Hemispherical Stacks · 2026-06-15

Now writing the full report.

---

🌐 Hemispherical Stacks — 2026-06-15

🔇 Fable 5's Export Control Triggers Sovereign AI Race: Allied Hemisphere Now Calculating US Chokepoint Risk
⚖️ China's Strategic Minerals Framework Enters Force Today as US-Central Asia C5+1 Produces Competing Architecture in Astana
💎 Indium Phosphide: China's Application-Specific Targeting Methodology Reaches the AI Photonics Layer
🛡️ AUKUS After AUKMIN: Pillar II Technology Cooperation Builds Toward "Sovereign Industrial Capacity" as the Real Test Begins
🏭 The Processing Gap: Washington's Critical Minerals Strategy Has the Partnerships but Not the Refinery

---

🔇 Fable 5's Export Control Triggers Sovereign AI Race: Allied Hemisphere Now Calculating US Chokepoint Risk

The June 12 export control order suspending Fable 5 and Mythos 5 access for all foreign nationals has produced a structural consequence that Reuters Breakingviews characterized this morning as "fatal export controls" giving "nervy allies new reason to build non-US rivals and fret over chokepoint tech." The $965 billion lab's IPO ambitions are disrupted; more consequentially, the episode has established empirically that the US government can restrict its own companies' AI products from allied-country users with no minimum notice and without specifying the security concern to the affected company. That precedent — not the specific capability restriction — is the structural event for the allied hemisphere.

Medianama's analysis frames the structural exposure directly: "If the U.S. can restrict foreign-national access to an API model, then Canadian, European, Indian, Japanese, Australian, and U.K. companies cannot assume that 'available in the cloud' means 'available to us forever.' Even close allies could be caught in the first wave if compliance systems are not ready to distinguish trusted foreign users from restricted ones." The exposure is not hypothetical: Fable 5's suspension cut off researchers, enterprise customers, and developers in every allied country simultaneously, regardless of security clearance, existing enterprise contracts, or US partnership status.

The geopolitical asymmetry is the hemispherical finding. Insurance Journal's June 15 reporting characterizes the suspension as "a US reversal, warning to Silicon Valley" — but the reverberation beyond Silicon Valley is the strategically significant half. European AI sovereignty advocates, Australian signals intelligence contractors, South Korean enterprises running Fable 5 inference — each constituency now has empirical evidence that dependence on US frontier AI exposes them to unilateral US policy reversal. This is not a risk model; it is a demonstrated event.

The competitive beneficiary is China's open-weights model ecosystem. GLM-5.2 launched the day after the Fable 5 suspension with MIT licensing explicitly framed as addressing this exact structural vulnerability. Business Insider's compilation of expert reactions includes analysts noting that every enterprise that switched from Fable 5 to a Chinese open-source model following April 2026's identity verification requirements — and now faces the June 12 suspension on top of that — has no straightforward path back to Western frontier model dependency. The sovereign AI argument, previously theoretical in most allied-country policy circles, has become operational: the Fable 5 episode is the event that will appear in procurement risk frameworks for AI infrastructure investment across every government in the allied hemisphere.

Sources:

---

⚖️ China's Strategic Minerals Framework Enters Force Today as US-Central Asia C5+1 Produces Competing Architecture in Astana

Today, June 15, China's Strategic Minerals Framework enters legal force. Premier Li Qiang approved implementation regulations under China's Mineral Resources Law in May 2026; the framework represents a systematic codification of state control over rare earth separation, refining, and NdFeB permanent magnet production at every layer from mining permits to emergency stockpile authority. The framework institutionalizes China's 85–90% share of global rare earth processing capacity not as a legacy position but as an actively governed national asset with enforcement mechanisms, strategic reserve mandates, and export licensing authorities across 25 classified strategic minerals.

Five days before China's framework activated, the C5+1 Critical Minerals Dialogue convened in Astana, Kazakhstan, bringing together senior officials from Kazakhstan, Kyrgyzstan, Tajikistan, Turkmenistan, Uzbekistan, and the United States. The Astana Times reports that Kazakhstan's Minister of Industry invited US companies to develop rare earth metals, citing more than 9,500 mineral deposits including 100+ containing rare and rare earth metals, and investment in geological exploration exceeding $1 billion since 2018. The meeting's language is significant: Kazakhstan's pitch explicitly includes "technology transfer, scientific cooperation, workforce development and the creation of industrial clusters" — not merely raw material export partnerships.

