global semiconductor industry in a new tranche of rules released October 17. The new package of rules appears designed to:

  • Deter the world’s leading semiconductor firms from developing high-performance chips for the China market;
  • Preempt circumvention by expanding the country scope for restrictions on advanced chips and semiconductor manufacturing equipment;
  • Further constrain China’s ability to rely on older deep ultraviolet (DUV) immersion lithography equipment to produce advanced chips; and
  • Slow down China’s cutting-edge AI chip design firms on national security grounds.

The 400-plus-page package of regulations takes care this time to lay out justifications for the controls and exemptions for US partners and certain “consumer-grade” chips. But the rules sustain the spirit of the original October 7, 2022 controls in their willingness to assert US extraterritoriality in novel ways, preempt technological advancements, and break new ground—for example, by inching toward regulating chips-as-a-service.

US policymakers are placing a big bet with these new rules. In designing the controls, the US is making an implicit assumption that industry and partner resistance will eventually give way to alignment with US regulations. The US is also wagering that time and innovation are on its side, even as China and others try to accelerate the substitution of US-origin technologies. Most of all, the US administration is assuming that tightening US-led tech controls on an economically insecure China amid an intensifying cycle of retribution can be managed with diplomatic guardrails.


Packed with punch

The October 17 package of export controls on AI chips, supercomputing, and semiconductor manufacturing equipment exceeds the ambition of the original October 7, 2022, semiconductor controls to constrain China’s ability to manufacture advanced chips and access high-performance computing power. As is typical with such dense regulations, tiny details buried in the text can carry a big punch.

Table 1

Don’t even bother

The US is sending a clear signal to the world’s cutting-edge chip developers: don’t even bother putting time and resources into developing a high-performance computing chip for the China market. The rules will catch up.

Indeed, after watching Nvidia and Intel release cut-down versions of their deep-learning AI processors within months of the original October 7 controls, Nvidia’s A800 and H800 graphics processing units (GPUs) and Intel’s Gaudi2 system-on-chip (SoC) appear to now be banned for all Chinese entities and their subsidiaries. AMD had been debating developing a variant of its MI250 and MI300 GPUs for the China market, but those plans have now been short-circuited.

Agility is not typically a word used to describe a regulatory agency. In this case, however, the US Commerce Bureau of Industry and Security (BIS) recognized the imperfection of its original performance thresholds and iterated on them substantially within a year.

BIS focused on two key factors in setting the new parameters for high-performance compute chips under export control classification ECCN 3A090: (1) what is the raw compute power of the chip, and (2) what is the compute power in relation to its physical size?

In the new rules, BIS slightly tweaked the existing October 7, 2022 formula for determining raw compute power (now called “total processing performance,”1 or TPP). If 1) the TPP of an AI accelerator is at or above 4800 or 2) the TPP is at or above 1600 but the chip’s performance density (see below) is at or above 5.92, it can only be sold to China after acquiring an export license on presumption of denial.

With the newly introduced performance density parameter, BIS is examining the compute power relative to the size of the chip. The point is to discourage AI chip companies from combining multiple smaller chips to achieve comparable compute power. Performance density (PD) measures a chip’s compute power per square millimeter (TPP divided by the die area). In BIS’s own words, “a performance density parameter prevents the workaround of simply purchasing a larger number of smaller datacenter AI chips which, if combined, would be equally powerful as restricted chips.” BIS set the performance density threshold at 5.92, just slightly below the performance density of Nvidia’s A100 chip at 6.04 (see Table 1). As a result, if a chip’s TPP falls below 4800 but its performance density is above 5.92, the chip will be restricted under ECCN 3A090.a.

Notably, BIS has added another export control classification, ECCN 3A090.b, to control chips that fall just below the TPP and PD threshold. Chips that meet the 3A090.b requirements are in a grey zone: notification is required for export to China, Macao, and other destinations at higher risk of leaking controlled technology to China. Any high-end chipmaker trying to sell to China would have to calculate the risk of BIS updating the TPP and/or PD metrics after having some time to observe industry workarounds.

Table 2

Notably, the October 17 rules focus on integrated circuits (ICs) that are “designed or marketed for use in data centers” to differentiate from “consumer-grade ICs” with AI capabilities that may be eligible for exemption. The theory of harm that BIS offers is that advanced computing chips that meet the new performance threshold and are used in data centers can be used to power AI frontier models with potential military applications, including unmanned intelligent combat systems, enhanced battlefield “situational awareness and decision-making,” and cyber warfare.

On these national security grounds, the US has not only tightened high-end chip access to China but has also directly targeted China’s premier AI chip designers: Biren and Moore Threads. These firms and their subsidiaries have been added to the BIS Entity List with a Footnote 4 Foreign Direct Product Rule designation, barring them from access to US-origin software and technologies. Even if Biren and Moore Threads (along with other Chinese AI chip designers who could land on the BIS Entity List for similar reasons) develop their own software stacks to compete with Nvidia’s powerful CUDA ecosystem, for example, they will still need leading-edge foreign fabs like TSMC to manufacture their chips. That supply chain node is now in jeopardy. US AI chip designers being told to keep out of the China market may at least take some solace in the fact that their biggest Chinese competitors are also getting pummeled by the new controls.

All eyes on SMIC (and ASML)

That is, unless China’s leading foundries SMIC and Hua Hong can advance their own manufacturing capabilities to produce indigenously designed advanced chips. But here again, the US has dealt a heavy blow in the updated rule for semiconductor manufacturing equipment.

In the new rules, the US instituted a 0% de minimis rule for certain lithography equipment but provides an exemption for exporting countries with equivalent export controls. Lithography, dominated by the Netherlands’ ASML, is the biggest semiconductor manufacturing equipment chokepoint in the value chain. In applying 0% de minimis, the US is effectively asserting jurisdiction over foreign-made lithography equipment regardless of whether there is an actual nexus to US content.

As we noted last month (see “Déjà Vu: Round Two of Chip Controls”), US regulators could exercise leverage over Cymer, a San Diego-based firm acquired by ASML in 2013 that produces excimer lasers for ASML’s lithography machines. Applying a 0% de minimis rule captures any shred of US technology or software linked to ASML, including those Cymer-made lasers, and even cases where no US product, service, technology, or person is involved.

Although the Netherlands begrudgingly joined the United States and Japan in issuing its own version of semiconductor export controls (the Dutch controls went into effect on September 1, with a grace period until January 1, 2024), the United States has apparently assessed that these controls did not go far enough. The ball is now back in The Hague’s court to “harmonize” their rules with those of the US.

That is not the only detail causing heartburn for ASML and Dutch regulators, however. In the updated technical specs for lithography equipment in the US rules, BIS tweaked a key metric, called dedicated chuck overlay, or DCO. DCO accuracy (sometimes also called single machine overlay, or SMO) is a measure of how precisely a lithography machine can position the wafer during exposure.

A high DCO accuracy is critical to producing logic chips at or below 16/14nm and advanced memory chips using older lithography machinery. In a process called multi-patterning, the pattern is split into multiple parts, or sequences, to create structures on the wafer that would normally be too fine for an older lithography tool to produce. In this process, the wafer runs through the lithography machine multiple times to build a single layer. High DCO accuracy is therefore necessary to perfectly align the wafer under the photomask containing the pattern each time. Each generation of ASML’s DUV immersion machines has significantly improved upon DCO accuracy (see Table 3).

Chinese chipmakers have used multi-patterning to produce more advanced chips without the need for cutting-edge extreme ultraviolet (EUV) lithography machines. But multi-patterning with DUV machines is slower, more costly, more resource-intensive, and results in lower yields. The US October 17 rule update reveals a US concern over China’s continued progress toward advanced process nodes using DUV.

As shown in Table 3, the Dutch government drew the line at ASML’s NXT:2000i machine, released in 2018. In the Dutch export controls announced in June 2023, any DUV immersion lithography machine with a DCO accuracy at or below 1.5nm requires an export license starting Jan. 1, 2024. But the US Oct. 17 rules cover lithography equipment with a DCO accuracy “of less than or equal to 1.50 nm” (3B001 f.1.b.2.a.) or with a DCO accuracy “greater than 1.50 nm but less than or equal to 2.4 nm” (3B001 f.1.b.2.b.). In effect, this would capture the export of ASML’s NXT:1970Ci—a machine that is two generations behind the NXT:2000i, where the Dutch drew their line.

Table 3

Dutch vs. American red lines

Why did the US feel compelled to extend a long arm over DUV lithography? Both the US and Dutch governments focused on the same technical features (DCO and minimum resolvable feature size) to design their controls, yet reached fundamentally different conclusions on where the red line should be drawn. Both may agree on what SMIC is capable of achieving with DUV for high-volume production of advanced chips, but where the US differs is in the length it is willing to go to try and slow that timeline.

It is no secret that Chinese authorities are pouring resources into ramping up indigenous production of semiconductor manufacturing tools. Breakthroughs like the Huawei Mate 60 Pro Hisilicon Kirin 9000s processor (presumably made on SMIC’s N+2 process node using DUV lithography) and, more recently, an advanced 3D NAND memory chip by China’s leading memory chipmaker YMTC (also on the BIS Entity List) are evidence of just how rapidly China’s chip champions are moving. But having already set a bold line in the October 7 rules on what the US considers “advanced node” semiconductors,2 the US is now doubling down on efforts to constrict China’s chipmaking supply chain.

As the Dutch government, ASML, and many industry experts will argue, any semiconductor manufacturer (or university lab, for that matter) can produce 16/14nm “advanced node” logic chips and even 7nm-class chips using DUV lithography machines. With the US drawing the line at 16/14 nm instead of 5nm and gate-all-around transistor architecture (where production is more clearly tied to EUV lithography requirements), critics will argue that the US designed an inevitably leaky export control regime that fails to deny China advanced node chip production for potential military end-use, and may even accelerate China’s chipmaking innovation.

SMIC only has to look to TSMC for inspiration for what’s possible in high-volume 7nm production. TSMC’s N7 (2018) and N7P (2019) processes did not rely at all on EUV equipment in producing Apple’s A12 processor for the iPhone XS/XS Max and XR, as well as the Apple A13 processor in the iPhone 11 series. Apple sold more than 116 million smartphones with an A12 or A13 processor in 2019 alone, not including iPads that use the same chips. At least for relatively small smartphone system-on-a-chip integrated circuits (SoCs), SMIC appears capable of building up high-volume 7nm front-end capacity solely with DUV equipment (assuming it can source spare parts and service its machines under the tightened controls).

Figure 1

However, producing high-performance AI accelerators and server processors using DUV is much harder than producing smartphone SoCs with DUV. This is because production of AI accelerators and high-performance server chips tends to rely on larger dies (the “die” is the piece of silicon in the chip containing the integrated circuit itself). Larger dies allow more surface area to pack in more transistors (and thus computing power), but they are more complicated to produce. This is especially true for multi-patterning processes that incur a higher potential for error. In our earlier example of Apple’s A13 processor, the die area was 98.48mm², eight times smaller than Nvidia’s A100 AI accelerator, also produced on TSMC’s N7 node using DUV with a die area of 826mm².

The larger the die, the higher the stakes. Semiconductor yield is calculated by the number of functional dies per wafer produced. If 600 dies of an A13 chip fit on a 12-inch wafer, and we assume 20 random defects, this would produce 580 known good dies, or 97% yield. By contrast, only 66 dies of Nvidia’s larger A100 chip fit on a single 12-inch wafer, so 20 random defects would result in 46 known good dies, or 70% yield (see Figure 1). This example roughly illustrates why it is more costly to produce larger dies on smaller process nodes and why Chinese chipmakers will face challenges in manufacturing high-performance computing chips on older DUV equipment.

Will SMIC eventually be able to indigenously produce Biren’s AI accelerator on its 7nm DUV process? With enough resources and an acceptance of high costs and lower yield, it might. But the US controls reaching back further on DUV restrictions are designed to make that road as long and arduous as possible.

SMIC and its partners in China’s advanced packaging ecosystem like JCET are working on ways to address the limitations of its older equipment. Innovative “chiplet” designs allow manufacturers to split up the chip into smaller functional blocks that are separately manufactured to improve yield, and then combined again in the final chip through advanced packaging techniques like heterogenous integration. US controls for now appear to be more focused on front-end manufacturing inputs than back-end advanced packaging techniques, with the important caveat that exemptions apply so long as advanced packaging activities do not modify the integrated circuit’s technical performance level. The lengthy October 17 update to the original October 7 rules is a reminder of how US restrictions are likely to tighten alongside China’s advances in chipmaking, whether in advanced packaging, utilization of non-silicon materials, or photonic chips.

A growing compliance burden

With the US tightening controls to cover older-generation DUV machines, China’s ability to source spare parts and maintain the DUV machines already in its possession is in question. The BIS rules qualify that the restrictions apply to facilities involved in the “development” or “production” of “advanced-node integrated circuits” and state an intent to avoid “restricting servicing (including installation) at legacy-node facilities,” likely with an eye toward avoiding disruptions in existing electronics supply chains. But the language makes clear that the burden remains on exporters like ASML and their Chinese clients to verify that “100% of the items” are not being used for advanced node production, or else they will face licensing restrictions. With China’s chipmakers under the gun to accelerate advanced node production, parsing between advanced node and legacy chip production at foundry sites with multiple production lines could prove to be a heavy, ongoing compliance burden for foreign SME suppliers.

Figure 2

BIS expects technology firms to also keep abreast of China’s innovation in meeting evolving compliance standards. The new rules include “red flags” for companies to help them screen against circumvention risks. In trying to account for inevitable Chinese technological leaps, “red flag 15” in the advanced semiconductor rule describes a scenario where “a customer’s website or other marketing materials indicated that the company had advertised or otherwise indicated its capability for “developing” or “producing” “advanced-node integrated circuits.” Companies applying the new red flag screening would thus have to account for scenarios like the Huawei Mate 60 Pro Hisilicon Kirin 9000s processor in reviewing whether the advancement of Chinese firms requires fresh compliance checks for companies in the supply chain that could have enabled that progress.

