U.S. China: What are Chances for a Deal Around Rare Earths and Semiconductors?

Short-term progress yes, but the status of long-term US access to critical minerals remains unclear

Recently one version of this piece was shared with a group of key stakeholders of the Center for China Analysis (CCA). Here is the full text with an update based on initial feedback from that process. By Paul Triolo, Honarary Senior Fellow for the Asia Society Policy Institute, Center for China Analysis.

In early April 2025 China’s Ministry of Commerc (MOFCOM) tightened export licensing on seven medium‑to‑heavy rare earths—samarium, gadolinium, terbium, dysprosium, lutetium, scandium and yttrium—and it could do so because Chinese companies still dominate the entire chain: they mine about 60 % of all rare‑earth ore, refine roughly 87 % of it, and handle an estimated 99 % of global heavy‑REE separation; for the controlled elements specifically, China supplies 100 % of samarium and virtually the only commercial output of dysprosium and terbium. Downstream, Chinese firms churn out more than 90 % of the world’s rare‑earth permanent magnets, including Sm‑Co magnets that keep guidance fins and F‑35 actuators working at 300 °C, and Dy/Tb‑doped NdFeB magnets that let EV traction motors, wind‑turbine generators, drones and robotics run without bulky cooling systems. The same seven elements feed Y/Sc‑strengthened aluminium fuselage alloys, Gd‑based MRI contrast agents, YAG lasers, LED phosphors and Lu‑rich LYSO scintillators for PET scanners—so Beijing’s grip on their supply cascades into everything from green power to precision‑guided munitions and medical imaging.

On April 4, 2025, MOFCOM put in place a global ban on shipments of these 7 REEs and associated products. Since that time, the issue of REEs and particularly magnets, has dominated US China relations, requiring face to face meetings in Geneva and London between US and Chinese officials. The issue is complicated by Chinese bureaucratic challenges arising from the imposition of a new export control regime, with MOFCOM having to scramble to add personnel to review license applications, and from lack of clear coordination between MOFCOM and Chinese customs officials. As of mid-July there has been progress as a result of a still unreleased “framework agreement” coming out of the London meeting. Chinese officials came in with what have been described as “aggressive demands” related to US export controls, and while licenses for commercial end users are now being issued more regulalry, it remains unclear if suppliers to US military and defense industrial end users will be able to ever get a license approved. Some in the industry think resumption of licensing for US military end users is not likely, but because it remains unclear what if any concessions the US side may have been in the negotiations in London and subsequent discussions, the situation remains murky.

We will know within a period of months whether major US defense contractors will be able to continue to source REEs and magnets from Chinese suppliers. In a likely sign of how now focused and concerned Pentagon official are about the issue, this week a new investment from the Department of Defense in rear earths major MP Materials, which operates the only rare earths mine in the US, was announced. Under the deal the Pentagon will become MP’s largest shareholder. Chinese rare earths giant Shenghe is also a shareholder in MP. The essay below tackles how a broader deal around export controls now in place on both sides could be shaped. US officials are clearly reluctant to talk about negotiations involving export controls, which before Geneva and London, had not been part of any similar negotiations. A written version of the framework agreement exists, according to the the White House spokesperson, but it appears both sides have agreed not to make the document public. The rare earth issue will continue to be a major topic in US China relations going forward.

Questions for initial feedback on this issue from select readers:

Why did the US not do more on REEs and magenets after what happened to Japan in 2010?

Complicated question, the short answer is that the problem of seriously reducing dependence on one country would require a sustained multi-year effort, at least 10-15 years, involving both public-private parnterships, government incentives/subsidies, and the parallel development of a supportive ecosystem, including things like mining and processing facility permitting, rare earths processing engineering at the university level to produce a pipeline of qualified engineers across the “mine to magnet” ecosystem, and more. China has been at this for at least two decades, and there are more than 120,000 individuals dedicated to the rare earth material supply chain. Nearly 40 university programs crank out around 2000 graduate in critical fields related to the industry every year, sustaining and expanding the pool of REE industry personnel. Without the industrial policy base and industry that would be requried to develop this type of education pipeline, and on again off again support for key companies in the industry in the US such as MP Materials, there has been no serious effort in the US to develop the type of holistic ecosystem that is required to have a serious REE material and magnet industry. With Chinese companies so dominant in the industry, and an able to control prices, if has been very difficult since 2010 to get US companies to enter the industry, given lack of long-term government support and lack of a personnel pipeline.

