Japan's heavy rare earth stockpile is draining faster than its diversification plan
Fifteen years spent building a second source of the hardest magnet metals, and the one country that controls the first just closed the valve to see which runs out.
Two clocks are running in Tokyo and only one of them is being wound. China exported zero dysprosium and zero terbium oxide to Japan from November 2025 through May 2026, cutting the two metals every high-performance magnet needs by roughly three-quarters in 2025 before shutting the flow off entirely (Rare Earth Exchanges, Aug 14). Against that, Japan's own plan of record is a seabed mining test in February 2027 and an economic viability assessment due March 2028, with a Malaysian magnet plant that does not start production until late 2027 (JapanTalkback, Aug 2; Observer Research Foundation, Jul 27). The stockpile sits between the two clocks, and it is the only thing being spent.
The trigger came on January 6, when China's Ministry of Commerce tightened export controls on dual-use items to Japan, drawing the line by end user rather than by product list, so the same shipment can clear for one buyer and stall for another (JapanTalkback, Aug 2). The retaliation was aimed at Prime Minister Sanae Takaichi's November 7 statement that a Chinese invasion of Taiwan could constitute a survival-threatening situation for Japan. But the trigger is the small story. The slow pressure is fifteen years of diversification that fixed the wrong half of the problem: China's share of Japan's rare earth imports fell from about 90 percent toward roughly 60 percent overall, yet Japan still depends on China for nearly all of its dysprosium and terbium (JapanTalkback, Aug 2). Japan diversified the cheap part of the periodic table and left the expensive part exactly where it was.
The actors are few and their incentives are legible. Takaichi wants to hold the Taiwan line without presiding over stalled auto lines, which is why she met Keidanren chairman Yoshinobu Tsutsui on January 19 to push private procurement diversification while promising coordinated diplomacy (JapanTalkback, Aug 2). Beijing wants the statement withdrawn and the demonstration effect preserved for every other capital watching: the controls cost China revenue, so they only make sense as a lesson, not a business. The trading house Sojitz and JOGMEC, the state metals agency, hold the Lynas stake that entitles Japan to up to 65 percent of output of the dense-magnet oxides from Lynas's Malaysian plant, and in 2025 Japan invested in France's Caremag for a hoped 20 percent of its demand for them (JapanTalkback, Aug 2). Each of these is a real valve. None of them is open yet.
The numbers behind the stockpile explain the urgency. Japan's 2020 resource strategy set a stockpile target of 60 days of supply for standard-risk minerals and up to 180 days for high-risk ones, with rare earths in the higher tier (Rare Earth Mining, May 19). Nomura Research Institute estimated a three-month Chinese restriction would cost the Japanese economy about 660 billion yen, roughly 4.2 billion dollars, and shave 0.11 percent off GDP, with a full year of restrictions reaching 2.6 trillion yen and 0.43 percent of GDP (JapanTalkback, Aug 2). Argus now reports the dysprosium and terbium shortage stretching to at least 2027 because demand is outrunning every substitute (Caixin Global, Aug 21). A six-month embargo plus a months-of-cover stockpile plus a 2027 supply date is arithmetic, not politics: the buffer empties before the alternative arrives.
The history Japan keeps reaching for is its own. After the 2010 Senkaku trawler incident, China embargoed rare earth shipments to Japan for about two months, and Tokyo responded with a playbook: sue at the World Trade Organization, fund Lynas, stockpile, and redesign motors to use less material (East Asia Forum, Mar 19). That playbook worked, which is precisely the danger. It worked on light rare earths, where Australia had ore and separation was solvable with capital. Dysprosium and terbium are different: they come out of ion-adsorption clays mined almost entirely in southern China and Myanmar, and China controls an estimated 85 to 91 percent of global rare earth refining, the step that turns rock into metal (Observer Research Foundation, Jul 27). Japan spent fifteen years proving it could route around a blockade of the easy elements while the hard ones stayed put.
Japan diversified the cheap half of the periodic table and left the half that matters exactly where Beijing put it.
The counterargument deserves its day in court, because in 2010 the embargo itself collapsed. China lifted it within months under WTO pressure and lost the case, and it lost customers permanently, a trade Beijing has reason to remember. This time is different in one measurable way: the January controls are written by end user, not by product, so Beijing can calibrate the squeeze shipment by shipment and deny doing anything at all (JapanTalkback, Aug 2). A weapon that can be dialed is a weapon that can be held. The 2010 precedent argues China blinks; the 2026 drafting argues it does not have to.
Walk the chain forward and the first squeeze lands on magnet makers, not miners. The mines for these elements sit inside China's control, so the profit migrates to whoever can separate and alloy outside the blacklist: Lynas and South Korea's JS Link just contracted a 3,000-tonne magnet plant in Kuantan, Malaysia, with Lynas supplying rare earths through January 2038, while France's Carester and Malaco Mining work on a separation plant in Perak (Observer Research Foundation, Jul 27). In the other direction, Energy Fuels of the United States is buying Germany's Vacuumschmelze for roughly 1.9 billion dollars, stitching mines to magnets across allied soil (Observer Research Foundation, Jul 27). Who pays in the near term is the Japanese assembly line: hybrid and electric vehicle motors, MRI machines, wind turbine generators, all waiting on oxides that only one country currently refines.
The G7 has noticed, and its answer is a number with a deadline. At Évian in June, leaders committed that no member would source more than 60 percent of its rare earth and magnet imports from any single supplier by 2030, approved 195 projects worth 64 billion euros, and set up pilot stockpile coordination with the International Energy Agency (Observer Research Foundation, Jul 27). The Center for Strategic and International Studies has already flagged the gap between signatures and tonnes: most of those projects will not reach commercial production within four years (Observer Research Foundation, Jul 27). Meanwhile Japan's most interesting hedge is not a mine at all. Daido Steel is developing magnets that need no Chinese dirt at all, and the deep-sea drillship Chikyu lifted about 50 tonnes of rare-earth-bearing mud from 5,569 meters near Minamitorishima in February, mud that runs about 54 percent medium and heavy elements (Observer Research Foundation, Jul 27; JapanTalkback, Aug 2). Both are real. Neither is 2026.
So what breaks first? Not the factory gate this quarter, but the quiet decisions upstream: a procurement officer at a Japanese automaker specifying a heavier magnet for a 2028 model, a European supplier qualifying Malaysian oxides, a bank pricing a Carester offtake. Stockpiles drain by allocation, not announcement, and the first rationing memo will never be a press release. The pressure also runs backward on Beijing: every month the valve stays shut, the 64 billion euros of allied projects get their funding committees moved up a quarter, which is the one outcome China's export control was designed to prevent.
The judgment the record supports is uncomfortable for both capitals. Japan did nearly everything the 2010 crisis asked of it and still cannot make a hybrid car motor without Chinese ore, because it diversified supply while leaving refining where it was. Beijing, for its part, has learned that a blockade it can dial is worth more than one it must lift, and it is now spending its monopoly to teach that lesson to everyone watching Tokyo. The stockpile is the scoreboard: when Japan starts publishing drawdown figures, the countdown it reveals is the real state of the trade war, no communiqué required.