In October 2025, on the sidelines of the Asia-Pacific Economic Cooperation (APEC) summit in Busan, South Korea, Presidents Donald Trump and Xi Jinping struck a trade truce that included a striking agricultural commitment: Beijing would purchase at least 12 million metric tons (mmt) of American soybeans before the end of that year, and at least 25 mmt annually through 2028. China’s Ministry of Commerce (MOFCOM confirmed) the outcome from its side days later, with spokesperson He Yadong stating that the two countries’ economic and trade teams, meeting in Kuala Lumpur, had reached consensus including an agreement to expand agricultural trade. Seven months later, the White House put China’s total annual agricultural purchases at no less than US$17 billion through 2028. On the surface, this looked like cooperation reasserting itself in a relationship widely described as entering structural rivalry. Underneath it, in the same months, China’s soybean sourcing was moving decisively the other way. This apparent contradiction is not new, and it is not really a contradiction at all: it reflects a long-term Chinese food security strategy that predates the current trade dispute by years and understanding it changes how the entire U.S.-China trade relationship should be read.
This essay makes a narrow proposal: a standing technical working group, housed inside an existing multilateral institution, to align how Washington and Beijing count and report soybean purchases. But the stakes of getting this right extend well beyond soybeans. The world has entered a geo-economic era in which diversifying supply chains has become standard practice for any state managing exposure to a single trading partner — and in which that same diversification is routinely mistaken for a signal that a state intends to renege on its trade commitments. Soybeans are simply where that confusion is currently most visible.
A crop China cannot grow its way out of
Soybeans occupy a key place in China’s food security architecture. Despite being one of the world’s largest soybean producers, it is also one of the biggest importers. Domestic output reached 20.65 million tons in 2024, Han Jun confirmed, up from roughly 16 million tons earlier in the decade. Against annual demand of well over 100 million tons, that leaves a self-sufficiency rate estimated at only 14 percent to 16 percent. The gap reflects a divide that is regulatory as much as agronomic. China permits the import of genetically modified soybeans for processing but has never approved a GM soybean variety for commercial cultivation domestically, so homegrown beans remain non-GM by law. The result is two effectively separate markets: imported, genetically modified soybeans are crushed almost entirely into soybean meal for animal feed — which sustains China’s vast pork, poultry, and aquaculture industries — and into edible cooking oil, while domestically grown, non-modified soybeans go overwhelmingly toward direct human consumption (such as tofu and soy milk). Roughly 13 percent of China’s total soybean demand is met through that direct food channel; the remaining 84 percent is processed into meal and oil, feeding a livestock sector that has expanded rapidly alongside rising incomes. Huang Jikun, Dean of the College of Modern Agriculture at Peking University and founding director of its China Center for Agricultural Policy, has put the stakes plainly: soybean imports, at roughly 100 million tons a year and overwhelmingly destined for feed, make soybeans China’s single largest agricultural import category, so that for China, food security is substantially soybean security. This is the backdrop against which every purchase pledge, and every diversification move, must be read: Beijing is not choosing between growing its own soybeans and importing them. It has already concluded it cannot do the former at scale and has built its food security strategy around managing the latter as carefully as it can.
Pursuing an import diversification policy
That management did not begin with the 2025 trade truce, or even the 2018 tariff dispute. It reflects a sustained, state-directed effort to reduce dependence on any single external supplier of a commodity Beijing considers strategically sensitive. Following the 2019 Central No. 1 Document — the Chinese Communist Party’s annual flagship policy statement on agriculture and rural affairs — and its call to implement a soybean revitalization plan, the Ministry of Agriculture and Rural Affairs (MARA) issued the scheme on March 15, 2019, setting 13th Five-Year Plan targets for expanding domestic acreage, raising yields, and improving product quality by 2020 and 2022. The plan’s own language describes the goal as gradually forming a pattern in which domestic planting structure adjusts dynamically alongside international market changes, an ordinary feature of long-term planning rather than a response to any single trade partner.