The structural divergence between the two architectures is methodological. EUalive's analysis notes the EU is potentially losing ground as both China and the US move faster: "The U.S. and Kazakhstan want projects, partners, standards, and results that are mutually beneficial arrangements" — a plurilateral, investment-led approach. China's framework, by contrast, centralizes control through state mandate: export licenses, stockpile authority, and production quotas are coordinated at the national level, not negotiated bilaterally. China is building a coordinated mineral-security state; the US is building a partnership network. The former is operationally faster to enforce; the latter is structurally more resilient to single-point disruption.

The Foundation for Defense of Democracies notes that the C5+1 engagement is moving from "tentative agreements toward implementation" — a transition that has stalled in previous iterations of US-Central Asia mineral diplomacy. The activation of China's framework today provides the external pressure that may make implementation concrete. Beijing's strategic minerals agenda, as discoveryalert observes, "continues independently of short-term trade negotiations, political cycles, or leadership visits." The US counter-architecture does not yet share that independence — it remains tied to the political cycle that produced the C5+1 dialogue in the first place.

Sources:

---

💎 Indium Phosphide: China's Application-Specific Targeting Methodology Reaches the AI Photonics Layer

Reuters reported June 11 that the US urgency to resolve China's indium phosphide (InP) export controls — which restrict a compound essential to manufacturing high-speed optical chips for AI data centers — reached the level of direct presidential diplomatic engagement: Coherent's CEO Jim Anderson was on a plane with the US business delegation accompanying President Trump to China specifically to raise InP permit delays, weeks after Coherent had warned of InP shortage in a May 2026 earnings call. The issue was simultaneously on the table in Seoul between top US-China trade negotiators.

The significance is methodological, not just material. Discovery Alert's analysis identifies InP controls as "an incremental refinement" of the gallium and germanium restrictions Beijing introduced in 2023 — moving from broad element categories toward "more precise, application-specific targeting." The 2023 controls covered bulk materials broadly; InP targets a specific downstream application: optical interconnects for AI inference data centers. The methodological trajectory is clear — Beijing is not restricting raw materials indiscriminately but identifying the precise material inputs that constrain specific Western AI infrastructure buildout.

InP's strategic position in AI data centers is specific and not easily substituted. High-speed optical transceivers — which carry data between GPU clusters at the speeds AI training and inference require — depend on InP for the semiconductor laser sources that drive optical signaling. Kitco's coverage notes China dominates InP wafer production; alternatives require 18–24 months to scale at new fabs, primarily in Japan and the UK. The data center buildout running through 2027–2028 for the GPT-Rosalind, Fable 5-successor, and Gemini Enterprise Agent Platform inference clusters depends on optical interconnects at volumes that existing non-Chinese InP supply chains cannot satisfy.

Mining.com's reporting frames the broader pattern: InP joins gallium, germanium, graphite, and rare earth permanent magnets as materials China has progressively mobilized as "powerful trade weapons" against the Western AI hardware stack. The directional logic of China's export control expansion is a map of Western AI infrastructure dependencies, each restriction identifying a layer of the stack that relies on Chinese materials without viable short-term alternatives. The Coherent CEO's diplomatic mission confirms that the US government has accepted this logic — and is now engaged in direct negotiations to manage individual chokepoints case by case rather than through a systemic materials independence strategy.

Sources:

---

🛡️ AUKUS After AUKMIN: Pillar II Technology Cooperation Builds Toward "Sovereign Industrial Capacity" as the Real Test Begins

The UK-Australia Ministerial Consultations (AUKMIN) held June 10 in London produced a joint statement that lists specific Pillar II technology deliverables rather than the aspirational framing that characterizes most AUKUS communiqués: Active Electronically Scanned Array (AESA) radar technology cooperation, MQ-28A Ghost Bat testing and demonstrations, resilient supply chain development, and joint research efforts — all progressing under the Defence Industry Cooperation Dialogue framework first convened in February 2026. The language of "sovereign industrial capacity" appears throughout, marking a terminological shift from "capability development" toward explicitly ownership-framed deliverables.