Extending a long(er) arm

Among the most novel features of the original October 7 rules was the application of Section 744.6 of the EAR, which restricts the activities of US persons, to apply extraterritorial controls. This resulted in an overnight scramble for licenses by China-based fabs owned by Samsung, SK Hynix, and TSMC, and it set off months of negotiations to determine whether US regulators could trust foreign-owned fabs to keep restricted items out of the hands of Chinese chipmakers. The US persons move had a far-reaching impact, prompting companies in other industries to audit their vulnerabilities to the potential future application of US persons controls. In the new October 17 rules, BIS creates carveouts for partners that aim to mitigate the unintended consequences of the US persons rule marking US talent as liabilities for foreign firms trying to keep a foothold in the China market.

In the October 17 update, BIS put less emphasis on the US persons rule but still got creative in expanding long-arm controls beyond China (and Macau) to intermediary countries that might help China circumvent export controls. At the same time, BIS attempted to apply a more constructive tone to the rules in highlighting the benefits of being a US partner when it comes to licensing exemptions and streamlined processing for when licenses are warranted.

The Commerce Department determines which countries are subject to stricter or more permissive export control by categorizing them into Country Groups. In the October 17 rules, BIS imposed different restrictions on Group A countries (more trusted) versus Group D countries (less trusted) with the aim of mitigating the risk of China circumventing of export controls through third countries (see Table 4). In drawing a distinction between Commerce Group A countries for US partners eligible for licenses and exemptions and Group D countries for unreliable entities under US arms embargo or other restrictions for national security reasons, the US is effectively sketching a geopolitical map for US-China technology competition (see Figure 3).

Table 4
Figure 3

In some cases, the new rules are given global reach to prevent export control circumvention. Restrictions can apply worldwide for cases in which the “direct product” of high-performance compute controlled items (EECN 3E001 technology for ECCN 3A090 items) is destined to an “entity headquartered in (or whose ultimate parent company is headquartered in)” a Country Group D:5 country, including China and Macau. The worldwide application is triggered any time these entities are a party to the transaction involving the foreign-produced item, including as a “purchaser,” “intermediate consignee,” “ultimate consignee,” or “end-user.”

BIS invokes this description of “entities that are headquartered in, or whose ultimate parent company is headquartered in” Macau or Country Group D:5, throughout the rules, both in qualifying partner exemptions and in specifying end-user controls. This phrasing appears designed to deter third parties from hosting or collaborating with Chinese firms to produce high-performance chips that meet the updated BIS parameters.

This “entities” provision raises serious questions on the sustainability of MNC partnerships with China-headquartered firms and their subsidiaries in developing commercial applications involving high-performance chips and related services (for instance, advanced chips used in autonomous driving that may not yet meet current parameters but that might cross that line within 3-5 years as technology advances.)

Dear partners, you’re welcome…?

The new country restrictions and exemptions are replete with nuances and caveats meant to encourage diversification away from China without upending global supply chains or overly antagonizing US partners. For example, the rules create a new Temporary General License (TGL) for US-headquartered companies and companies headquartered in Group A:5 and A:6 countries. It also makes clear that the December 31, 2025 expiry date is meant to give partners “additional time to identify alternative sources of supply.” In other words, South Korean memory chipmakers and TSMC are still on the BIS licensing hook to continue operating in China and are facing a deadline to diversify.

Finally, the 0% de minimis rule for certain lithography equipment discussed above is remarkable. This rule effectively asserts that even if no obvious US linkage exists via a person, technology, product, or service, the US nonetheless “retains jurisdiction over such foreign-made equipment to protect US national security and foreign policy interests.” To what extent US technology and security partners, such as the Netherlands, push back on the long-arm lithography restrictions remains to be seen. We see potential for the US and Netherlands to negotiate a middle ground (for example, ongoing due diligence requirements for a handful of fabs in question in China.)

The year behind and ahead

The geopolitical climate has evolved considerably since the US shook the world with the October 7 controls one year ago. Reflecting on the past year, the following themes stand out:

  • Chinese retribution is on full display. From symbolic moves like Chinese legislators codifying the right to countermeasures to concrete actions like a selective product ban on US memory chipmaker Micron and export restrictions on gallium, germanium, and graphite (critical inputs for semiconductor and battery production) Beijing has been telegraphing its options to strike back at US-led chip controls and disrupt technology value chains.
  • But Chinese retaliatory moves did little to temper US-led technology and investment controls. In addition to the October 17 package of rules discussed here, the US rolled out a new outbound investment screening tool in August with a broader scope for semiconductors and unanswered questions on AI (see August 11, “Big Strides in a Small Yard”). Final CHIPS Act guardrails applied stringent conditions on CHIPS Act recipients to deter future investments in China.
  • The US resurrected diplomatic “guardrails” with China. US cabinet officials stepped up their outreach to China, with three secretary-level visits since June. One of these visits established an export control working group to communicate intentions and red lines on tech restrictions, though Beijing may be questioning the value of this high-level diplomatic channel given how much punch is packed into the details of the October 17 controls.
  • China was able to demonstrate technological breakthroughs despite US and partner chip controls. Recent teardowns of the SMIC-manufactured Kirin 9000s SoC and YMTC-manufactured 3D NAND flash memory chip revealed the tenacity of China’s chipmakers (all while stockpiling of SME and China-compliant chips soared.)
  • The US was largely able to steer partners into alignment. Japan and the Netherlands followed through with their own export controls (even as the US is now spinning up another round of negotiations with The Hague with its new long-arm lithography controls). The US was also able to bring together Japan and South Korea in a trilateral security summit where supply chain resilience for semiconductor manufacturing is a key feature of these talks (alongside parallel discussions with Taiwan.)
  • The EU activated its regulatory toolkit in targeting China-made EVs with an anti-subsidies investigation, passing a new Anti-Coercion Instrument, and rolling out its new Carbon Border Adjustment Mechanism. With the EU now leaning in with measures to “level the playing field” and the US and Japan following closely behind, China faces a much bigger challenge in trying to sow divisions among G7 countries. At the same time, European nations, especially Germany will be reticent to mirror US technology restrictions while the EV probe is underway and retaliatory risks run high.
  • AI innovation surges while regulators try to keep up. The past year has seen a range of regulatory actions toward generative AI, with China and Europe leaning toward a more restrictive approach, India scaling back after initially applying a heavy hand, and the US trying to find a balance between stimulating innovation and drawing boundaries with a new Executive Order on AI. The pace of technological development is catalyzing US and G7 partner discussions on AI standards and norms, how to govern cross-border data flows and define dual-use AI foundation models. This dynamic is pushing China toward a potential ‘breaking point’ in its own data security regime as it weighs the economic consequences of being excluded from emerging US and partner AI ecosystems.

Looking ahead

These geopolitical features will inform the next regulatory, diplomatic, and retributive moves to come. Our watch list for the post-October 17 world includes the following:

  • Partner negotiations with a political deadline. The US-Netherlands negotiation on licensing conditions for lithography equipment will be most urgent, but the Dutch won’t be alone. Germany (whose companies like Trumpf produce machine tools and lasers, and Zeiss, which produces optics for semiconductor manufacturing) and South Korea, whose leading SME players include SEMES, Wonik IPS, PSK, and Eugene Tech, will be high on the US priority list for discussions on export control alignment. And with the November 2024 US presidential election looming, the US administration can put it to partners on whether they would rather reach an understanding with the current administration or take a gamble on the next one.
  • Controls creeping into the cloud space. The October 17 controls contain a seedling of restrictions on “chips as a service” that is likely to grow bigger in time. The new rules include an end-use control on high-performance computing items to prevent companies headquartered in D:5 countries, including China, to set up cloud or data servers in other countries. The reason cited is that China could access high-performance compute capacity “to train AI models in ways that would be contrary to national security.” In its solicitation for public comment, BIS also reveals an intention to regulate infrastructure-as-a-service (IaaS) providers to restrict customers involved in dual-use AI foundation model production. This could escalate US-China tech competition to another level as Chinese cloud service providers run into compute limitations, and US cloud service providers in China face tighter restrictions from Chinese authorities in response.
  • Where to draw the line on consumer-grade AI, integrated circuits, and connected systems? The October 17 rules grapple with how to distinguish between consumer and potential dual-use applications for AI and advanced computing chips in trying to determine where restrictions should kick in. For example, data center usage (plus performance threshold) is a key factor for BIS in determining which transactions involving high-performance computing chips will be restricted outright, and which now require notification. BIS is also soliciting public comment from cloud service providers on how they would monitor and flag to BIS when their customers are “developing” or “producing” a dual-use AI foundation model. Parallel to the export control buildout, outstanding questions remain over how Treasury will scope dual-use AI applications in new outbound screening rules. The BIS rules include provisions highlighting exceptions for “consumer communications devices” in the October 17 rules. BIS also notes that it has not (yet) classified network security appliances or DNA sequencing machines. That said, BIS has advised companies to screen for encryption functionalities in determining whether a technology product or service could come under restriction (the US applies the rare 0% de minimis provision for certain encryption products.) Notably, BIS also added new ‘.z’ designations to ECCNs to clarify controlled items that also meet the high-performance compute threshold. Given the fast pace in which AI-powered technologies are proliferating, BIS is trying to lay the groundwork through these rules to ensure the US list of controlled items can keep pace with potential dual-use applications.
  • Minding the legacy chip conundrum. US regulators appear mindful of sticky dependencies on China for legacy chipmaking (see March 3, “Running on Ice: China’s Chipmakers in a Post-October 7 World”). Given BIS’s export control mandate, the rules focus on where the agency has drawn the line on advanced process node chips while carving out exceptions for legacy chip node production, including exemptions for servicing lithography equipment not used for advanced node production. BIS has also so far refrained from elevating controls on SMIC from its current BIS Entity Listing to Foreign Direct Product rule designation. BIS restraint is likely linked to concerns over creating bigger supply chain disruptions in electronics supply chains given China’s growing role in legacy chip manufacturing. At the same time, US policymakers have voiced concern over China growing its “coercive leverage” in commodity chips and are pointing to China’s export restrictions on critical materials as evidence of the risk. This concern will be dealt with via other regulatory tools, such as conditions on US CHIPS Act fund recipients to deter them from investing in expanding China’s semiconductor manufacturing capacity, incentives for fund recipients to reserve capacity for legacy chipmaking in the name of supply chain resilience, and potential “level the playing field” measures targeting China’s chipmakers over market distortions created by subsidies.
  • The retribution factor. As noted above, China has already been demonstrating its retaliatory leverage at the lower end of electronics, chip, and battery supply chains. The question ahead is whether China’s retribution intensifies as the US proceeds apace with tech controls. Foreign technology firms are at the highest risk of becoming targets in areas where Chinese companies pose credible competition (for example, OPPO/Xiaomi/Vivo/Huawei versus Apple, Luxshare versus Foxconn, or Dell and HP versus Lenovo). US and partner-headquartered firms may face a higher risk of product bans, either informally or explicitly, on cybersecurity grounds. MNCs relying on US cloud service providers in China or Chinese cloud service providers with operations overseas may face challenges ahead as new restrictions risk setting off a chain reaction for retaliatory moves in the cloud space.

Placing a big bet

The Sullivan doctrine is alive and well in the October 17 controls. The new rules emphasize the US National Security Advisor’s “preempt and deter” posture toward protecting a US and partner tech ecosystem and the accepted ambiguity that comes with regulating fast-moving disruptive technologies. This is a paradigm shift that the corporate world is still trying to internalize with each regulatory rollout, whether in export controls, industrial policy guardrails, investment screening, trade defensive measures, or emerging data and AI governance policies.

For China, the stakes are clearly rising. Even as semiconductor manufacturing tops China’s priority list for industrial policy and tech self-reliance, China’s fiscal capacity at the local government level is highly constrained (see August 23, “The Myth of China’s Fiscal Space”). This dynamic can lead to trade-offs at the local government level between science and technology-related spending and other vital funding demands like pensions and social security, not to mention interest payments on rising local government debt. With inbound foreign investment stagnating, Beijing must make trade-offs between showcasing retaliatory leverage and sustaining China’s appeal to foreign investors. For example, even as China is restricting critical inputs to tech supply chains (and inadvertently energizing G7 de-risking policies), Chinese regulators are trying to reassure foreign investors with a partial easing of cross-border data regulations.

But China’s tech champions face an uncertain future in the wake of the October 17 controls. On the one hand, state support and retaliatory moves that cut down foreign competition can carve out more market share for domestic players. On the other hand, China faces critical vulnerabilities in semiconductor supply chains that are foundational to its economic growth. If, for example, China struggles to indigenously produce chips at scale for its data centers, Xi Jinping’s vision of “digital China” to spur new economic growth with big data infrastructure buildouts will come into question. With companies like Alibaba, Tencent, and Baidu innovating at the nexus of AI and cloud computing, will US-led dual-use controls begin to impact China’s tech champions more directly, pushing Beijing toward more aggressive forms of retaliation? And will a cornered Beijing view earnest US appeals for diplomatic and military guardrails amid this tech escalation as leverage to demonstrate the consequences of not picking up the phone when a security crisis emerges?

US National Security Advisor Jake Sullivan appears to be placing a bet that time and innovation is on the United States’ side while China’s statist approach to managing its economy will result in more fumbles than breakthroughs. As Sullivan wrote in a recent Foreign Affairs essay, “we need a sustained sense of confidence in our capacity to outcompete any country. The past two and half years have upended assumptions on the relative trajectories of the United States and China.” The October 17 controls are brimming with such confidence, all while China’s structural economic challenges have been brought into sharp relief. US faith in diplomatic guardrails to manage this escalation, however, may be premature.


This note has been prepared in collaboration with Jan-Peter Kleinhans at Stiftung Neue Verantwortung.


