China exported less rare earth material in the first half of 2026 than a year earlier, but the headline tonnage hides the real move: Beijing is throttling the separated oxides and metals the West still cannot make at scale, while flooding the world with the finished magnets that lock in its downstream advantage. For the AI buildout, which needs those magnets in data center cooling, storage, and comms hardware as much as defense needs them in guidance and radar, the supply risk did not ease. It relocated.

What’s happening

China’s total rare earth exports fell to 30,482.8 tonnes in the first half of 2026, down 6.4 percent from 32,569.3 tonnes in the same period of 2025, according to General Administration of Customs data reported by Global Times on July 14, 2026. June exports were 5,104.8 tonnes.

The aggregate decline masks a split by product and by destination. Digitimes, reporting the same customs data on July 21, 2026, said Chinese exports of rare earth permanent magnets reached record levels in the first half, flowing freely to most of the world but throttled to the United States and, since January, newly restricted to Japan.

The upstream squeeze is where the leverage sits. Silverado Policy Accelerator, drawing on Chinese customs data through May, reported that shipments of export-controlled rare earth compounds and metals to the United States fell to zero in May 2026, the first zero reading since September 2025, after a brief recovery in March and April. Terbium exports globally stayed under 1 tonne in May. Dysprosium, the exception, ran near historical levels, most of it going to South Korea (Silverado, June 26, 2026).

Japan is the other pressure point. China’s exports of rare earth compounds and metals to Japan have stayed minimal since Beijing imposed new limits in January 2026 (Silverado, June 26, 2026). At a recent G7 session, Japanese Prime Minister Sanae Takaichi raised concerns over China’s restrictions; a Chinese Foreign Ministry spokesperson said on June 18 that China had banned dual-use exports to Japanese military users (Global Times, July 14, 2026).

The pattern is deliberate. By restricting upstream oxides and metals while growing downstream magnet shipments, Beijing applies pressure exactly where foreign manufacturers depend on Chinese inputs, and competes against those same manufacturers in the finished goods they still need to buy.

Brazil angle

Brazil holds one of the most credible non-Asian answers to the heavy rare earth chokepoint. Serra Verde’s Pela Ema operation in Goiás, in commercial production since 2024, produces the full magnetic rare earth basket, including the heavy elements dysprosium and terbium that few facilities outside China make at commercial scale. It is targeting 6,400 tonnes of rare earth oxides a year by the end of 2027 over an initial 25-year mine life (Discovery Alert, April 24, 2026).

That is precisely the heavy basket, terbium and dysprosium, that China just proved it can zero out. The strategic value did not go unnoticed: USA Rare Earth agreed in April 2026 to acquire Serra Verde for about 2.8 billion US dollars, backed by roughly 565 million US dollars in US International Development Finance Corporation financing, to build a mine-to-magnet chain from Brazilian ore to US processing (Discovery Alert, April 24, 2026). The gap Brazil still has to close is separation and refining on its own soil, a final investment decision the company has signaled for early 2027. Until that exists, Brazilian ore still travels to be turned into oxide elsewhere.

Africa is racing on the same track. At Angola’s Longonjo project, Pensana reported in June that construction was 22 percent complete, with first mixed rare earth carbonate targeted for 2027, a 250 million US dollar build with a 20-year mine life aimed at supplying US, European, and Japanese magnet makers and railed to the Atlantic through the Lobito Corridor (Mining Weekly, July 10, 2026). Longonjo and Serra Verde are both Global South projects trying to exist at scale before the next Chinese cutoff, and separated heavy oxide capacity outside China remains the missing link for both.

US angle

Washington has moved from rhetoric to price support. The Department of Defense partnership with MP Materials set a 10-year floor of 110 US dollars per kilogram for neodymium-praseodymium oxide, well above the sub-60 US dollars per kilogram that NdPr fetches inside China, effectively a subsidy to stand up domestic magnet supply (Center on Global Energy Policy, Columbia University). The Serra Verde acquisition extends the same logic across a border: US capital and US development finance securing a Brazilian heavy rare earth resource because the domestic mine base cannot supply dysprosium and terbium on its own. The May zero reading for US-bound compounds and metals is the reason the urgency is real rather than theoretical.

China angle

Beijing is running a value-chain strategy, not a blunt embargo. Growing magnet exports keep global buyers hooked on Chinese finished goods and undercut nascent Western magnet makers on price, while zeroing out oxides and metals to specific destinations denies those same rivals the feedstock to build independence. The Japan military end-use ban and the US compounds cutoff show the tap is now a policy instrument keyed to destination and end use, not a market clearing at a price. China still controls the separation step that turns mixed concentrate into the individual heavy oxides the rest of the world lacks, which is why record magnet volumes and targeted upstream cutoffs are not a contradiction. They are the same play.

What it means

The concentration risk that SOV50 tracks, with rare earths dominated by China at roughly 70 percent of supply, is being actively priced by policy, not just structure. The diversification thesis behind SDX now has a concrete test case in Brazil: whether Serra Verde and its US acquirer can convert ore into separated heavy oxides outside China before the next cutoff. Magnets flowing at record volume is not reassurance. It is the sound of the dependency being reinforced one shipment at a time.

What to watch