On the import side, Brazil has been the primary beneficiary: in 2024, China imported a record 105 mmt of soybeans, of which roughly 71 percent came from Brazil and about 21 percent from the United States, a marked change from 2016, when the American share stood closer to 40 percent. The pattern extends further. The average U.S. share of its own soybean exports to China fell from roughly 60 percent before the 2018 trade dispute to roughly 47 percent after it, while Brazil’s already-substantial share held steady near three-quarters. At the same time, China has sought stronger trade relationships with both Argentina and Uruguay, importing a combined 5 million tons from the two countries between September 2024 and July 2025 alone. By the first five months of 2026, more than 60 percent of China’s soybean imports still came from Brazil, versus roughly 23 percent from the U.S. and 10 percent from Argentina. Purchases under the new October 2025 commitments arrived more slowly than the headline targets implied. For instance, at the time of writing, by mid-2026 China remained behind pace even as buying continued. This pattern — diversifying steadily regardless of how any single bilateral deal happens to be faring — is best read as evidence of a durable food security strategy that exists independently of, and will outlast, any particular agreement with Washington. That kind of behavior fits a broader pattern political scientists have long studied.
The theory behind the behavior
Political scientists have a name for the vulnerability this diversification responds to. Henry Farrell and Abraham Newman’s concept of “weaponized interdependence” describes how globally networked economies contain structural chokepoints that states controlling those nodes can exploit for coercive advantage. In this light, diversification is the obvious response: states with concentrated dependence build redundancy regardless of whether the relationship with the concentrated supplier happens to be cooperative or adversarial at a given moment. Daniel Drezner’s critique of the framework is instructive here — several of its most-cited cases underperformed as coercive tools, including China’s own 2010 rare-earth restrictions on Japan, which weakened Beijing’s leverage by accelerating Japanese diversification away from Chinese supply. Diversification, in other words, is frequently just the rational response of any state to concentrated dependence, and its presence does not by itself signal that a relationship is deteriorating. Reading China’s soybean diversification as evidence its purchase commitments to the U.S. are hollow, on this logic, misreads both the diversification and the commitments.
Why would China accept a multilateral fix?
Any proposal for closer multilateral engagement around this dispute must reckon with why China would accept the exposure that comes with it. Farrell and Newman’s framework identifies a second mechanism worth considering: the panopticon effect, whereby hubs gain power not only by cutting off flows but by surveilling them. Given this context, it is reasonable to ask why Beijing would agree to any multilateral mechanism that gives a U.S.-anchored system more visibility into its trade behavior. It is also worth noting that the current deal has already produced exactly the kind of dispute such a mechanism would need to resolve. For instance, U.S. officials have publicly disagreed over what “meeting the commitment” means, in terms of purchase timelines and marketing-year benchmarks.
To this end, China’s track record with comparable institutions is reassuring on the question of willingness, at least. It holds the largest single earmarked commitment of any member state — 300,000 tons — in the ASEAN Plus Three Emergency Rice Reserve (APTERR), a legally binding mechanism it signed alongside Japan and South Korea in October 2011 and has funded real disaster-relief releases through it. For instance, China donated 800 tons to typhoon-affected areas of the Philippines in 2014. This indicates that Beijing is prepared to accept the material cost to belong to multilateral food-security architecture in addition to funding it. This pattern is further seen in China’s role as APEC’s 2026 host, where food security sits among the forum’s functional cooperation priorities and where Beijing is due to convene a Food Security Ministerial Meeting in Hangzhou later this month.