1. According to the new rules, total processing performance (TPP) = 2 * MacTOPS * bit length of the operation (aggregated processing units on the integrated circuit). “Mac” stands for “multiply-accumulate computation” (D = A × B + C) and BIS states that “2 * MacTOPS may correspond to the reported TOPS or FLOPS on a datasheet.” The intent is to capture the theoretical peak number of Tera operations per second for multiply-accumulate computation.

2. The EAR defines “advanced-node” integrated circuit to include ICs that meet any of the following criteria: (1) logic ICs using a non-planar transistor architecture or with a “production” technology node of 16/14 nanometers or less; (2) NOT AND (NAND) memory ICs with 128 layers or more; or (3) dynamic random-access memory (DRAM) ICs using a “production” technology node of 18 nanometer half-pitch or less

Sustainable Aviation Fuel Workforce Development: Opportunities by Occupation

Aviation is one of the fastest-growing emissions sources in the US, as well as one of the hardest to abate. With air travel demands unlikely to slow down, decarbonizing aviation will be critical in achieving a net-zero economy by mid-century. A leading solution for decarbonizing aviation is the use of sustainable aviation fuels (SAF) as a drop-in replacement for conventional jet fuel. Doing so can dramatically reduce life-cycle carbon dioxide emissions compared to conventional jet fuel. In recent years, we have witnessed growing US policy support and private investment to help expedite SAF commercialization and scale-up. However, scaling up the industry will require a robust workforce comprised of skilled laborers across a variety of occupations.

In this note, we assess the employment benefits associated with a commercial-scale SAF production facility—focusing on the total number of jobs created as well as the types of jobs created. We find that constructing and operating a commercial facility results in an annual average of 2,210 plant investment jobs and 1,440 ongoing jobs. It also leads to the employment of a wide range of professions including construction workers, engineers, farm laborers, and other agricultural workers.

Getting the sustainable aviation fuel industry to scale

The US aviation sector is the third largest source of US transportation emissions, accounting for approximately 7% of total emissions from the sector. It is also one of the fastest-growing sectors. Emissions are expected to increase as air travel and air freight transport demands continue to rise.  Decarbonizing aviation is particularly difficult in large part due to commercial aircrafts operating, on average, 20 to 30 years. Considering this, if the US relied only on improvements in aircraft design and efficiency and/or the electrification of aircrafts, it would take a considerable amount of time to decarbonize the sector, not to mention that it is unclear if electric-powered airplanes will ever be able to service long-haul markets like existing commercial airliners. Instead, the most viable option for mid-century decarbonization could be the use of sustainable aviation fuels (SAF). SAF is a type of “drop-in” fuel produced from either biological or non-biological feedstocks. Being a drop-in fuel means that it can easily and safely be dropped into existing jet engines and fueling infrastructure in place of jet fuel. Compared to conventional jet fuel, SAF can reduce total life-cycle carbon dioxide (CO2) emissions by up to 100%, with some technology options for net-negative jet fuel. Its drop-in status and lower life-cycle carbon footprint suggest that its use in place of jet fuel can produce significant reductions in aviation emissions.

In recent years, several new SAF production facilities have been announced in the US. Total capacity is expected to reach 2.7 billion gallons per year (BGY) by 2030 (Figure 1). This would satisfy only about 7% of projected US 2030 jet fuel consumption. The hydroprocessed esters and fatty acids (HEFA) process is currently the only commercial SAF production pathway. However, the availability of sustainable feedstocks, and competition with other sectors for HEFA-based fuels, limit its use for aviation. Investment in other, less mature production pathways like bio-based synthesis gas Fischer-Tropsch (Gas-FT) and alcohol-to-jet (AtJ) will be needed if expanding SAF production capacity is to continue. We have chosen to focus on the AtJ pathway for this analysis given recent shifts in investment toward the commercialization of this technology.


Production capacity levels in the US are far from being on track for what is needed to achieve mid-century decarbonization targets for aviation. If SAF is the predominant way to decarbonize, we estimate 33 to 36 billions of gallons per year will be needed to meet US aviation demand in a zero-emissions aviation sector (Figure 2). For SAF to scale exponentially, policy and private sector support needs to accelerate.


Here in the US, we are seeing more policy support and funding opportunities at both the federal and state levels to boost the nascent industry. Federal support includes: (1) new RD&D funding for SAF-related projects; (2) establishment of the SAF Grand Challenge; and (3) last year’s passage of the Inflation Reduction Act, which establishes SAF production tax credits through 2027. Other supporting federal policies include the Renewable Fuels Standard (RFS). The RFS allows SAF producers to “opt-in” and generate renewable identification numbers (RINs) that can then be sold to refiners to help offset their renewable blending requirements. The average price of a biomass-based diesel RIN (D4 RIN) in 2022 was $1.66. Unlike obligated parties, SAF producers do not face volume obligations under the RFS. State-level policies include California’s Low-Carbon Fuels Standard (LCFS). Like the RFS, it does not hold SAF producers to any volume obligations but allows them to generate and sell LCFS credits to obligated participants for revenue. Another state-level policy is the Invest in Illinois Act (“Invest Act”). The legislation offers a SAF purchase credit for airlines through 2033.

These policies in conjunction with increasing US airline offtake agreements and private sector investment are driving the US SAF capacity expansion. In 2022, there were 42 announced offtake agreements, the highest of any year in the last decade. This includes an agreement between the major US carrier Delta Airlines and SAF producer GEVO. Starting in 2026 Delta will receive 75 million gallons of SAF annually over the next seven years from GEVO. US carrier United Airlines also announced a joint agreement with Dimension Energy in which they agreed to purchase at least 300 million gallons of SAF over the next couple of decades. But even more support across the public, private, and market dimensions is needed if a net-zero aviation sector is ever to exist.

Another major requirement for scaling up SAF will be the availability of a capable workforce. In our recent 2022 report, Sustainable Aviation Fuels: The Key to Decarbonizing Aviation, we investigated the job creation potential in the US for a range of SAF production pathways. Although our prior work shed light on potential workforce levels associated with commercial SAF deployment, it lacked details on the types of jobs employed. In this note, we provide additional analysis of workforce development in the SAF industry, including the types of jobs created.

SAF workforce opportunities

To successfully scale up SAF by mid-century, we need a capable workforce with skilled laborers in a variety of occupations. For this analysis, we estimated the total number of jobs created by the construction and operation of a 50 million gallon per year SAF facility utilizing the AtJ production process. The selected production capacity falls within the range of capacities for announced SAF projects. As of today, there are no commercial AtJ facilities in operation. However, new commercial-scale projects have been announced in the last few years. The world’s first commercial AtJ facility will be LanzaJet’s Freedom Pines Fuels facility located in Soperton, GA. The ethanol-based production facility is on track for commissioning later this year.

Because there are currently no active commercial facilities, there is no empirical data we can utilize to conduct our analysis. Therefore, we developed a methodology that estimates SAF employment based on adjacent industries. We begin by determining the capital and operating costs for each production process based on expert interviews, literature reviews, and Rhodium analysis. Next, we conduct in-depth research on how the distribution of cost varies by production method and link these costs to the appropriate industries reflected in the input-output economic model IMPLAN. We estimate employment numbers by inputting our cost data into IMPLAN using its national-level dataset. Occupational results are also extracted from the model and supplemented with Bureau of Labor Statistics data. Our employment numbers capture both the on- and off-site jobs associated with the construction and operation of a commercial SAF facility. They also capture jobs associated with upstream supply chain activities.


We find that the facility generates, on average, 2,210 plant investment jobs each year over a three-year construction period (Figure 3). Plant investment refers to the construction, engineering, materials, and any equipment needed to build the SAF facility. About two-thirds of these are facility-level jobs as they are directly related to the on-site construction of the facility and its many systems. All remaining jobs support related supply chain activities (e.g., business support services). A total of 1,440 jobs are associated with ongoing facility operations and maintenance (O&M). 38% (540 jobs) of those jobs occur on-site. Most of the O&M jobs occur off-site and support related supply chain activities. A further drill down reveals that a significant portion of the supplier jobs are associated with the production and transport of the corn stover feedstock.

Job breakdown by occupation

The purpose of our occupational analysis is to show the types of jobs stemming from the building and operation of a commercial AtJ SAF facility, with the focus being on plant investment and ongoing O&M employment.[1] Figure 4 shows the top five occupational categories associated with plant investment. They are: (1) construction trades; (2) metal workers and assemblers; (3) engineers; (4) executives and business operations; and (5) machinery installers, maintenance, and repairers. Together, these occupations account for about 61% of all non-proprietor plant investment employment. Non-proprietor employment refers to wage and salary workers—in other words employees that are not owners.


Construction is the largest occupation category, which is not surprising given that it captures all physical labor types involved in the construction of the facility’s exterior shell, building envelope, and many of its complex systems. Roughly 43% of all construction trade workers are either construction laborers or site managers/supervisors (Figure 5). Other prominent occupations include carpenters, electricians, and wall and flooring workers.


The largest ongoing O&M occupational category is agricultural workers and managers (Figure 6). The AtJ process converts alcohol feedstocks from biomass (i.e., sugars, starches, hydrolyzed cellulose) into SAF. Therefore, the majority of the O&M activities are associated with the processing and handling of biomass feedstock inputs. The next largest category is machinery installers, maintenance, and repairers, which includes industrial machinery mechanics and mobile equipment mechanics.[2]


Overall, our findings suggest that large-scale SAF deployment has the potential to generate thousands of jobs while establishing a workforce that is characterized by a wide range of occupations. Many of these occupations will require both technical and extensive training. As the industry continues to expand, it is imperative that a wide range of resources be directed towards the development and support of new on-the-job training assistance programs. This will ensure that the SAF workforce is well equipped with the skills and knowledge necessary to perform any essential duties.

Forthcoming analysis

So far, there has been limited work exploring the types of occupations associated with the construction and operation of a commercial SAF facility. Our estimates are based on the most up-to-date data in a dynamic and fast-moving technology space. As the SAF industry develops we will refine and expand our analysis. We will also continue to expand upon our employment analysis to investigate important factors including wages, union labor, and required skills.

This note serves as the last in a three-part series on the employment and occupational impacts of selecting emerging clean technologies. The other two notes assess workforce development associated with clean hydrogen and direct air capture facilities.

[1] Supplier jobs are not considered in our occupational analysis.

[2] For additional information on any of the occupations discussed in this note, such as general job duties, wages, or working conditions, refer to the latest edition of the US Bureau of Labor Statistics’ Occupational Outlook Handbook

This nonpartisan, independent research was conducted with support from Breakthrough Energy. The results presented reflect the views of the authors and not necessarily those of supporting organizations.

Opening Salvo: The EU’s Electric Vehicle Probe and What Comes Next

On October 4, 2023, the European Commission launched an anti-subsidy investigation into imports of passenger battery electric vehicles (BEVs) made in China. The investigation underscores mounting concerns in the EU about the impact that a wave of cheap, subsidized BEVs from China could have on Europe’s fledgling industry. While BEVs occupy the spotlight at the moment, the EU is also reportedly considering trade defense measures against Chinese wind turbines and steel. Many more industries are recipients of China’s industrial policy largesse, and could become targets for EU policymakers in the years ahead. In this note, we explain the EU’s concerns about China’s BEV industry, present the likely strengths and weaknesses of the Commission’s case, and explore how trade measures could be expanded, both by the EU and its G7 allies, to encompass a broader set of subsidized industries in China.

EU concerns: Subsidies, over-capacity, and a rise in cheap exports

There are three main concerns informing the EU’s decision to open an investigation into imports of China-made BEVs: substantial Chinese state support to its electric vehicle industry, a rapid increase in cheap exports to Europe, and mounting over-capacity in China that could accelerate those exports. Taken together, these factors could threaten what the Commission views as “an already vulnerable EU industry” that is still years away from breaking even. At the very least, an influx of cheap Chinese BEVs could hinder the investments needed to scale up Europe’s home-grown industry.

By launching a rare ex officio investigation, rather than responding to industry appeals for a case, the Commission is sending a signal that it is prepared to use the tools at its disposal to protect and bolster the competitiveness of European industry—a major theme in Commission President Ursula von der Leyen’s state of the union speech. Importantly, the Commission is taking a preemptive approach by basing the case on the threat of injury, rather than already documented harm, in an effort to avoid a repeat of Europe’s experience with the solar photovoltaic (PV) sector a decade ago, when state-sponsored Chinese players shattered Europe’s domestic industry.

In the late 2000s, Chinese solar PV manufacturers leveraged foreign technology and abundant state financial support to achieve dominance across the PV value chain. China’s share of global PV cell production surged from 14% in 2006 to 60% in 2013. From 2013 to 2018, the European Union, following the United States, imposed anti-dumping and countervailing duties on solar panels, cells, and wafers from China. These moves temporarily halted the growth in China’s share of global exports (Figure 1), but failed to restore the share lost by the EU or US.

Figure 1
Figure 2

By moving early to impose countervailing duties (CVDs) on BEV imports from China, the European Commission hopes to avoid a similar fate for Europe’s BEV industry. Trade data shows that China’s BEV exports have grown at lightning speed over the past three years (Figure 2). China’s share of global BEV exports grew from just 4% in 2020 to 21% in 2022, and China’s share of EU BEV demand grew from 4% to 16%.1 It is important to note that a majority of BEVs exported from China to the EU are currently foreign brands—chief among them Tesla, but also European brands with joint ventures or contract manufacturing agreements in China, such as Renault and BMW. Renault’s Dacia Spring, produced in Hubei, was the 10th best-selling BEV model in Europe in H1 2022 with more than 19,000 cars sold; and almost 9,000 of BMW’s China-made iX3 models were sold in Europe during the same period. But Chinese brands are growing increasingly competitive. Their share of total EU car sales grew from 0.4% in 2019 to 8% in 2023. According to KPMG, Chinese groups such as BYD and SAIC could capture as much as 15% of new car sales in Europe within the next two years.

