The material assembled here does not describe two separate industrial disputes; it describes a single structure of reciprocal containment that now conditions advanced technology supply chains. China has concentrated the midstream processing of rare earths to a degree that allows relatively small export actions to disrupt far larger downstream markets, while the United States has built a parallel control position over advanced semiconductors and the equipment used to make them. Both levers have already produced production stoppages and forced sourcing shifts, and both are currently held in temporary suspension rather than resolved.
The concentration of rare-earth processing is the most heavily corroborated fact in the evidence. In 2024 China accounted for roughly 91% of global rare-earth refining and separation 4,15 and 94% of global sintered permanent-magnet production 4,15. The same structure appears in multiple independent measurements: 60% of mining for magnet rare earths 32, about 91% of global rare earth separation and refining 32, more than 90% of global rare-earth processing capacity distinct from mining 14, and 94% of sintered magnet output cited to the IEA 30,37. The United States, by contrast, depended on imports for more than two-thirds of the rare earth elements it used in 2025 32, and China processed most minerals used globally 32.
The interesting question is not whether that concentration is large, but why it persists and what it permits. China’s position is not primarily a mining monopoly; it is a midstream processing and manufacturing dominance that combines capacity with know-how in separation, metallisation, alloy and magnet manufacturing 15 and dominates the heavy rare earths dysprosium and terbium 20. The IEA commentary supplies the reason small volumes matter: export restrictions on relatively small mineral volumes can have outsized impacts on industries worth trillions 7, with full implementation potentially putting USD 6.5 trillion in annual sales at risk across energy, automotive, electronics and defence outside China 7, against an annual magnet rare earth market of less than USD 10 billion 7.
That is a structure in which the short-run elasticity of substitution is very low: a buyer cannot quickly replace separated heavy rare earths or certified finished magnets when licensing is tightened. The long-run adjustment, by contrast, is exactly what the United States is now attempting, and it operates on a different time horizon from the export action itself.
The Mechanics of Export Control, and What They Have Already Done
Beijing operationalized its position through layered controls beginning in 2025. In April 2025 China imposed licensing requirements on exports of seven heavier rare earths 32, also described as export controls on seven heavy rare-earth elements and related magnets 4,7. On October 9, 2025 China’s Ministry of Commerce announced six export control measures 32; under Announcement No. 61, Chinese-origin rare earths are subject to control when they make up 0.1% or more of a product’s value 32, meaning a foreign factory needs a Chinese license to export if that threshold is met 32 and thereby affecting which products factories abroad may export 32. Beijing had already banned exports of separation technology in December 2023 32, and on June 22, 2026 placed 10 US companies on its export control list 32, a targeted ban covering 10 US companies 32 under which Chinese and foreign firms are prohibited from selling the listed firms goods with civilian and military uses 32.
It is important to distinguish the parts of this apparatus that are paused from the parts that are not. The April heavy-rare-earth licensing was not suspended 32 and would still apply even if broader suspensions are extended 32, as would the June 22 targeted-firm restrictions 32. The truce is therefore not a clean return to the status quo ante; it is a partial relaxation with residual constraints still in force.
The disruption already documented is not hypothetical. China’s 2025 controls demonstrated strategic importance through concentration 15, with companies outside China including automakers struggling to obtain permanent magnets 15, some automakers reducing output when controls tightened in 2025 28, cutting output or halting production for lack of supply 37, and reducing or halting production in 2025 according to the X post account 30. Ford reportedly had to idle its Chicago assembly plant for a week in May 2025 because of a magnet shortage tied to export rules 32 and could resume only with Beijing’s approval 32. Rare-earth licensing was explicitly seen as potential bargaining leverage in US-China talks 12, while China’s rare-earth magnet exports to the United States fell 21% month over month to 512 tonnes in August 12.
The American Rebuild: A Long-Run Adjustment with Short-Run Costs
The persistence of the leverage is a problem of time horizons. Rebuilding a separated rare-earth supply chain involves longer timelines and billions in startup costs 32, faces long lead times, expertise gaps, technology restrictions and expensive environmentally challenging separation 32, and requires separating minerals from ore and from each other, an intensely difficult process 32. A 2010 Government Accountability Office estimate put rebuilding at up to 15 years 32, repeated as as long as 15 years 32. The human and technological constraints are specific: China has thousands of experienced engineers while the United States has far fewer 32, and industrial-scale cutting-edge technologies were developed in China 32.