How to evaluate how vulnerable US is now? and related How to have a realizable timeline?

Another very complex question, which differs by rare earth, light or heavy, which have different processing chains and significantly different end user requirements and profiles. In many cases, more than one rare earth is combined as part of a final product, for example. US commercial and military/defense industry vulnerability to Chinese company dominance is very high, with few alternative suppliers. China’s new export control regime resulted in the partial shutdown of production at a major US automaker’s facility with around two months! That is a vulnerability. For military end users, the vulnerabilities are equally as severe, but there has been much less public discussion of these. Beijing is very serious about restricting exports to military end users and is requiring substantial documentation from end users about how REEs and product such as magnets will be used. While US defense industrial firms scramble for alternatives, there are few alternative suppliers—Japanese companies have significant know-how in area such as magnet manufacturing, have some stockpiles, and are likely to be one short-term source of alternative supplies for some specific applications. A long-term solution to the supply chain vulnerabilities will require a serious industrial policy effort, coupled with a plan to manage relations with China and maintain access to REEs and critical products such as REE magnets for a sustained period of time while efforts are made to build alternative supply chains. Right now, there is no coherent plan for either of these two requirements.

The article:

U.S. China Export Control Impasse: What are Chances for a Deal Around Rare Earths and Semiconductors?

INTRODUCTION

The United States and China continue to grapple with how to address the new situation created by Beijing’s establishment of a major export control regime around rare earth elements (REEs) and products such as magnets. It is now clear that both sides possess significant leverage over technology supply chains and are willing to use it. Plant closures in both countries—caused by tariffs, broader trade issues, and now rare earths—are detrimental to leaders in both nations and to the prospects for economic growth. While both sides stepped back from the brink in Geneva and London, pressures remain that could return the bilateral relationship to an even more dangerous downward spiral. Is there a way out of this? “

ESCALATE TO DE-ESCALATE” WILL NOT BE A WINNING FORMULA

Coming out of the May trade meeting in Geneva and the subsequent London meeting two weeks later—which focused on REEs and magnets on the U.S. side and U.S. export controls on semiconductors and their manufacturing tools on the Chinese side—there is still no clarity about the full scope of the “framework” announced in London, which was intended to revive the “consensus” established in Geneva. Industry sources suggest that there was no real “deal” at either meeting, and no details of any “framework” allegedly agreed upon in London have been released. While U.S. officials such as President Donald Trump and Secretary of Commerce Howard Lutnick have made broad statements that China agreed to free up export licenses for REEs and magnets, this clearly has not yet happened on a broad scale. Chinese officials have so far said nothing about the details of a notional “framework” deal.

As occurred between Geneva and London, both sides appear to be considering further escalation to force the other into making concessions. U.S. officials seem to cling to the idea that Beijing has agreed to roll back REE controls to where they were before April 4. This is not likely, as Beijing has for some time viewed strong control over REEs and related products as a national security priority. Radical measures are under consideration, such as suspending all licenses for U.S. technology exports to China in sectors such as semiconductor manufacturing tools. The first package included restrictions on aircraft parts, chemicals (including ethane), semiconductor design tools, and high-purity quartz. So far, none of these restrictions appear to have been lifted. Their relaxation was linked to progress by the Chinese Ministry of Commerce in releasing more licenses for REEs and magnets. U.S. Department of Commerce (DOC) officials continue to collect licensing information from U.S. companies, and the picture remains decidedly mixed. For example, what metric is the DOC looking for before easing the post-Geneva controls? [This situation has improved since the initial circulation of this article]

For Beijing, the REE and magnet issue is now a national security priority, and the licensing regime being constructed around the industry will not be rolled back. The global stoppage of REE and magnet shipments on April 4 has disrupted multiple industries dependent on these products, and the market may take years to recover—even if Beijing loosens restrictions slightly in the coming months. A new interagency group, the National Export Control Coordination Mechanism, is now reviewing license applications with the aim of limiting exports of critical materials to any U.S. military or defense industrial user. This policy will remain in place regardless of any commercial agreements that may be reached regarding end users in the automotive semiconductor and consumer electronics sectors. Beijing is also aiming to control exports to third countries that might redirect products to sensitive end users. While the control regime is still in its early stages, the involvement of the Ministry of State Security and the Ministry of Public Security in the interagency review process highlights how seriously Beijing takes the issue. Hardliners in Beijing are pushing to use this newfound leverage to gain concessions from Washington on export controls—one of the key issues complicating implementation of the handshake agreement in London on REEs and magnets.