The Commission argues that rapidly rising Chinese exports of lower-priced BEVs (about 20% below European unit prices), and the sharp rise in China-based production capacity constitute an imminent threat to Europe’s auto industry. In late 2021, capacity utilization rates were only 58.4% for new energy passenger vehicle manufacturing in China. Persistently high levels of investment in the sector (+56% y/y capex growth in H1 2023) could exacerbate overcapacity in the years ahead, especially against a backdrop of slowing Chinese growth. The Commission fears many of these artificially low-priced vehicles will find their way to Europe.

Making the case: Documenting China’s support to the industry

It is noteworthy that the Commission chose to launch an anti-subsidy rather than an anti-dumping investigation into imports of China-made BEVs. In the 2012 case against China-made solar PVs, the Commission launched both simultaneously. There are two plausible explanations for this choice. First, anti-dumping investigations are best applied to basic products like solar panels or raw materials. Because of their complexity, BEVs might not lend themselves as easily to an anti-dumping case. Second, countervailing duties tend to be lower than anti-dumping duties. In bringing an anti-subsidy case, the Commission may be signaling to a reluctant country like Germany that it will remain measured in its response.

In order to impose countervailing duties without running the risk of a major rebuke from national European courts (a more immediate threat than a WTO counter-case, which can take years), the Commission will need to prove three things: 1) that exporters of China-made cars received countervailable subsidies from the Chinese government, 2) that Europe’s industry is under imminent threat of injury, and 3) that there is a causal link between the two.

The existence of state subsidization should not be difficult to show. The BEV industry became a priority for Beijing in the 2010s, as it became clear to China’s leaders that the country’s efforts to achieve competitiveness in internal combustion engines had stalled. Several policy schemes to support BEV makers, including purchase subsidies paid directly to China-based manufacturers, are openly disclosed by the Chinese government. In fact, the EU’s initiation notice points to publicly available evidence of a wide variety of support tools, including direct grants and subsidies but also a range of tax exemptions, low-cost financing, and provision of inputs at below-market prices.

However, measuring the full extent of China’s subsidization to provide sufficient evidence of potential imminent harm might be more difficult. There is abundant quantitative evidence of direct support instruments such as grants and purchase subsidies (Figure 3), which other countries also use to prop up their BEV industries. But most of China’s support to strategic industries tends to take the shape of government-sponsored access to cheap credit and equity, as uncovered by recent OECD work on the aluminum, rolling-stock, and semiconductor sectors. These non-conventional support channels can be much harder to quantify because they require comparing actual costs of capital with theoretically expected costs based on future market conditions.

In its case, the Commission will have to contend with the fact that non-Chinese carmakers (including those based in the US and Europe) have also received significant state support for their China operations, as a means to encourage them to shift production and export capacity from Europe to China. After BYD, Tesla was the second-largest recipient of pre-allocated purchase subsidies in 2020-2022, receiving a total of RMB 1.24 billion (USD 192 million). Volkswagen’s joint ventures with SAIC and FAW, meanwhile, pulled in RMB 0.58 billion (USD 79 million) in such subsidies over the same period.

Figure 3

Because batteries account for around 25-35% of the cost of a BEV, China’s support for this industry will also be important for Brussels to consider. We estimate that Chinese firms BYD and CATL, both of which produce batteries, received approximately USD 4.3 billion and USD 2.5 billion in state support between 2015 and 2020, respectively, more than half of it through below-market equity.2 It is therefore quite likely that the Commission will factor in upstream subsidization (battery manufacturing, and possibly also the processing of critical raw material inputs).

The Commission will also have to show that such practices have contributed to a surge in low-priced exports, which constitute an imminent threat to European industry. It will likely rely on evidence of production overcapacity and signs that this overcapacity is being (or will be) exported to Europe. Although BEVs continue to receive subsidies through tax exemptions and local government programs, the end of the flagship central government purchase subsidies program in January 2023 and slowing domestic demand could increase the temptation to export. The current price war between BEV makers—with Tesla, BYD, and smaller firms like Nio announcing large price cuts in 2023—is a sign that supply and demand are out of whack in the Chinese market, providing companies with a powerful incentive to look for more lucrative markets overseas.

Targets on the horizon: Wind turbines and heat pumps?

While the Commission is focused on BEVs for now, we expect it to eventually turn to other green tech industries where state support in China is distorting global markets. There is, for example, abundant evidence of state support for China’s domestic wind turbine industry, which benefitted from a combination of demand-side feed-in tariffs and strong local content requirements until 2009.3 After that, China-based wind turbine manufacturers benefitted from abundant production subsidies, with a marked uptick in the last couple of years (Figure 4). Over the same period, China’s share of global exports of wind turbines doubled, from 12% in 2019 to 21% in 2022 (Figure 5). China’s share of EU wind turbine demand grew even faster, from 6% in 2018 to 28% in 2022. China’s increased production capacity contributed to a major drop in the price of wind turbines in China, from USD 700,000 per megawatt in early 2020 to USD 470,000 in late 2021, putting Chinese manufacturers at an advantage compared to foreign—and especially European—manufacturers struggling with rising production costs. It is no surprise, then, that the Commission is reportedly considering an investigation into imports of China-made wind turbines.

Figure 4
Figure 5

It could be harder to build a case against China’s heat pump industry, which benefited from state support, but much of it from demand-side subsidies, which are less likely to trigger countervailing duties. Still, domestic companies often benefitted more from purchase subsidies than foreign firms. For example, most of the heat pump companies (23 out of 26) selected by the government for heat pump water heater purchase subsidies were domestic brands, and no foreign firm was eligible for the highest amounts of subsidies. This support propped up demand and, in turn, boosted the revenues and profitability of domestic manufacturers. China’s share of global heat pump exports expanded in ways similar to BEVs or wind turbines—more than doubling in the past four years (Figure 5). The majority of the export surge has come from Chinese OEMs working with foreign brands, but the exports of local brands are also booming, with Midea’s air source heat pump water heater exports, for example, increasing more than 200% year-on-year in the first half of 2022. The sector, finally, is ripe with overcapacity, which Chinese demand might be incapable of absorbing. This situation could trigger a similar reaction from Brussels.

The case of lithium-ion batteries

As mentioned above, the battery sector has been a major recipient of Chinese state support over the past decade, resulting in rapidly falling prices and substantial overcapacity. In 2022, China’s production of lithium-ion batteries was 1.9 times the cumulative installed volume. Despite this, China’s support to the sector is still growing, with direct support mechanisms such as grants on the rise—even as those to BEV makers have come to plateau (Figure 6).

However, there are two important differences between batteries and the other green tech sectors mentioned above, and these have implications for how the EU might respond. First, China’s dominance in the sector means that a trade investigation and resulting duties on China-made batteries could create major short-term inflationary pressures on downstream European industries, as alternative sources of supply are limited and more expensive. China today accounts for nearly half of global battery cell production, in addition to dominating the global market for processed battery components such as anodes, cathodes, electrolytes, and separators. Second, Chinese battery makers have invested heavily in Europe, contributing to local economic activity and leading to an on-shoring of battery capacity—a form of de-risking, even if plants are Chinese-owned. It seems unlikely that the EU will take action against Chinese battery makers in the short- to medium-term, including through its new foreign subsidies instrument, which allows the EU to target subsidized production in Europe.

Figure 6
Figure 7

This could change in the longer term, however. As new battery manufacturing hubs emerge in Europe, the price effect of trade action will subside, while the incentive to preserve these hubs will increase. Poland, Hungary, and the Czech Republic’s combined market share reached 36% of EU battery demand in 2021, up from 7% in 2017, and largely driven by investment from South Korean companies Samsung, SK Innovation, and LG Energy. But that share has stagnated over the past two years, while the share of Chinese batteries in EU imports jumped to 46% in 2022, from 31% in 2021. This was largely due to price differences: in 2022, Chinese batteries were 33% cheaper than those produced in Europe. As Europe strives to create a home-grown battery industry, Chinese-owned mega-factories on the continent—including the likes of CATL’s €8 billion Hungary factory—may be seen to pose a threat to emerging European champions like Northvolt or Verkor, especially if they put depress prices in ways unsustainable for competitors.

European divisions vs. Brussels’ momentum

The Commission has 13 months starting October 4 to make a final decision on the case, and potentially impose permanent countervailing duties on imports of China-made BEVs. While this timeline pushes the final decision to the next Commission, the von der Leyen Commission could impose provisional CVDs within a two-to-nine-month window following the launch of the investigation. An early decision to impose duties might win favor in France, but it would likely rile other member states, chief among them Germany.

Berlin has sent mixed signals about the probe, reflecting divisions within the coalition government. But there is significant concern in Germany that EU duties would trigger retaliation against its carmakers, which are deeply dependent on the Chinese market. France, on the other hand, has strongly supported the investigation, which it views as an opportunity to bring manufacturing jobs back and increase Europe’s resilience in green technologies. Paris has even gone a step further, proposing to introduce national tax credits for electric vehicle purchases based on EV makers’ carbon footprint. France’s methodology for carbon scoring would cover manufacturing and transport (including batteries) and could set a precedent for other member states. Italy is reportedly considering a similar scheme. Nordic states, which are trying to get battery manufacturing off the ground and may be wary of retaliatory moves by China to restrict critical inputs, have not reacted as strongly to the EV probe. Central and Eastern European countries, which are highly integrated with carmakers in Germany, could choose to align with Berlin. But they might also welcome new BEV investments from China, prompted by Europe’s trade measures.

Regardless of how member states position themselves, the Commission could decide to plow ahead, as it did with solar panels a decade ago. Its biggest battle will be in European national courts, where cases will likely be launched by Chinese and European automakers alike. In this event, the Commission will need to prove not just subsidization but that subsidies have produced a significant price differential between Europe-based and China-based BEV makers, and that this poses an imminent threat to home-grown players. The standard will be high.

If the Commission’s case is successfully challenged in courts, political support for future cases could wane. Beyond the BEV probe, and potential wind turbine and steel investigations, we could envision, in the longer term, the use of various EU instruments to take aim at China’s support to its heat pumps, medical devices, and rolling stock industries. In addition to traditional trade defense tools and the new foreign subsidies regulation (discussed above), the EU also has available its new international procurement instrument (a tool to deny or limit third country firm access to EU public procurement markets where third market public procurement is closed to EU firms). But significant additional action would likely hinge on the success of the BEV case.

How will China respond?

Beijing is clearly alarmed by the momentum in Europe behind the use of defensive trade measures. In response to the EV investigation, Chinese authorities have accused the Commission of “naked protectionism” and signaled a willingness to retaliate. In the near term, while the EU investigation is underway, we believe that Beijing is unlikely to announce formal countermeasures. Instead, it may resort to closed-door threats or informal measures, as it tries to use its leverage with member states like Germany to head off the imposition of EV duties and other EU measures designed to level the economic playing field. As it has in the past, China could impose targeted product bans or encourage consumer boycotts (aimed at French or Italian luxury goods or agricultural products, for example).

While conventional wisdom holds that German carmakers would be first in the firing line, China could, in an initial phase, turn to other targets. Volkswagen, BMW, and Mercedes-Benz have continued to invest heavily in China in recent years, including in technology partnerships with Chinese firms, in a bid to stay competitive in a fast-evolving EV market. At a time when it is struggling to rebuild trust among foreign investors, Beijing may decide that punishing those firms whose trust has never wavered is short-sighted. Still, because they are highly exposed to the Chinese market, the big German carmakers are susceptible to a range of coercive actions, including licensing revocations and crackdowns on data security and cybersecurity grounds.

Beijing has many other factors to consider as it weighs how to respond. China has been developing legislative tools of its own, such as the newly enacted Foreign Relations Law, to send a message about its right to strike back against what it perceives as unfair trade and investment restrictions (see August 17, Codifying Retribution). Its move earlier this year to introduce export controls on germanium and gallium was seen as a warning to Europe and others about China’s willingness to restrict access to critical inputs that it dominates. In an extreme scenario, China could restrict exports of critical materials that the world needs for EV manufacturing (for example, graphite, lithium, cobalt, and copper processing) and EV-related inputs and technologies like neodymium magnets and vehicle-mounted laser detection and ranging systems. Chinese regulators could also become more defensive of battery technologies and restrict outbound licensing deals in an attempt to keep investment in China. Policies like the Inflation Reduction Act, which seeks to reduce Chinese inputs in US EV supply chains, and trade defense moves like the EU’s EV probe, are both likely to steer EV-related investment away from China—a shift Beijing will be keen to counter.

Nevertheless, while retaliation in the EV space may be Beijing’s most powerful weapon, it could also end up turbocharging de-risking policies in Europe. Alienating foreign automakers and suppliers in the EV and autonomous driving space could inhibit China’s prospects of co-developing standards to stay integrated in global EV supply chains. And licensing restrictions on Chinese leading battery makers would hinder their ability to ride friend-shoring waves in auto clusters abroad. Ultimately, from a geopolitical perspective, China will be mindful that an escalating trade conflict with Europe could push capitals like Berlin and Paris closer to Washington. Beijing has economic leverage, particularly in green technologies, but the risks of weaponizing that leverage are high.


1. Share of EU demand is proxied using China’s share of combined intra-EU and extra-EU imports.

2. Drawing on a methodology developed by the Organization for Economic Cooperation and Development (OECD), we gathered micro-level insights into China’s subsidization of the BEV battery sector. We relied on public information about the amounts disbursed to companies for some (though not all) of these conventional and non-conventional policy instruments.

3. We’ve shown in previous work how the combination of a large and closed domestic market can act as a form of state support.

4. Battery makers include Ningbo Shanshan, Zhuhai CosMX Battery, Gotion High-tech, CATL, EVE Energy, and Hunan Corun New Energy. BEV makers include BYD, SAIC Motor Corporation, FAX Jiefang Group, Chongqing Changan Automobile, Beiqi Foton Motor, and Guangzhou Automobile.