The critical vulnerability is not mining or finished-magnet assembly alone; it is midstream chemical separation and refining 16. China holds an effective near-monopoly in separating individual elements including neodymium, praseodymium, dysprosium and terbium 16. Separation requires specialized engineering 16, hydrometallurgical processing 16 and management of toxic or radioactive waste 16, with high capital burdens deterring private investment without government support 16. The United States has deposits but few mines 32, and historically buyers could obtain material below self-production cost 32 while avoiding cleanup bills and bad press 32, a pattern through which the United States lost leverage by outsourcing processing 32 until separation capability became concentrated 32.
The adjustment has now begun, but it is early. More than US$7 billion has been committed since April 2025 to rebuild capacity 32, also stated as spent since April 2025 32, including a 2026 Commerce Department letter of intent for US$277 million in direct funding to USA Rare Earth 32. MP Materials and USA Rare Earth are identified as the federally backed firms furthest along in building a mine-to-magnet chain 32, while Australia and Japan are starting capacity 32, Japan is described as having some success 32, and Lynas is cited as allied capacity 16 including Lynas’s Kalgoorlie facility as allied mixed carbonate production 16. These are real, but they are long-run increments to a stock of capability that the short run has not yet accumulated.
The Hard Deadline at January 1, 2027
A specific regulatory deadline converts that slow rebuild into near-term demand risk. On January 1, 2027 DFARS sourcing requirements are expected to expand to cover the full chain from mining through finished product 30, also stated as extending across the entire chain from mining through finished magnets 28,37. Existing rules already bar magnets melted or produced in covered countries including China 28,37, and the January 1, 2027 extension creates a defined deadline 37. The X post framing is defense procurement and magnet supply 30, arguing the deadline could create demand for compliant non-China suppliers 30 and linking compliant-magnet demand to that date 20, while defense primes without secured supply face an approaching deadline 20. The resulting demand signal is not a general preference for non-China material; it is a specific requirement for compliant, certified, finished magnets.
The problem is that the number of non-China suppliers able to deliver certified sintered magnets at commercial scale under full-chain rules is vanishingly small 30. New mining capacity alone does not provide finished magnets 20; finished magnets require separation, refining, alloying, pressing and sintering 20, and downstream qualified products complement upstream supply 20. Heavy rare earths are a supply constraint 20, with availability of compliant dysprosium and terbium a strategic risk 20. Global non-China heavy-rare-earth output is lagging magnet additions 20, even as 20,000–30,000 tonnes of non-China magnet capacity comes online 20; further growth through 2030–2035 would require stable non-China heavy supply 20.
Evolution Metals illustrates the kind of incremental supply response the deadline is beginning to elicit. The firm is preparing to meet restrictions beginning January 1, 2027 37, aligning non-China capability with that deadline 37 and expanding into processing 37. It has entered a supply contract for non-China critical metals 37 and received a first non-China NdPr shipment intended for defense-compliant production 37, with 5 metric tonnes of NdPr from SRE Vietnam delivered July 22, 2026 20, secured from SRE Vietnam in July 30 and produced entirely from non-China sources from ore through oxide and metal 33. Future Brazilian or US heavy supplies remain contingent 20.
The Semiconductor Mirror: Containment Without Completion
The same pattern of concentrated leverage and incomplete adjustment appears on the semiconductor side, with the positions reversed. Since 2022 the Bureau of Industry and Security has restricted exports of the most advanced AI-training chips and chip-manufacturing equipment 11, primarily for destinations including China 11; Washington continues to restrict advanced computing chips 22, and US controls restrict supply of advanced AI accelerator chips to China and certain markets 39. December 2024 additions covered semiconductor manufacturing equipment 6, 24 categories of equipment 19 and three software-tool categories 19, under Commerce controls covering lithography and memory 21.
Yet containment has been only partly effective. Chinese firms continued to produce increasingly capable models despite controls 10 that were sometimes more cost-efficient 10, pursued self-reliance amid controls 13, and sought overseas computing because of domestic shortages and tightening US controls 29. AI workloads moving offshore have been attributed to restricted advanced-chip access inside China 25, and there are claims that destination controls have not prevented remote access to chips in third countries 36. Controls also reached models and people, with a June 12 Commerce directive suspending foreign nationals’ access to Anthropic’s Fable 5 and Mythos 5 17, lifted 18 days later on June 30 when access was restored 5, also noted as lifted as of June 30 3,17, alongside outbound-investment restrictions for semiconductors 9 and quantum computing 9 involving China, Hong Kong or Macau parties.