As hardliners on both sides dig in, is there any chance for a broader deal? For China, U.S. export controls on semiconductors have long since become a red line—viewed by President Xi Jinping as an effort to constrain China’s development and likely touched on during a June 5 phone call with President Trump that set the stage for the London meeting. Beijing’s concerns center not only on post-Geneva measures by the DOC but also on the entire framework of controls initiated since October 2022, targeting advanced AI hardware and semiconductor manufacturing tools. Beijing is also worried about a new wave of controls in the pipeline, including some currently under consideration as part of the “escalate to de escalate” package in Washington.

Is there a chance for an broader deal to be made about licenses and controls? The answer is that it is complicated. U.S. export control proponents—led by Under Secretary of Commerce for Industry and Security Jeff Kessler, National Security Council staffer and former Select Committee on China official Landon Heid, and backed by elements of the U.S. intelligence community and Washington think tank advocates of maintaining and expanding the existing controls to contain the ability of Chinese companies to develop advanced AI models —appear ready to push forward with more controls. Secretary of the Treasury Scott Bessent, U.S. Trade Representative Jamieson Greer, and other administration moderates seeking to align with President Trump’s preference for some type of deal with Beijing appear increasingly concerned about the impact of REE and magnet controls on U.S. companies. They believe that Beijing could significantly raise the stakes if the administration proceeds with further escalation.

WHAT COULD A DEAL LOOK LIKE?

The window for de-escalation around the REE and magnet issue is rapidly closing, especially as elements of the escalation measures are leaked to the media. Kessler has indicated he is ready to suspend another key export control policy—previously granted exemptions to South Korean and Taiwanese semiconductor manufacturers in China under the Validated End User program—replacing it with a licensing requirement that includes a presumption of denial for any products that could enhance Chinese innovation capacity. Washington’s decoupling faction supports this measure as a logical continuation of the October 2022 rationale for slowing China’s AI advancement. However, this approach is now generating mounting awareness in Washington of the significant collateral damage to U.S. technology leaders. A renewed push would further harm U.S. technological leadership without yielding national security benefits. These measures would go far beyond the “small yard, high fence” approach of early Biden administration policy and instead compound what I have called the “small gain, high cost” nature of more recent U.S. export control actions. The rare earth dimension adds even more urgency and risk to continued escalation.

There may be a different and novel approach to defuse the downward spiral in bilateral relations. A deal focused on licensing—covering both REEs and magnets, as well as semiconductor manufacturing equipment—could benefit U.S. technology leaders, global supply chains, and U.S.-China relations. This approach might also appeal to President Trump, who still hopes for a broader deal with Beijing. While many in the administration favor deeper decoupling, the president does not, and he, along with Bessent and Greer, may be looking for a resolution to the REE and magnet standoff that minimizes damage to U.S. industry while maintaining access to Chinese supply for U.S. defense and military users. But this will require the Trump team to roll back some of the export controls that Beijing finds especially provocative. The Trump administration has already rescinded the Biden-era AI Diffusion Rule, which had been hastily implemented without industry input. AI and crypto czar David Sacks has vocally criticized both the rule and the Biden national security team behind it, including on X and in a recent podcast.

One potential framework for a deal would be for the administration to agree to roll back selective export controls put in place since the November election—a series of controls rushed out unilaterally without industry input that remain on the books. These controls have not demonstrably impacted the ability of Chinese companies to develop advanced AI models but have had a huge negative impact on U.S. industry, in addition to contributing to Beijing’s drive to control REEs and magnets. In addition, DOC officials are sitting on tens of billions of dollars in licenses for U.S. technology companies, including some of the leading U.S. firms in the semiconductor manufacturing space. These U.S. controls are also driving U.S. companies offshore, while Chinese and other foreign technology firms continue to design out U.S. technology to avoid future restrictions—a bad long-term dynamic for U.S. technology leadership.

A deal focused on licensing—covering both REEs and magnets, as well as semiconductor manufacturing equipment—could benefit U.S. technology leaders, global supply chains, and U.S.- China relations. The United States could agree to roll back controls that have not produced any measurable impact on China’s AI capabilities but have galvanized China’s domestic semiconductor industry to make major improvements in innovation and capacity—without giving up all controls on the most advanced AI semiconductors from October 2022 or on manufacturing tools.