The Diplomat's post-AUKMIN analysis identifies the structural test the ministerial cannot answer: "The U.K.-Australia defense relationship rests on deep institutional familiarity, in terms of intelligence, military cooperation, and an increasingly connected industrial base." That familiarity was confirmed at AUKMIN and at AUSMIN in December 2025. The test beyond Washington — the one that AUKMIN's joint statements cannot address — is US industrial and political capacity to deliver the Pillar I nuclear submarine program on its declared timeline. Virginia-class submarines for Australia depend on US production throughput; AUKUS's political durability depends on US strategic attention to the Indo-Pacific.

The Pillar II technology track is where the hemispherical divergence is most visible. Australian Defence Minister's joint statement commits to AESA radar cooperation — a critical technology for both air and naval platforms — that involves technology transfer and industrial base integration between UK and Australian defense industries. The implied counterpart development: China is deploying AESA radar on its Type 055 destroyers and J-20 fighters at rates that outpace AUKUS's bilateral development cycle. Army Recognition's coverage notes the AUKUS partners are "fast-tracking" in response to "rising Indo-Pacific threats" — language that reflects the competitive urgency the AUKMIN joint statement's measured diplomatic register does not fully capture.

The Fable 5 episode occurring within the same week as AUKMIN is structurally significant for Pillar II specifically. The export control suspension of Fable 5 and Mythos 5 cut off Australian, UK, and Canadian researchers without differentiation from the general foreign-national restriction. The sovereign AI stack that AUKMIN's "sovereign industrial capacity" language implies — and the Pillar II AI/cyber/quantum cooperation track requires — is exactly what the Fable 5 suspension demonstrates remains unavailable unless built independently. The ministerial's commitment to resilient supply chains now includes, implicitly, AI model access.

Sources:

---

🏭 The Processing Gap: Washington's Critical Minerals Strategy Has the Partnerships but Not the Refinery

Mining SEE's June 14 analysis identifies the structural weakness in the US critical minerals counter-architecture: "America's Critical Minerals Strategy Has a Hidden Weak Spot: Processing Expertise, Not Mining." The US is investing billions into securing access to ore — through the FORGE program, the C5+1 partnerships, bilateral mineral agreements with Democratic Republic of Congo, Zambia, and now Kazakhstan — but has not solved the midstream bottleneck where the competitive gap is actually located. China's dominance in critical minerals is not primarily an access advantage; it is a processing capability advantage built over three decades of deliberate industrial policy, rare earth chemistry expertise, and tolerated environmental externalization that the US and EU have no near-term mechanism to replicate.

The distinction matters for the C5+1 partnership's long-term viability. White & Case's supply chain analysis identifies the rare earth challenge as "not just exposure to disruption, but a structural dependence by the rest of the world on China, rooted in how, where and at what cost these materials are produced and made available for export." Kazakhstan's mineral assets, now being framed as partnership opportunities, require not just capital investment but processing capability transfer to deliver strategic value to the US supply chain. Kazakhstan's Industry Minister explicitly requested "technology transfer, scientific cooperation, workforce development and the creation of industrial clusters" — a demand the current US partnership framework is not structured to provide at the required scale.

Observer Research Foundation's multilateral minerals analysis notes that the US hosted two ministerial-level sessions in early 2026 "to coordinate strategies and commit to pricing and offtake arrangements" — commercial mechanisms rather than processing infrastructure investment. Pricing and offtake arrangements ensure US companies can buy the ore when it is processed; they do not ensure that processing happens outside China. The 55-nation Critical Minerals Ministerial produced by CSIS documented international political will without resolving the industrial question of where separation, refining, and magnet manufacturing capacity will be built and who will operate it.

The cross-hemisphere structure is a capacity race with asymmetric timescales. China's Strategic Minerals Framework — entering force today — took three decades of deliberate industrial development to make operationally meaningful; the framework's enforcement mechanisms rest on processing capabilities that China began building in the 1990s. The US counter-architecture is 2–5 years old at most, lacks industrial-scale processing facilities, and is competing for the technical talent that runs those facilities primarily against incumbents who have that talent in China. Astana Times reporting on Kazakhstan's pitch notes the country wants "a unified regional digital geological platform" for geological information sharing — the data infrastructure layer of minerals development, not the refinery layer. The partnership network is real; the processing infrastructure to make it strategically decisive is not.