The lesson is not that export controls do nothing; it is that their effects are gradual, uneven and capable of being routed around in specific ways. The short-run friction is real, but the adjustment on both sides is also real. That is exactly what one should expect in a market where substitution possibilities are not zero and where remote compute can partially offset local chip constraints.
The Truce as a Temporary Equilibrium
The current calm is best understood as a pause, not a settlement. China’s rare-earth controls are scheduled to remain in suspension through November 10, 2026 1,2,32, the suspension of six controls is scheduled to last until that date 32, agreed after the October 2025 Trump-Xi meeting 32, with suspended rules returning automatically unless both governments extend pauses 32 and the November 10 status potentially linked to the September 24, 2026 Xi-Trump meeting 32. More recent material reports the truce was extended by two months through January 2027 38; the US and China extended the truce to January 10, 2027 34, an arrangement that keeps rare-earth controls suspended 24 and runs through January 10, 2027 24, with rare-earth restrictions paused through January 10, 2027 8 and the truce pausing rare-earth restrictions until that date 8.
The underlying disagreements have not been resolved. Washington is pressing Beijing to loosen controls critical to US manufacturing 23, officials are expected to press for delivery of rare earth minerals 38, and disagreements persist over rare-earth supplies 18. Both chip controls 34 and rare-earth licensing 34 remain unresolved. The truce is therefore better modelled as a temporary equilibrium, a halt in action during which the structural sources of conflict remain intact, than as a new stable state.
A parallel energy signal from October 2026 shows how quickly state supply management can tighten markets even beyond rare earths. Reuters-cited sources reported that Chinese refiners suspended October gasoline exports outside Hong Kong and Macau 27, with refineries halting product exports pending Beijing instructions 26 and PetroChina canceling several cargoes 27 and October gasoline shipments 26,35. Beijing did not formally confirm the restriction 27, the resumption date was unclear 26 and dependent on inventories and post-holiday output 26, reflecting domestic supply concerns 26. This coincided with Russia extending its direct-producer diesel, marine fuel and gas oil ban until October 31 26 under a broader ban until January 31, 2027 26. The episode does not directly determine Alphabet’s supply chain, but it is a reminder that state-managed energy and mineral flows can become binding constraints on short notice.
What This Means for Alphabet
For Alphabet, the collective meaning is that AI leadership will be paced as much by supply-chain and export-license access as by model architecture. That is not a claim about a single quarter; it is a structural statement about the constraints under which compute capacity must be planned. Persistent US chip controls, combined with Chinese self-reliance and remote-compute workarounds, point to a bifurcated compute market. Frontier training capacity, compliance exposure for cloud and data-center operators serving China through overseas facilities 31, and monitoring of China equipment restrictions will shape cost and deployment choices over the relevant planning horizon.
The rare-earth side of the ledger enters through hardware and robotics. Motors depend on rare-earth processing 14, and demand for magnet rare earths has doubled since 2015 37 with a further one-third rise projected by 2030 37. If Alphabet’s hardware and robotics ambitions require permanent magnets, then their cost and availability will be governed by the same midstream bottleneck and the same January 2027 defense and licensing reset as the broader market. The scarcity is not in the ground but in the ability to separate, refine and certify non-China material in commercial volumes, a distinction that survives any temporary truce.
We must therefore distinguish between what the truce does and what it does not do. It buys time; it does not create new non-China separation capacity. It pauses some restrictions; it does not remove the unsuspended April heavy-rare-earth licensing or the June 22 targeted-firm bans. It shifts the immediate risk to January 2027; it does not eliminate the snap-back mechanism by which suspended rules return automatically unless extended. The decision for Alphabet, as for any large buyer of compute and precision hardware, is whether to treat the pause as an opportunity to secure compliant supply or as a reason to defer adjustment. The evidence assembled here suggests the former is the more defensible reading: the long run is arriving, but its arrival is gated by very specific dates, very specific materials, and very specific unresolved disputes.