In exchange for a long-term commitment from Beijing to ease license approvals, potentially develop a “white list of U.S. companies,” and agree on long-term licensing arrangements to restore a hugely disrupted global supply chain and market around REEs and magnets, the Trump administration could selectively rescind provisions of the most recent set of export controls—primarily Biden-era rules—and frame the pullback as consistent with the Trump team’s preference for reducing regulation, simplifying rules, and responding to industry concerns. It remains unclear what types of guarantees Beijing would be willing to accept for such an agreement, which would require considerable negotiation. It could include rolling back changes to end-use controls on DRAM pushed out in January 2025, rescinding new foreign direct product rules released as part of a major package in December 2024, and removing some or all Chinese firms from the Entity List—firms that were added in December 2024 and span broad swathes of the semiconductor supply chain. This represents a glaring example of the abandonment of the “small yard, high fence” policy outlined by former National Security Advisor Jack Sullivan in 2022. In addition, the Trump administration could rescind the “is informed” letters sent to Nvidia, AMD, and Intel in April, which warned that a new rule would preclude shipments of GPUs such as the Nvidia H20 to China. These GPUs are far from the most advanced and were redesigned by Nvidia specifically to comply with DOC controls—the H20s are primarily useful for AI inference, not for training the most advanced models.

Finally, the administration could begin granting licenses for some of the billions in shipments to Chinese companies that are already in the Bureau of Industry and Security pipeline and have been for months, prioritizing less sensitive tools, commodity products, and non military-related end users in ways that mirror what the U.S. side is asking Beijing to do with rare earth materials and magnets. Easing licensing for suppliers to companies such as Huawei and Semiconductor Manufacturing International Corporation, which have been tagged as “military-related” or contributing to China’s civilian military fusion program, could be negotiated with Beijing in return for lifting restrictions on rare earths and magnets for U.S. dual-use defense industrial and military end users. Without such concessions, Beijing will not ease up on what it views as national security restrictions on REEs that mirror U.S. controls on semiconductors and semiconductor tools going to what Washington considers military related end users.

In summary, the United States could agree to roll back controls that have not produced any measurable impact on China’s AI capabilities but have galvanized China’s domestic semiconductor industry to make major improvements in innovation and capacity—without giving up all controls on the most advanced AI semiconductors from October 2022 or on manufacturing tools, including advanced lithography equipment first implemented in 2019 before the major expansion in October 2022. In return, U.S. companies would regain long term access to critical minerals and products such as magnets. This type of arrangement would also address how to bridge the period required for the United States and its allies to begin reducing their heavy dependence on Chinese firms for critical minerals, REEs, and products such as magnets—this is a “managed coupled derisking” approach. It is at least a five-year project, if not longer, and will require up to $1 trillion in investment over a decade, according to industry estimates, and off take committments of the type apparently in the new Pentagon investment in MP. Without such an agreement with Beijing in place, U.S. companies will continue to face the constant specter of both cutoffs of REEs and magnets and the slow rollout of licenses, along with an increasing probability of further retaliation from Beijing against U.S. company operations in China—particularly if the DOC and the decouplers succeed in pushing forward more export controls.

At the same time, further controls will continue to degrade the position of leading U.S. firms in China. Already, Nvidia has had to write off at least $4.5 billion—and potentially as much as $20 billion overall—related to the H20, and U.S. toolmakers have license applications pending at the DOC totaling multiple billions of dollars. In addition, as part of any deal, U.S. officials could consider removing “domestic persons” controls for U.S. toolmakers to level the playing field with Japanese and Dutch competitors, who were not required to pull personnel from facilities in China. This would help prevent further hemorrhaging of U.S. know-how caused by the forced removal of U.S. persons from Chinese semiconductor facilities, as occurred in October 2022—allowing competitors access to sensitive equipment. One industry official has called this policy the largest and most consequential unforced technology transfer to China in history. It is time for a deal that benefits all parties, stabilizes complex supply chains and markets, restores U.S.-China relations to a more stable footing, and paves the way for broader agreements on tariffs, fentanyl, and Chinese investment in green technology facilities in the United States—investments that benefit U.S. consumers and create jobs. If the REE and semiconductor controls have revealed anything, it is that the cost of decoupling would be extremely high for both economies—and that there are creative ways to address both national and economic security concerns without incurring even greater costs than those already racked up by nearly three years of damaging export controls.