Going Beyond Carbon: Closing the Non-CO2 Ambition Gap

At the upcoming COP28 climate summit in Dubai, the world will take stock of progress toward meeting the Paris Agreement goal of limiting warming to well below 2º Celsius. This Global Stocktake will be an opportunity to identify gaps in the collective effort to address global climate change and ways to enhance ambition. The Group of 20 (G20) highlighted one such gap in its most recent Leaders Declaration, noting many countries do not yet have economy-wide absolute emission reduction targets and calling for them to be reflected in nationally determined contributions (NDCs) “in light of different national circumstances.” This means reducing not just carbon dioxide (CO2), but also short-lived climate forcers like methane, hydrofluorocarbons, and other non-CO2 gases, which contribute roughly a quarter of global greenhouse gas emissions. To provide context on the contribution of these gases, in this note we assess the scale and key sources of non-CO2 emissions from some of the world’s largest greenhouse gas emitters—the G20 members themselves.

Global non-CO2 emissions

Efforts to mitigate greenhouse gas (GHG) emissions over the past few decades have largely focused on curbing CO2 emissions. Indeed, CO2 has contributed the most to warming experienced to date and contributes three-quarters of total global GHG emissions. Keeping warming below 2ºC also requires significant and sustained reductions of non-CO2 emissions as well. According to the Intergovernmental Panel on Climate Change, reducing emissions of short-lived climate forcers is “critical to meet long-term climate goals and might help reduce the rate of climate change in the short term.” Reductions in methane emissions are particularly critical to keeping the Paris Agreement goals within reach. Due to its much shorter lifetime, methane has a disproportionate impact on near-term temperature and is estimated to account for almost a third of the warming observed to date.

Globally, non-CO2 gases contribute roughly a quarter of GHG emissions today, but there is wide variation among countries. Among the G20 economies, non-CO2 shares of total GHG emissions range from only 9% in Japan to as high as 51% in Argentina (Figure 1). Most G20 economies’ non-CO2 shares are on par with the global average, though several countries have much higher-than-average shares, including Argentina, Brazil (44%), Russia (42%) and France (36%).


To account for methane and other short-lived climate forcers’ relatively short atmospheric lifetimes, a different metric (GWP20) can be used to compare non-CO2 gases to CO2 equivalents using a 20-year lifetime rather than 100 years (as in the conventional approach of using GWP100 values). The choice of metric, including time horizon, should reflect the policy objectives for which the metric is applied.[1] According to the IPCC, “all metrics have limitations and uncertainties, given that they simplify the complexity of the physical climate system and its response to past and future GHG emissions. No single metric is well-suited to all applications in climate policy.” When methane and nitrous oxide are considered on a 20-year horizon, non-CO2 gases represent a much larger share of total GHG emissions. Based on this metric, the share of non-CO2 gases rises to over 30% for the majority of G20 economies. Half of the G20 members have non-CO2 shares that exceed the global average of 46% non-CO2 based on a 20-year lifetime (Figure 2).


For some G20 economies, even if non-CO2 emissions represent a quarter or less of total emissions, on an absolute basis these emissions are quite sizeable (Figure 3). Even if you excluded China’s CO2 emissions entirely, its non-CO2 emissions would rank as the world’s third largest GHG emitter (using both 20- and 100-year GWP values). The non-CO2 emissions from the US and India would rank them as the world’s sixth and seventh largest emitters, respectively, and Russia’s non-CO2 emissions alone would rank it the world’s 10th largest emitter. The EU-27’s non-CO2 emissions would earn the rank of 11th largest emitter, while Brazil’s would land at 13th.


When considering a 20-year lifetime for methane (CH4) and nitrous oxide (N2O), the total CO2 equivalent of non-CO2 emissions jumps considerably (Figure 4).


Methane is a majority

Methane is by far the largest source of non-CO2 emissions, contributing 69% of all non-CO2 gases. In all but two G20 economies, methane makes up the majority of non-CO2 emissions. In some countries (e.g., Russia, Brazil, Indonesia), methane makes up more than three-quarters of total non-CO2 emissions (on a GWP100 basis). In 2021 at COP26, a coalition of countries launched the Global Methane Pledge with the aim of reducing global methane emissions at least 30% from 2020 levels by 2030. Since then, over 150 countries have signed the pledge to contribute to collective action to reach the 2030 goal. All G20 member states—with the exception of China, India, Russia, and Turkey—have signed the pledge.

The two primary sources of methane emissions are 1) agriculture, land use, and waste; and 2) industrial sources including the production and use of fossil fuels. For most G20 member economies, the majority of methane emissions come from agriculture, land use, and waste. Several fossil fuel-producing members—including China, the US, Russia, Canada, Australia, and Saudi Arabia—have sizeable methane emissions from the fossil fuel industry.


Closing the non-CO2 ambition gap

At the conclusion of the first Global Stocktake at COP28 in Dubai this year, the world should walk away with a clear sense of what remains to be done to meet the Paris Agreement goals of limiting warming to well below 2ºC. One critical next step is to enhance the ambition of countries’ nationally determined contributions (NDCs). As the Paris Agreement itself lays out, each country’s successive NDC should represent a progression beyond their earlier commitments, reflecting their highest possible ambition. More specifically, Article 4.4 provides that “Developed country Parties should continue taking the lead by undertaking economy-wide absolute emission reduction targets. Developing country Parties should continue enhancing their mitigation efforts, and are encouraged to move over time towards economy-wide emission reduction or limitation targets in the light of different national circumstances.” To date, the NDCs of two G20 member states—India and China—do not include coverage of non-CO2 gases. Given the significant contribution of emissions of non-CO2 gases to global GHG emissions and their meaningful short-term impact on warming, integrating absolute reductions of non-CO2 emissions into countries’ NDCs (for those that do not currently) and adopting progressively ambitious absolute reduction targets that include non-CO2 reductions in current and future NDCs will be critical to closing the ambition gap and putting the world on track to meet the Paris Agreement long-term goals.

[1] According to the IPCC, the GWP time horizon can be linked to the discount rate used to evaluate economic damages from each emission. For methane, GWP100 implies a social discount rate of about 3–5% depending on the assumed damage function, whereas GWP20 implies a much higher discount rate, greater than 10%.

Direct Air Capture Workforce Development: Opportunities by Occupation

Direct air capture (DAC), a carbon dioxide removal solution that captures carbon dioxide directly from the atmosphere, has the potential to play a pivotal role in meeting the US goal of net-zero greenhouse gas emissions by 2050. Alongside the scale-up of clean electricity generation, electrification of end-uses, and other emissions abatement strategies, DAC could be a complementary technology that allows for emissions offsets from hard-to-abate sectors, and that can remove emissions that already exist in the atmosphere. Over the past few years, the Energy Act of 2020, the Infrastructure Investment and Jobs Act (IIJA), and the Inflation Reduction Act (IRA) have ushered in a wave of DAC policy support in the United States. Last month, the Department of Energy announced DAC hub funding (a program incentivized by IIJA) to support the development of two commercial-scale DAC facilities and 19 early-stage projects across the US.

Assuming this momentum continues and policy support is extended long-term, DAC deployment at scale will require a large, well-trained workforce to build and operate DAC facilities. In this note, we assess the job creation and workforce development benefits associated with a commercial-scale DAC facility. We find that the construction and engineering of a DAC plant creates 1,215 annual average jobs over the roughly five-year time period it takes to build the facility. After the plant is built, we estimate there are approximately 340 jobs needed to operate the facility over its lifetime. We also dig into the types of occupations that will benefit from DAC scale-up and find that the industry will support a diverse set of construction trades, maintenance workers, and business operators.

Scaling the direct air capture industry

Direct air capture (DAC) is a technology similar to traditional carbon capture that captures carbon dioxide (CO2) from ambient air instead of an industrial or electric power flue gas stream. Because ambient air concentrations of CO2 are much more dilute than flue gas, the costs tend to be higher for DAC than traditional carbon capture. That said, unlike traditional carbon capture, the main utility of DAC is that it results in net-removal of CO2 from the atmosphere—not just abatement like traditional carbon capture. In a 2019 Rhodium report on advancing direct air capture, we estimate that the US needs to scale up to between 690 and 2,260 million metric tons (MMT) of annual DAC capacity in order to decarbonize the US by 2050 (Figure 1). The range of DAC capacity needed represents how quickly we can scale other decarbonization strategies such as energy efficiency, electrification, decarbonization of the electric power system, and the availability of other carbon dioxide removal (CDR) systems. The slower other decarbonization strategies are in scaling up, the more DAC the US will need to meet its goal of net-zero emissions by 2050.


Since we released our 2019 report, unprecedented policy support along with a surge in DAC investment has fundamentally changed the landscape for DAC. For example, as part of their pledge to invest one billion dollars in CDR, Microsoft has purchased carbon removal from three major DAC companies thus far—Climeworks, Heirloom, and CarbonCapture. Most recently, the Department of Energy (DOE) announced funding site selections for two major commercial-scale DAC hubs—one in Texas and the other in Louisiana—that will together have the capability to remove at least two million metric tons of CO2 once fully operational. In addition to these plants, 19 other potential DAC sites across the US were selected to receive DOE funding that assists with early-stage project development. In total, the Infrastructure Investment and Jobs Act (IIJA) passed in 2021 provides roughly $3.5 billion for four commercial-scale regional DAC hubs, $100 million for a commercial DAC prize, and another $4 million for a pre-commercial DAC prize. In addition, the 45Q tax credit enhancement in the Inflation Reduction Act (IRA), passed last year, allows DAC developers to receive $180 per ton of CO2 sequestered. Both pieces of legislation include elements that help to boost DAC deployment and build out supporting infrastructure like geologic storage and CO2 transportation pipelines. It’s also worth noting that the Energy Act of 2020 reorganized the Office of Fossil Energy at DOE to be inclusive of carbon management technologies and, subsequently, set up the first CDR research and development (R&D) program.

In the near term, this policy support creates the economic conditions necessary to establish a true market for DAC. We estimate that in total, these policies will drive 1 to 8 MMT of DAC capacity deployment in 2030 and 5 to 84 MMT of deployment by 2035 (Figure 1). These ranges represent the speed at which the supporting industries such as construction, engineering and manufacturing can scale to support the DAC industry as well as uncertainty surrounding regulatory and permitting reform. Though this is momentous for an industry that was hardly heard of five years ago, the direct air capture industry still has a long way to go to achieve gigaton scale by mid-century. Ultimately, longer-term policy support, private investment, and workforce development will be needed. As a starting point, we explore the opportunities for workforce development from scaling up the DAC industry.

Direct air capture workforce opportunities

Large-scale deployment of DAC has the potential to generate significant economic and employment opportunities in the communities where plants are sited and in the industries that support DAC construction and operation. For this analysis, we evaluated a range of DAC technological processes assuming they are at commercial scale. The results provided below reflect the median values across all DAC processes. Because no large-scale commercial DAC plants exist today, there is no empirical employment data to use in this analysis. We have developed a methodology that proxies DAC employment based on adjacent industries. To do this, we determine the capital and operating costs of each process based on expert interviews, literature, and public announcements. Next, we conduct in-depth research on how the distribution of costs varies by process and input those costs into the appropriate industry associated with the input-output model IMPLAN. We produce occupational results from IMPLAN and supplement these outputs with Bureau of Labor Statistics data.

We estimate the total number of jobs created from the construction and operation of a single commercial DAC facility with an annual capture capacity of 500 kilotons (kt) because that will be the size of the first large-scale commercial DAC facility in the US. For reference, this is a plant size that captures emissions equivalent to about half those produced annually by an average-sized natural gas combined-cycle power plant.


We find that a commercial 500 kt DAC facility generates, on average, 1,215 plant investment jobs each year over the course of the facility’s five-year construction period (Figure 2). Plant investment refers to the jobs associated with construction, engineering, materials, and any equipment needed to build the DAC facility. 63% (765 jobs) of these jobs are considered facility-level jobs in that they are directly related to the construction of the DAC facility. All remaining plant investment jobs support supply chain activities like freight transport and equipment manufacturing.

Ongoing operations and maintenance (O&M) over the lifetime of the facility generates an additional 340 jobs, with roughly 65% (220 jobs) being on-site primarily maintaining the DAC equipment over the facility’s lifetime.

Direct air capture plant construction occupations

For the first time, we take a deeper dive into the types of jobs associated with the construction and operation of a commercial-scale DAC facility. We exclude occupational detail on supply chain jobs from this analysis. The top five occupational categories associated with the construction of a typical DAC facility are: (1) construction trades; (2) metal workers and assemblers; (3) engineers; (4) executives and business operations; and (5) machinery installers, maintenance, and repairers (Figure 3). Collectively, they account for roughly 63% of all non-proprietor employment associated with the facility’s construction. Non-proprietor employment refers to wage and salary workers—in other words, employees who are not owners.


Construction trades are the largest occupational category, followed by metal workers and assemblers, and engineers. Metal workers include welders, solderers, and machinists. Civil, mechanical, industrial, and electrical engineers, along with engineering technicians, make up the third largest group of occupations associated with building a DAC plant.

Since the construction trades represent such a diverse set of skills and they are responsible for a large subset of the DAC plant construction workforce, we’ve provided more detail for this particular occupation group (Figure 4). 40% of construction trade jobs required to build a DAC plant are either general laborers or managers. Carpenters and electricians account for 26% of construction trade jobs, and wall and flooring workers account for 14%. Plumbers, pipelayers, masons, and roofers account for the remaining construction trade labor associated with building a DAC plant.


Direct air capture plant operation occupations

In the case of ongoing jobs associated with operations and maintenance of a DAC plant, the top occupations are machinery installers, maintenance and repairers, which includes industrial mechanics and industrial line supervisors. The next largest categories of occupations include executive and business operations occupations, metal workers and assemblers, production occupations and freight movers (Figure 5).