A narrow fix
The Agricultural Market Information System, known as AMIS, is the institution that can absorb this dispute, and it does not need to be built from scratch. Established by the Group of Twenty (G20) in 2011 in direct response to the 2007–2008 and 2010–2011 food price crises, AMIS monitors global markets for wheat, maize, rice, and soybeans and works to improve policy coordination during volatility. It has already shown its transparency function can hold under conditions far harder than present U.S.-China friction: on March 5, 2022, days after Russia’s invasion of Ukraine — a war between two of AMIS’s own participating countries — its Rapid Response Forum convened an extraordinary session, bringing senior officials from Argentina, Australia, Brazil, Canada, the European Union (EU), Japan, South Korea, Russia, Ukraine, the U.S., and others to jointly review wheat, maize, and sunflower oil supply data in real time. The Food and Agriculture Organization (FAO) credits this series of dialogues with helping reduce the conflict’s impact on global food trade, alongside AMIS’s earlier COVID-19-era assessments, which found supplies adequate and helped calm markets. AMIS kept functioning, and demonstrably helped stabilize markets, while two of its own members were in active armed conflict. The same 2011 process also named a companion Agriculture and Food Security Risk Management Toolbox, meant to give AMIS a risk-management counterpart, but the Toolbox was never resourced, and its intended contents are difficult to document in detail today. In this light, it provides further reasons to build on what AMIS has already proven rather than on what its unbuilt companion was only ever meant to become.
The proposal is a standing U.S.-China technical working group housed inside AMIS, with a mandate limited to three things: harmonizing what counts as a “marketing year” for reporting purposes, agreeing when a purchase is recorded as fulfilled, and establishing a common reporting template both countries’ customs data can be checked against. This is smaller than it might sound, and deliberately so. Both countries’ customs authorities already publish soybean trade data publicly and monthly. In this context, the working group would not hand either side new visibility into the other’s trade. Instead, it would align the definitions each side already uses to describe numbers both already disclose. Set against APTERR’s earmarked physical stocks, that is a modest ask, which is exactly why it is achievable now, without waiting for the years of accumulated trust such a mechanism required.
The obvious objection is that this undersells the scale of the problem. A technical reporting fix does nothing to stop the next round of diversification-driven friction. That is fair, and the honest answer is that this proposal is not meant to solve that larger problem. Instead, it is meant to solve the one concrete, current dispute where a solution is available, without waiting on either Beijing or Washington’s broader posture toward multilateral institutions.
The same logic, reversed
Washington is proving this same point right now, on a different commodity, inside the very agreement this essay has examined throughout. China refines roughly 90 percent of the world’s rare earths, and the same Busan agreement that produced the soybean commitments also included pledges to keep rare-earth exports flowing to the U.S. Even as it made that pledge, Washington spent the following months building the rare-earth equivalent of Brazil: bilateral critical-minerals frameworks with Japan and Australia in October 2025, a memorandum with Malaysia the same month, and in February 2026 a Critical Minerals Ministerial in Washington gathering more than fifty countries, explicitly aimed, in Rubio’s words, at a supply chain “heavily concentrated in the hands of one country.” Between January 2025 and June 2026, the U.S. government committed roughly US$10 billion to the critical-minerals sector under this push. Few would call this bad faith. Washington is honoring its bilateral supply arrangement with China even as it builds redundancy away from it — the same combination this essay has argued is not a contradiction at all.
Both countries are managing concentrated exposure to a single trading partner while simultaneously deepening a bilateral commitment with that same partner. This is not a contradiction. Rather, it is the ordinary shape of trade policy under conditions of chokepoint vulnerability. Reading diversification as bad faith, whichever government is doing the diversifying, misses that — and the cost of that misreading will only grow as more governments diversify more supply chains at once.
What prevents that misreading from hardening into policy is not restraint or good faith on either side, but institutions capable of showing, in real time, that diversification and commitment can run on parallel tracks. AMIS already does this for food markets; nothing comparable yet exists for the commodities likely to matter next. Fixing how two governments count soybean shipments will not, on its own, build that capacity. But it is a place to start that neither side has an excuse to refuse.
Genevieve Donnellon-May is a Researcher at the Centre for Contemporary Chinese Studies, University of Melbourne’s Asia Institute, and a Non-Resident Fellow at the Asia Society Policy Institute’s Centre for China Analysis.