This first-of-its-kind occupational analysis provides insight into not only the total number of jobs but also the types of jobs created from the construction and operation of a commercial DAC facility. Given the current state of the industry, our occupational analysis should be seen as a first-order estimate, but we expect the overall takeaways to be the same as we continue to improve our analysis. As the DAC industry scales, we know that there will be a huge demand for a workforce to build and operate the plants. That workforce represents a diverse set of occupations that exist today. Even so, many of these occupations will require technical and extensive on-the-job training. To prepare for a scale-up of these emerging technologies, occupational training programs will be crucial. Planning will be key to make sure that the DAC workforce is abundantly available and trained in the locations where developers are looking to build DAC plants.

Forthcoming analysis

This analysis is in a three-part series assessing the workforce development opportunities from scaling up emerging climate technologies in the US. The first part, looking at the clean hydrogen industry, is available here, and our third part on the sustainable aviation fuel industry is forthcoming.

In addition, we plan to refine our approach and understanding of the DAC workforce as facilities get built and more experience and innovation on DAC technology occurs with scale-up. We will also continue to expand upon our employment analysis to investigate important factors including wages, union labor, and required skills. Additionally, later this year we will be releasing an analysis assessing opportunities for building up the DAC industry within individual states, including the associated economic and employment benefits. Finally, we are planning on a full employment analysis for the selected DAC hubs that we’ve discussed in this note.

This nonpartisan, independent research was conducted with support from Breakthrough Energy. The results presented reflect the views of the authors and not necessarily those of supporting organizations.

Clean Hydrogen Workforce Development: Opportunities by Occupation

The United States is positioning itself to be a leader in clean hydrogen production thanks to a wave of public policy support under the Inflation Reduction Act and the Infrastructure Investment and Jobs Act. However, the US will not be able to achieve its goal of significantly scaling up clean hydrogen without a robust and effective workforce. A clean hydrogen economy will require skilled laborers in an array of occupations.

In this note, we explore potential job opportunities from building up the clean hydrogen industry in the US. We assess the total number of jobs associated with example individual commercial-scale facilities, as well as the types of occupations that comprise these employment numbers. Building a commercial-scale electrolytic hydrogen facility is associated with an annual average of 330 plant investment jobs and 45 ongoing jobs. Moreover, retrofitting a traditional hydrogen facility with carbon capture is associated with an annual average of 520 plant investment jobs and 80 ongoing jobs. Many of these occupations will require technical and extensive on-the-job training. To prepare for an expansion of these emerging technologies, occupational training programs will be imperative.

Policy support for clean hydrogen continues to grow

In recent years, clean hydrogen has become the focus of increased international attention as a versatile tool for deep decarbonization. In many ways, the United States is positioning itself as a leader in this space. In 2021, the Department of Energy announced its Hydrogen Shot goal to bring the cost of clean hydrogen down to $1/kg H2 in the next ten years. If achieved, this would make clean hydrogen price-competitive with today’s conventional—and emission-intensive—means of hydrogen production via steam methane reformation (SMR), often referred to as gray hydrogen.

In line with this goal, the Biden administration and congressional leaders incorporated support for clean hydrogen into two major clean energy bills—the Inflation Reduction Act (IRA) and the Infrastructure Investment and Jobs Act (IIJA). The IRA includes a clean hydrogen tax credit, 45V, which can bring down the cost of clean hydrogen by up to $3/kg H2 depending on the carbon intensity of the production pathway. Rhodium analysis has found that this incentive can drive down costs and accelerate long-term adoption of clean hydrogen both in the United States and abroad. Additionally, the IRA increased the value of the 45Q tax credit for carbon capture to $85/ton CO2 sequestered. This provides hydrogen from SMR with carbon capture, commonly known as blue hydrogen, with an alternative economic incentive to 45V if a developer prefers to claim it. Moreover, the IIJA includes $8 billion in funding for a clean hydrogen hubs program. This funding will go towards establishing six to ten clean hydrogen facilities across the country to help accelerate adoption in the US energy system. It’s anticipated that these hub locations will be announced towards the end of the year.

Demand growth for clean hydrogen is exponential

While numerous aspirational use cases exist for clean hydrogen in the long-term, the preeminent near-term end use is to replace existing demand for conventional hydrogen in the industrial sector with this low-carbon alternative. This can be done by lowering the carbon intensity of existing facilities by installing carbon capture, and by building new production facilities for electrolytic hydrogen that can potentially displace conventional hydrogen. Electrolytic hydrogen is hydrogen produced from water via electrolysis powered by clean energy, often referred to as green hydrogen.

Under current policies including the IIJA and the IRA, Rhodium estimates up to 16 million metric tons of carbon capture capacity will be installed on existing conventional hydrogen facilities by 2035. We also project between 15.3 to 23.2 GW of electrolytic hydrogen deployment in 2035. The range reflects uncertainty around infrastructure scale-up and technology costs. This is roughly 1,000 to 1,500 times the current electrolyzer capacity in the US and is enough capacity to produce approximately 2 to 3 million metric tons of clean hydrogen. By comparison, the US currently produces roughly 10 million metric tons of conventional hydrogen. Reaching those projections will require an 80% annual average increase in electrolytic hydrogen capacity over the next 12 years. Though that level of scale-up is ambitious, this type of growth is not unprecedented in the clean technology space. For comparison, utility-scale solar grew by an average of 97% per year during its fastest 10-year scale-up.

Under an array of scenarios for meeting net-zero emissions by mid-century, Evolved Energy Research finds that the US will need to produce between 25 to 140 million metric tons of clean hydrogen in 2050. Depending on the scenario, they find that 13 to 136 million metric tons of this hydrogen is from electrolysis (Figure 1). To achieve this high level of deployment, further policy support will be required.

Figure 1

Workforce readiness is part of the clean hydrogen scale-up challenge

The US will not be able to achieve the dramatic scale-up of clean hydrogen required to reach net-zero emissions by mid-century without the proper workforce. A clean hydrogen economy will require skilled laborers in discrete occupations. Below we explore the total jobs associated with an initial commercial-scale facility as well as the types of occupations that comprise these employment numbers.

For this analysis, we determined each production method’s capital and operating costs based on expert interviews, literature, and Rhodium analysis. The costs are informed by the size of the facilities and technology maturity. Next, we conducted in-depth research on how the distribution of costs varies by production method and sorted these costs into the appropriate industry codes associated with the input-output model IMPLAN. To estimate employment numbers, we input Rhodium’s cost data into IMPLAN using its national-level tools. We pulled occupational results from IMPLAN and supplemented these outputs with Bureau of Labor Statistics data. Our job numbers capture both the on-site and off-site jobs behind installing and maintaining a clean hydrogen facility, as well as the upstream supply chain jobs. We distinguish job types into two categories: plant investment jobs associated with the construction, engineering, materials, equipment, and supply chains required to build the facility; and operations & maintenance jobs, which are ongoing jobs over the lifetime of the plant.

This analysis accounts for the recent economic shifts, including rising inflation numbers, supply chain constraints, and overall higher capital costs compared to our previous analysis. We also assume larger hydrogen facilities to keep up with the rapidly evolving hydrogen landscape. This section is separated by hydrogen production method since each facility type has a distinct job profile.

Electrolytic hydrogen workforce opportunities

For our analysis, we use an example 100 MW size facility, which is based roughly on an expected initial commercial-scale facility seen in current literature and announced project plans. While a 100 MW size plant is much larger than any existing facility in the US today, it is still relatively small compared to where the industry is heading. Developers and manufacturers hope to eventually increase electrolyzer plants to the GW scale, at least 10 times the size of this sample initial facility. We find that a 100 MW polymer electrolyte membrane (PEM) hydrogen plant is associated with an annual average of 330 plant investment jobs over the 2-year construction period (Figure 2). 150 of these jobs are associated with the construction, engineering, equipment, and materials to build the facility, and 190 come from the supply chain requirements. Additionally, we see 45 ongoing jobs over the lifetime of the plant associated with operating and maintaining the facility and related supplier activities.

Figure 2

Job breakdown by occupation

To gain a better understanding of the skilled labor needed to fill these positions, we analyzed the underlying occupations that make up the plant investment and O&M jobs. Supplier jobs are not considered in our occupational analysis. The top five occupations associated with building a new electrolytic hydrogen facility are, in order: (1) metal workers and assemblers, (2) legal workers, (3) engineers, (4) executive and business operations, and (5) production occupations (Figure 3). The top category—metal workers and assemblers—primarily consists of skilled workers including welders and machinists. Electrical, industrial, and mechanical engineers are the largest subcategory of engineers required to build a PEM facility. Production occupations include jobs such as inspectors, testers, processing technicians, and chemical equipment operators and tenders.


The top five occupations for ongoing O&M jobs are (1) installers, maintenance, and repairers, (2) production occupations (3) executive and business operations, (4) engineers, and (5) plant system operators (Figure 4). The top job category—installers, maintenance, and repairers—includes jobs maintaining and repairing the necessary electrical systems and machinery.


Hydrogen with carbon capture workforce opportunities

As more carbon capture projects break ground, we have noticed a trend towards companies reporting higher costs for carbon capture than previous literature had estimated. These changes are reflected here in our employment analysis. To estimate the jobs associated with a carbon capture retrofit of a conventional hydrogen production facility, we use a facility with 500 kilotons of annual CO2 capture capacity as representative of cost and employment trends for the broader industry. We estimate an average of 520 jobs per year are associated with adding carbon capture to an SMR facility of this size over the 4-year construction period (Figure 5). 220 of these jobs are associated with the construction and engineering of a retrofit project and the remaining 300 jobs are associated with materials and equipment used in the project via the supply chain. Moreover, we estimate 80 ongoing jobs over the lifetime of the facility to operate and maintain the carbon capture equipment. These estimates do not include ongoing jobs retained at the hydrogen production facility prior to the installation of carbon capture.


Carbon capture retrofits will happen where there are existing SMR facilities. Thus, much of this workforce opportunity is expected in the Gulf Coast, Midwest, and California.

Job breakdown by occupation

We also analyzed the underlying occupations that make up the plant investment and O&M jobs for hydrogen with carbon capture. Supplier jobs are not considered in our occupational analysis. The top five occupations associated with retrofitting a hydrogen facility with carbon capture are: (1) construction trades, (2) metal workers and assemblers, (3) executive and business operations, (4) engineers, and (5) machinery installers, maintenance, and repairers (Figure 6).


Construction trades include a variety of occupations, including construction laborers and managers, carpenters, electricians, plumbers, and pipelayers (Figure 7). Metal workers and assemblers include occupations such as welding, soldering and electrical equipment assembling. Mechanical, civil and industrial engineers represent the bulk of engineers required to retrofit conventional hydrogen production with carbon capture.

Figure 7

In terms of ongoing jobs, the top five occupations are: (1) machinery installers, maintenance, and repairers, (2) metal workers and assemblers, (3) executives and business operations, (4) freight movers, and (5) production occupations (Figure 8).


Reaching our estimate of up to 16 million metric tons of blue hydrogen capture in 2035 would require 32 facilities of this size—28% of hydrogen facilities reporting to EPA’s facility level data. This translates to over 16,000 annual average plant investment jobs and 2,560 ongoing jobs across the industry.

Across all types of clean hydrogen production, many of these occupations will require technical and extensive on-the-job training. To prepare for a scale-up of these emerging technologies, occupational training programs will be imperative.

Forthcoming analysis

With clean hydrogen being a critical tool of a clean energy future, Rhodium Group will continue to follow this topic closely. We have forthcoming analysis exploring the economic opportunities associated with the Department of Energy’s hydrogen hubs. Additionally, as a continuation of our emerging climate technologies research, Rhodium Group will be releasing employment and occupational analysis for direct air capture and sustainable aviation fuels soon. We will also continue to expand upon our employment analysis to investigate important factors including wages, union labor, and required skills.

This nonpartisan, independent research was conducted with support from Breakthrough Energy. The results presented reflect the views of the authors and not necessarily those of supporting organizations.

Global Greenhouse Gas Emissions: 1990-2021 and Preliminary 2022 Estimates

Each year, Rhodium Group provides updated estimates for annual greenhouse gas (GHG) emissions at the global and every-country levels, including for all six main greenhouse gases and across all sectors of the economy. This regular tracking of greenhouse gas emissions provides valuable data for decision-makers in both the public and private sectors as they develop strategies to meet global net-zero emissions goals.

In our preliminary estimates for 2022, we find that global GHG emissions rose 1.1% from the previous year, to 50.6 gigatons of CO2-equivalent. Global emissions dropped steeply in 2020 due to the economic disruptions of the COVID-19 pandemic, followed by a nearly-equivalent rebound in 2021. In 2022, the rise in emissions was much smaller, but nonetheless emissions rebounded back to above pre-pandemic levels and reached a new high. Disruptions in global energy markets from the war in Ukraine resulted in a sharp decrease in natural gas consumption and growth in renewable energy consumption in several major economies in 2022, offset in part by increased coal consumption.

Preliminary estimates show emissions surpassed pre-pandemic levels in 2022

In 2022, just two years after the most marked decline in global GHG emissions seen in three decades, emissions rebounded back above pre-pandemic levels. Rhodium’s preliminary estimates indicate a 1.1% increase in global emissions—covering the six main greenhouse gases emitted from every sector of the economy, including land use, forestry, and international bunkers—from 50.1 gigatons of CO2e in 2021 to 50.6 gigatons in 2022 (Figure 1).

Seven economies were responsible for close to two-thirds of global emissions in 2022 (Figure 3). Relative emissions shares among the world’s major economies remained the same as previous years. China was the highest-emitting economy, contributing 26% of global emissions, followed by the US at 12%, and the EU and India each at 7%.

The vast majority of global GHG emissions are produced as CO2 from the combustion of fossil fuels (here referenced as “energy CO2”), which to date have powered the global economy. In 2020, energy CO2 declined 4.8% worldwide, with some major economies seeing reductions of as much as 12.7% (Figure 4). In 2021, global energy CO2 bounced back by 4.7%, with several major economies—India, China, Brazil and Russia—exceeding their pre-pandemic levels. In 2022, emissions from the combustion of fossil fuels continued to rise in India and the US, though US emissions did not exceed pre-pandemic levels. Energy CO2 remained relatively flat in Japan and China, while the EU, Brazil, and Russia each saw emissions decline. Russia experienced the most pronounced drop in energy CO2 with an 8.6% reduction, which can be largely attributed to international sanctions following its invasion of Ukraine, which contributed to a 14% reduction in natural gas production and a 7% decrease in coal production in 2022.

Digging deeper into the dynamics affecting energy CO2 emissions, the impact of the war in Ukraine significantly disrupted global fuel markets, inducing inflationary pressures worldwide and altering CO2 emission sources among major economies. As the second-largest natural gas exporter, Russia’s sanctioned trade impacted the global gas market. As a result, many of the world’s major economies saw a sharp decrease in natural gas consumption, leading to a 3% drop in global natural gas consumption in 2022. Nowhere was this more evident than in the EU, which reduced its consumption by 14% in 2022 (Figure 5). In Brazil, demand for natural gas dropped 21% as higher hydropower availability cut gas-fired power generation. Within Russia—a country that relies on natural gas for 55% of its total fuel consumption—gas demand dropped 14% last year.

Many countries that have traditionally relied heavily on natural gas were forced by high prices to turn to more carbon-intensive fuels like coal, demand for which increased 0.6% in 2022. China and India—both net natural gas importers—reported increases in coal usage by 1% and 4%, respectively. However, the US diverged from this trend. Coal supply chain disruptions in the US pushed for an increased reliance on both natural gas and renewables to meet energy demand.

Amid a global surge in natural gas prices, many nations pivoted to renewable energy to meet their energy needs. Excluding hydropower, worldwide consumption of renewable energy rose by 13%. Among the major economies, Russia led with a 31% jump in renewable consumption. China and India followed with an 18% increase, while the US registered a 13% rise. Collectively, this shift from natural gas to renewables significantly reduced the carbon intensity of the global economy.

Emissions by sector: Industry and electric power generate more than half of global emissions

In 2021, the most recent year with detailed sectoral data, industry and the electric power sector were neck-and-neck in their contribution to global GHG emissions, together constituting nearly 60% of the global total (Figure 6). The industry and the electric power sector each accounted for 29% of total GHG emissions, with the vast majority coming from coal combustion. Land use, agriculture, and waste collectively contributed 20% (though this excludes the impact of forest fires), while transportation and buildings contributed 15% and 7%, respectively.


The data described here are available on Rhodium’s ClimateDeck data platform, with support from Breakthrough Energy. The ClimateDeck features GHG emissions and energy data for all 190+ countries in the world and all 50 US states and provides users the ability to filter by region, GHG, sector and sub-sector, as well as socioeconomic indicators (e.g., emissions per capita and per GDP), and full inventory tables for each country.

These estimates differ slightly from previous versions of our global GHG emissions data due to methodological improvements and updates that incorporate new data sources. The most significant changes from our 2022 report are due to: 1) use of Global Warming Potentials (GWP) of non-CO2 gases from the Fifth Assessment Report (AR5) of the IPCC, in alignment with UNFCCC reporting guidelines; and 2) the transition toward third party datasets for Agriculture, Forestry and other Land Use (AFOLU) and energy CO2 to ensure methodological consistency across all countries (which means our data will not necessarily align with self-reported UNFCCC inventory data).

In deriving our global and country-specific GHG emissions estimates, we incorporate data from IEA’s energy consumption flows and BP’s 2023 Statistical Review for fossil fuel CO2. For emissions from AFOLU and waste, our primary data source is FAOSTAT and our preliminary estimates for 2022 also incorporate projections from IIASA’s GLOBIOM 2021. These estimates do not take into account emissions from forest fires. For all other non-CO2 gases, we use Annex I inventories and EPA’s Global Non-CO2 data.

For more information about the ClimateDeck, please email

Irrational Expectations: Long-Term Challenges of Diversification Away from China

In the first half of 2023,  Mexico overtook China as the main trading partner of the United States—the first time since 2005 that China did not hold the top spot. Some observers seized upon this development as proof that diversification away from China was well underway. Others cautioned against jumping to conclusions, pointing out that Mexican imports continue to rely heavily on Chinese inputs. For the diversification skeptics, global value chains are simply too integrated to untangle and the Biden administration’s de-risking policies are unlikely to succeed.

Which camp is right? There is almost as much data to support the idea that firms are diversifying away from China as there is showing that they are doubling down on the Chinese market. Reaching a verdict is further complicated by the fact that there is no consensus on how to define diversification—a concept that encompasses not only trade and investment but also supply chains and tougher-to-measure concepts like knowledge-sharing. Still, as companies around the world reassess their links to the Chinese market in response to an increasingly challenging business environment in China and rising geopolitical tensions, business leaders need to get a clearer sense of whether diversification is happening and what form it is taking. This note explores these issues using a wide range of available data points. We find that:

  • Firms—foreign and Chinese—are actively diversifying investment and sourcing away from China, in sensitive as well as less sensitive sectors. Because channeling new investments to new markets is easier than finding alternative suppliers, particularly for intermediate inputs, the trend is most visible in the FDI realm.
  • Because global value chains are so entangled with China, diversification will not necessarily result in a reduced reliance on Chinese inputs and suppliers in the short- to medium-term. A more comprehensive relocation of global supply chains is likely to unfold eventually, especially in sectors like EVs where co-localization is key, but this will happen over an extended period of time.
  • Because China’s manufacturing clout is so significant, even substantial shifts in production to alternative destinations may only result in small declines in China’s share of global exports, manufacturing or supply chains. These moves, however, can trigger major economic, industrial, and logistical impacts on new destination countries, given their much smaller sizes.
  • These realities mean that it will take years for advanced economies to achieve the objectives behind their “de-risking” policies—namely limiting trade, investment and supply chain exposure to China in a subset of critical areas. This does not mean that the broader diversification objective is misguided. But policymakers will need to adjust their expectations about the timeframe needed to substantially reduce dependencies. If they do not accept that diversification is a complex, long-term challenge, there is a risk that calls for bolder policy measures to reduce dependencies on China grow.

Diversification is well underway

A wealth of anecdotal evidence on investment, trade and supply chain patterns suggests that some degree of diversification from China is already underway. Business association surveys—early indicators of shifting sentiment—show that the appetite for making new investments in China is waning. They suggest that a growing share of EU, UK, US, and Japanese firms do not plan to increase investment in the country over the coming years (with German firms a notable exception, see Figure 1). One in five German and UK firms, and one in ten US firms, now say that they plan to decrease investment in China. And a growing number of firms from advanced economies say they are relocating or intend to move manufacturing or sourcing outside of China, most often to ASEAN (especially Vietnam) and near-shore destinations (Central and Eastern Europe for EU firms, Mexico for US ones).

Figure 1
Figure 2

The shift in corporate intentions toward China, evident in recent surveys, can also be seen in investment data. Our tracking of US and EU FDI into China shows that the number of FDI transactions in China by EU and US firms is trending down, and is now roughly half of levels seen a decade ago.1 Annual investment values are well below the levels seen in the early 2010s (see “Big Strides in a Small Yard” August 11, 2023). This compares with robust (and fast-growing in 2022) greenfield FDI to alternative destinations like Vietnam, Malaysia, Mexico, and especially India (Figure 2). These trends have resulted in a decline in China’s share of global greenfield FDI since 2004, a trend that has accelerated over the past five years, with that share declining from 11% in 2018 to just below 5% in 2021. This occurred despite the fact that China’s share of global GDP increased over this period, reaching 19% in 2022 (Figure 3).

Diversification in firms’ investment decisions alone is not necessarily enough to impact trade patterns, given that a majority of China-world trade today is driven by Chinese rather than foreign firms (72% of China’s exports as of March 2023). This is clear when looking at EU-China trade ties, which have deepened in the past three years, despite EU firms showing waning interest in China investment. But where policies have raised the costs of bilateral trade, MNCs have also started to seek alternative sourcing locations, especially for assembled goods, with a much more visible impact on trade patterns (Figure 4). Such policies have led to a clear diversification trend in US-China trade, with China’s share of US imports declining from 22% in 2018 to about 17% in 2022—much of it linked to reciprocal US-China tariffs imposed under the Trump administration.2 Japan, an early mover in implementing policies to encourage supply chain diversification, has also achieved marked change in its trade relationship with China. Apart from a short rebound during the pandemic, China’s share of Japanese imports dropped from 26% in 2016 to 21% in 2022.

Figure 3
Figure 4

Geopolitical dynamics have played a key role in influencing corporate sentiment toward China in recent years, with tariffs, export controls and other politically motivated restrictions on economic activity raising the costs associated with investment and trade in a range of sectors. Geopolitics has also influenced the debate in corporate boardrooms, raising the hurdles to capital expenditure plans in China and bolstering the attractiveness of alternative locations. Diversification is therefore most evident in geopolitically sensitive sectors. Annual investments in the information communications technology (ICT) sector in China by US firms plummeted from over $4 billion in 2016 to just over $2 billion in 2019, and to a couple hundred million dollars in 2020. As scrutiny of the semiconductor sector grew, chipmakers announced plans for major investments in the US and Europe, as well as alternative locations like India. As a result, China’s share of global inbound FDI in semiconductors plummeted to 1 percent in 2022, from 48 percent in 2018. Meanwhile, the US share rose to 37 percent from zero percent over the same period, and the combined share of India, Singapore, and Malaysia rose to 38 percent from 10 percent (Figure 5). Electric vehicle (EV) and battery investment intentions in the US, finally, have spiked in the wake of the Inflation Reduction Act (IRA), with $52 billion in related FDI announced within the first nine months of the legislation’s publication.

Figure 5

But geopolitics is not the only driver of diversification choices, and behaviors are changing across a much broader set of sectors. A deteriorating business environment has also driven firms in highly restricted sectors, from digital services to energy, to downgrade their expansion plans and refocus resources on other geographies. Rising competitive pressures have had an additional effect, with companies such as Samsung Electronics, LG, and Stellantis exiting the Chinese market amid declining market shares, and announcing large investment projects in India, Vietnam, Mexico, and other destinations. Slowing economic growth and rising production costs in China are increasingly important factors for companies as well, as they reassess their presence in the Chinese market. Rising costs were tolerable when China was delivering GDP growth rates in the high single digits and the business environment was relatively stable. They are harder to ignore when growth is in the low single digits and the market is highly politicized. All in all, these data points paint a picture of global MNCs actively diversifying away from China, due to a combination of geopolitical and commercial considerations.

A slow and China-linked diversification

Then why is there such an intense debate about whether or not diversification is happening? The reason is that, while individual companies are reassessing their approach to China, the country remains a vital hub for global manufacturing value chains. Diversification, especially of sourcing activities, will unfold gradually. Barring an acute geopolitical crisis, a clear decline in China exposure will take years. It would not be surprising to see China’s overall share of global exports, manufacturing and supply chains continue to rise, even as diversification away from China accelerates.

The evolution of labor-intensive industries, early movers out of China, illustrates this counterintuitive dynamic. Diversification has been taking place for years in these industries, but it has been obscured by a concomitant expansion of manufacturing value chains in China. Textiles, for example, were an important driver of Chinese export growth for years. Over the past decade, however, the share of Chinese textiles in US and Japanese imports has fallen sharply (by 15 and 20 percentage points respectively), replaced by alternative countries of origin like Vietnam or Bangladesh. At the same time, FDI in China’s textile sector has also fallen rapidly, from $1.2 billion in 2012 to less than half a billion in 2019, according to MOFCOM data. Yet these shifts are barely visible in the aggregate, with China’s share of global labor-intensive manufacturing exports holding steady at around 30% (Figure 6).

This can be explained by China’s size and manufacturing clout. Today, the country represents 31% of global manufacturing value-added. As a result, even seemingly significant FDI flows to Vietnam and Bangladesh end up shifting only small fractions of China’s output and export capacity offshore. Vietnam’s population, after all, is only about 7% of China’s, and its manufacturing sector is a mere 2%. In addition, while the shift in textiles has been clear-cut, countless other labor-intensive sectors, from toys to light consumer electronics, have remained in China thanks to its infrastructure, logistics, and scale advantages, a relatively stable business environment, low outbound shipping costs, and low-cost inland provinces, where foreign investment more than tripled between 2012 and 2021. Finally, much of the decline in foreign investment in China was compensated for by Chinese firms growing manufacturing capacity and market share.

This counterintuitive dynamic is likely to play out in other sectors as well. Firms in the electronics, ICT, autos, or machinery sectors—both foreign and Chinese—may have started the process of diversifying away from China, but it will take years for this trend to be clearly visible in aggregate data and produce a clear decrease in micro- and macro-level exposure to China.

Figure 6
Figure 7

A second reason why diversification is unlikely to result in a rapid decline in exposure to China is that increased investments in, or trade with, alternative countries are often accompanied by a deepening of economic ties between these destinations and China. Over the past 20 years, China has onshored not just final assembly capacity, but also the deep value chains attached to them (Figure 7). Replicating these outside of China, or finding alternative supply sources, takes time, comes at a high cost, and is hindered by a lack of alternatives. Those MNCs that have sought to reduce their China presence or rebalance their global footprint have therefore done so in a piecemeal manner, shifting out single supply chain segments—often final assembly—to alternative destinations, while diversifying suppliers for a subset of the most critical inputs.

As such, diversification away from China does not always result in a meaningful reduction in firms’ reliance on China or Chinese firms—especially in the short-term. There are numerous examples of MNCs following Chinese suppliers to other markets or encouraging them to relocate outside of China. Many continue to source from China suppliers with whom they have developed close, long-term relationships even as they set up new plants in ASEAN or Mexico. In fact, Chinese outbound manufacturing patterns indicate that Chinese firms have been at the forefront of the global diversification wave (Figure 8). No wonder then that rising US imports from Mexico and Vietnam over the past 5-7 years were matched extremely closely by an increase in Chinese exports to these markets. This stickiness in relation to Chinese inputs is also visible at the sectoral level, where the relocation of electronics value chains to India or the transfer of electric vehicle (EV) production to Poland have been accompanied by a surge in China-sourced inputs (Figure 9).

Figure 8
Figure 9

This does not mean that diversification is not happening. However, it does mean that some of the objectives behind “de-risking” policies in advanced economies are unlikely to be met for many years. China’s centrality in global manufacturing supply chains, and unmatched capacity in critical sectors like electronics, EV or solar, means diversification will remain China-linked for years to come. Manufacturers will need to rebuild their local supplier networks over time—relying on relocated Chinese firms in many cases—and countries will need to rebuild local capacities and skills if they want to absorb manufacturing activity. This takes time, and significant investment.

Diversification, continued

Given these patterns, what should we expect in the future? We have developed five main hypotheses about the outlook for diversification away from China:

First, because diversification is driven by a wide range of factors which are likely to endure, from rising geopolitical competition to a deteriorating economic and policy environment in China, MNCs will continue to look for alternative manufacturing bases and supply sources for their most critical products and inputs. This does not mean that diversification cannot be accelerated or arrested by shifts in policy or by major geopolitical events, but we expect it to continue, and most likely deepen, over a period of years, if not decades.

Second, diversification will be characterized by greater investment by global MNCs in Asia, with Vietnam and India topping the list, and Thailand, Bangladesh, and Malaysia playing a role for specific sectors like EVs, textiles, or chips, in which these countries (respectively) already benefit from adequate skills and manufacturing capacity. Mexico will attract significant additional US MNC investment and sourcing interest across a broad range of sectors, as will Central and Eastern Europe (especially Poland and Hungary) from European MNCs. Chinese firms will deploy in ASEAN (rather than India), but also opportunistically to European, North African or North American destinations, as they seek to serve the EU and US markets. Even if slow and partial, these movements out of China will trigger significant economic, industrial, and logistical impacts on these countries, given their much smaller sizes.

Third, diversification will happen much faster in sensitive sectors (semiconductors, EVs, ICT but also healthcare industries and green technologies). In less sensitive sectors that require skills and advanced equipment, the persistent attractiveness of China as a manufacturing hub will remain a drag on diversification efforts. Across sectors, we expect stronger diversification impulses in cost-sensitive manufacturing—especially for final assembly activities—and in sectors where shipping costs per unit are lower and manufacturing capacity and skills exist outside of China.

Fourth, China’s role in global value chains, and therefore in “diversified” trade and investment, will endure. China’s share of global manufacturing value-added and exports could remain at a high level or even grow further in the years ahead (once COVID effects on China’s export levels wane in 2023). Yet we also expect a slow but gradual onshoring of value chain segments in alternative other countries, motivated by economic benefits from agglomeration. All else equal, agglomeration should happen faster in sectors like electric vehicles or solar panels, where co-location of supply chains is a key element of competitiveness, than in electronics, where value chains have been global for decades. All this could happen while Chinese companies continue to increase global market shares, as these firms pick up manufacturing and export activities vacated by MNCs in China, gain competitive footing in higher value-added sectors and manufacturing segments in China, and gain ground in alternative destinations.

Fifth, the slow pace of diversification could trigger calls in the US (and possibly other G7 countries) for a more aggressive approach to de-risking. Policymakers will need to adjust their expectations about what diversification can achieve in a short period of time. There is a risk, if they do not view this as a long-term challenge, that demands for bolder policy measures (in the form of both incentives and barriers to trade and investment) grow in the years ahead.



1. Our databasing of EU and US FDI into China only covers transactions over USD 1 million and would therefore not capture a rise in smaller transactions, should it have happened over the same period. We doubt however that this is the case given smaller firms are the ones that have most reduced their FDI into China in recent years.

2. On average, US imports of goods affected by tariffs are about 20% below pre-trade war levels, according to PIIE research. Note that the real drop might be slightly milder, however, as China-based exporters sought workarounds in response to US-China tariffs—breaking down orders into smaller parcels that are not covered by tariffs, or shipping goods through third countries. But there is no question that the US has been progressively diversifying its trade over the past five years, at the expense of China.

Vanishing Act: The Shrinking Footprint of Chinese Companies in the US

The United States is experiencing a post-pandemic boom in foreign direct investment (FDI), driven by the resilience of the US economy as well as new industrial policies that incentivize US manufacturing investment such as the CHIPS Act and the Inflation Reduction Act (IRA). Chinese companies are notably missing from the party. This note uses a proprietary Rhodium dataset as well as new US government data to analyze latest patterns of China’s FDI in the US. The findings are:

  • New investment has slowed to a trickle: Both official and alternative data show a sustained slowdown of Chinese FDI in the US since 2017. Annual investment has dropped from $46 billion in 2016 to less than $5 billion in 2022. In the past seven years, China has gone from one of the top five US investors to a second-tier player surpassed by countries such as Qatar, Spain, and Norway.  
  • The US footprint of Chinese firms is shrinking: Not only has investment slowed, but assets, revenues, and employment at Chinese companies in the US have all declined in recent years. The retrenchment is more severe and prolonged than the temporary slowdown in business that other multinational corporations (MNCs) experienced during the pandemic, suggesting that the retreat of Chinese companies from the US market was driven by restrictive economic policies fueling US-China economic decoupling over the past five years.
  • Declining local job creation is further undermining political support for US-China economic engagement: As the prospects of Chinese firms serving as major local employers in the US dwindle, fewer local officials and businesses are willing to stand up against more restrictive US economic and national security policies towards China.
  • The retreat of Chinese firms from the US illustrates risks of fragmentation and political misalignment: Chinese FDI has remained more resilient in other markets, reflecting a lower degree of regulatory and political scrutiny. The divergence of Chinese FDI trajectories across the OECD—especially in high-growth sectors like electric vehicles—illustrates the risks of fragmented global value chains as well as misalignment of policy priorities between the US and important allied nations.

Boom and Bust: The Slowdown of Chinese FDI Since 2017

Chinese FDI in the US experienced a period of rapid growth after 2014, but has been on a downward trajectory since 2017. Investment was a trickle up until the mid-2000s but increased steadily after the global financial crisis in 2008 and 2009 as Chinese firms weathered the storm relatively well and sought opportunities abroad. After Beijing loosened restrictions on outward FDI in 2014, large M&A transactions drove Chinese FDI in the US up to a record $46 billion in 2016. A re-tightening of capital controls and crackdown on highly leveraged private investors by Beijing led to a sharp drop in China’s global outward FDI. This was felt in the US in 2017 and following years, especially in real estate, entertainment, and other industries scrutinized by the Chinese government.

At the same time, US policy and politics started to impact Chinese investors, especially in sectors where Beijing still allowed and supported outbound FDI, such as the acquisition of technology assets in line with national development priorities. US policymakers started to shut Chinese companies out of certain markets for national security reasons, for example banning the use of Chinese telecommunications equipment. The US screening process for inbound acquisitions was tightened further with the enactment of FIRRMA in 2018, and the political environment for Chinese companies became more hostile overall amid the Trump administration’s confrontational stance toward economic engagement with China. By 2019, Chinese FDI had slowed to a trickle of less than $10 billion per year.

In recent years, headwinds have only increased. The outbreak of the COVID pandemic weighed on FDI globally in 2020, further dragging down China’s outbound investment. While global FDI flows rebounded in 2021, Chinese investors remained stuck at home due to Beijing’s zero-COVID policy. US national and economic security policies also continued to weigh on the prospects of Chinese companies. The Biden administration further expanded export controls and sanctions against Chinese companies and recently expanded investment screening to outbound capital flows. Most importantly, Congress enacted powerful new industrial policies that incentivize large capital expenditures into US manufacturing, but explicitly forbid or restrict the participation of Chinese investors, most notably the CHIPS Act and the Inflation Reduction Act (IRA). As a result, Chinese FDI in the US remained lackluster through 2022 and 2023.

Figure 1

The latest data from the Bureau of Economic Analysis (BEA) on new foreign direct investment in the US, which uses company surveys to calculate firms’ new FDI expenditures, confirms this trajectory. First year FDI expenditures of Chinese firms grew rapidly 2014-2016 but then peaked and slowed dramatically in following years. Since 2019, annual investment has dropped below the billion-dollar mark and 2022 shows the lowest level of new FDI in at least a decade (Figure 2). Over the past four years (2019-2022), new Chinese FDI in the US has averaged only $667 million, which is dwarfed by the expenditures of other major economies and significantly below the investments by MNCs from smaller Asian or European economies such as Singapore, South Korea, or Spain (Figure 3).

Figure 2
Figure 3

A Shrinking Footprint: US Operations of Chinese Firms

Not only is new investment slowing dramatically, but there is increasing evidence that the operational footprint of Chinese firms in the US has been shrinking in recent years. The BEA recently published their latest batch of data on the operations of MNCs in the US, which is collected through mandatory surveys and includes a variety of data that illustrate the struggles of Chinese firms in the US market.

The assets of Chinese companies in the US grew rapidly from $19.3 billion in 2009 to $275.5 billion in 2017. Since then, the asset base of Chinese companies in the US has plateaued, suggesting that additional investment flows since then have been offset by divestitures or write-offs. As of 2021, the BEA survey puts Chinese firms’ asset base in the US at $282 billion. That is almost four times higher than ten years ago, but roughly at the same level as in 2017. China only accounts for 1.5% of the total asset base of multinational companies in the US—a meager number for the US’s largest trading partner.

Figure 4

Another data point that illustrates the challenges of Chinese firms in the US market is declining revenues. Sales of Chinese MNCs in the US rose quickly from $3.4 billion in 2009 to just over $90 billion in 2017. Revenues then stagnated from 2017 to 2019 and contracted in 2020 and 2021. The 2020 drop can at least partially be attributed to pandemic-related market volatility, but the revenue of Chinese firms does not show the post-pandemic rebound in sales in 2021 seen for the broader set of MNCs operating in the US.

Figure 5

A particularly important trend is the decline of employment at Chinese companies in the US. Employment at Chinese firms in the US had surged from next to nothing in 2007 to more than 229,000 in 2017. Since then, employment at Chinese companies in the US has fallen back to only 140,000. That compares to about a million American workers employed by MNCs from other major export nations such as Germany (929,000) or Japan (1,045,000). Chinese firms today provide fewer jobs locally in the US than Swedish companies and only slightly more than firms from Italy or Australia (Figure 6).

Figure 6

Implications for US China Policy: Less Ballast for Engagement

While the physical presence of Chinese companies in the US economy as investors and employers has shrunk since 2017, Chinese exports to the US have continued to grow to a record $564 billion in 2022. China remains a massive outlier in that it predominantly serves the US market through exports and has not followed other major trading partners in building out a significant local presence through FDI in local manufacturing and service operations (Figure 7).

The shrinking prospects of significant local job creation by Chinese firms have arguably further tilted the scale toward more restrictive economic policies toward China. Local policymakers and US companies have fewer incentives to spend their political capital on pushing back against policies in Washington that are increasingly shaped by national and economic security considerations.

Figure 7

Global Implications: The Fragmentation of China’s Outbound FDI

From a global perspective, the withdrawal of Chinese firms from the US economy illustrates the risk of an increasingly fragmented global economy, particularly in high-tech sectors with national and economic security relevance such as clean tech, telecommunications equipment, or semiconductors.

While the slowdown in China’s outbound FDI after 2017 is primarily a result of Chinese policies, the decline of Chinese FDI in the US has been deeper and more sustained than in other advanced economies, in part due to more restrictive US policies. In addition to investment screening by CFIUS, US policymakers have expanded their toolbox to explicitly or implicitly restrict market access and local R&D activities of Chinese firms through supply chain security policies, sanctions lists, export controls, and industrial policy provisions. Other countries have tightened their investment review processes but have not matched the US in other areas, contributing to a growing divergence in Chinese outbound FDI patterns.

One illustrative example is the electric vehicle industry. While overall Chinese outbound investment has declined, Chinese companies have invested billions of dollars in the past five years across the global electric vehicle value chain, from critical minerals mining to battery manufacturing and research and development operations. The US has experienced a boom of foreign investment in the electric vehicle industry driven by the IRA, but the role of Chinese investors has been limited due to provisions that exclude Chinese investors from certain government subsidies.

Stark differences in the value of announced Chinese FDI in the EV industry in Europe and North America illustrate these diverging trajectories. Since 2019, Chinese companies have announced investments exceeding $30 billion across Europe, including large manufacturing facilities like Envision AESC’s in France and the UK, as well as CATL’s in Hungary. Over the same period, Chinese FDI in North America totaled less than $7 billion (Figure 8). Among the small number of investments announced in the US, several large-scale Chinese battery plants—like the $2 billion facility by Gotion in Michigan—have triggered a political and regulatory backlash.

Figure 8
Figure 9

If this divergence persists, it could become a source of transatlantic tension, particularly if Europe’s embrace of Chinese green technology suppliers gives its home-grown manufacturers a leg up on US rivals that do not have access to the same critical inputs. Ultimately, this is also about jobs. Over the past five years we have observed a stark divergence between Europe and the United States when it comes to local job creation by Chinese firms: In 2017, Chinese companies employed roughly the same number of workers in Europe and the US (between 200,000 to 250,000). Today, more than 300,000 Europeans are employed by Chinese companies, while the number of US workers with Chinese employers has dropped to less than 140,000 (Figure 9).

At a time when the US and Europe (not to mention other G7 countries like Japan, the UK, and Canada) are ramping up efforts to coordinate their policies toward China, differences in their embrace of Chinese FDI could, in the long run, undermine this push. It will be especially important to monitor whether the US and its allies converge around a common approach to Chinese greenfield investment in high-growth, job-creating industries.