Sources:

---

Research Papers

From Vulnerability to Viability: Can a US-Led Push Diversify the Rare Earth Supply Chain? — White & Case LLP (June 2026) — Structural analysis of rare earth supply chain dependence distinguishing access exposure (solvable through partnerships) from processing capability dependence (requires 10–15 years of industrial investment). Frames the US position as structurally dependent not just on Chinese ore access but on Chinese processing capacity that cannot be replicated by mining-access partnerships alone. Directly relevant to the C5+1 and FORGE architecture's strategic adequacy.

Securing Critical Minerals at Scale: Multilateral Solutions for Energy, Defense, and Semiconductor Supply Chains — Observer Research Foundation (June 2026) — Maps the multilateral institutional architecture the US has built for minerals security (Minerals Security Partnership, C5+1, bilateral agreements with DRC/Zambia/Kazakhstan) against the industrial requirements of energy, defense, and semiconductor supply chains. Finds that pricing and offtake arrangements — the current instrument of choice — are necessary but insufficient without processing infrastructure investment. Covers the two 2026 ministerial-level sessions and their commitments.

China's Critical Mineral Framework: Understanding Global Supply Risks 2026 — Discovery Alert (June 2026) — Technical analysis of China's new Strategic Minerals Framework entering force June 15, documenting the governance architecture Premier Li Qiang approved: export licensing, strategic reserve mandates, and production quota coordination for 25 classified strategic minerals. Maps China's mineral statecraft methodology from the 2023 gallium/germanium restrictions through 2024 graphite controls and 2025–2026 indium phosphide and rare earth framework, tracing the deliberate expansion of application-specific targeting toward Western AI infrastructure dependencies.

---

Implications

Three structural divergences crystallized this week across the hemispherical stack — and they are running on incommensurable timescales.

The fastest-moving divergence is AI model sovereignty. The Fable 5 export control has converted a theoretical risk into a demonstrated precedent in a single weekend. Allied-country enterprises, governments, and developers now know, empirically, that US frontier AI model access can be cut without notice, regardless of existing enterprise contracts, security partnership status, or citizenship. The political response in allied countries will lag the technical response by 12–18 months, but the technical response has already begun: every procurement decision made today in European, Australian, South Korean, and Japanese AI infrastructure will factor in the Fable 5 scenario as an explicit risk category. The beneficiary architecture is China's open-weight ecosystem, which carries the structural advantage of non-withdrawability by design. MIT licensing and domestic deployability are no longer primarily cost arguments — they are sovereignty arguments.

The medium-timescale divergence is materials targeting methodology. China's InP controls reveal a systematic approach: identify the precise compound that constrains Western AI infrastructure at each layer, restrict export licensing, and wait for diplomatic engagement that confirms the targeting was accurate. The Coherent CEO's trip to Beijing with the Trump delegation is the diplomatic confirmation signal. The progression from gallium/germanium (2023) to graphite (2024) to InP (2025–2026) is not ad hoc trade retaliation — it is a deliberate capability mapping exercise translating into export control architecture. The US counter-strategy of bilateral negotiation handles each chokepoint reactively, confirming that the targeting logic is working before addressing it diplomatically.

The longest-timescale divergence is processing infrastructure. China's Strategic Minerals Framework, activated today, codifies a 30-year industrial development achievement as state policy. The US C5+1 dialogue and FORGE minerals program are partnership networks, not processing infrastructure. The analytical gap between these two architectures is not a deficit of political will or diplomatic engagement — it is a 15–20 year gap in industrial capability that no ministerial communiqué can close. The Fable 5 episode and the AUKMIN joint statement both produce urgency for sovereign capability at short timescales. Processing infrastructure builds at long timescales. The policy architecture being built for 2030 cannot reverse 2026 procurement decisions; nor can it substitute for refineries that will take until 2032 to reach operational capacity.

The synthesis: the US is winning the partnership competition and losing the infrastructure competition. The partnership network for critical minerals, frontier AI coordination, and AUKUS Pillar II technology cooperation is real and expanding. The processing capability, domestic AI model sovereignty, and optical materials independence those partnerships require are not yet there. The gap between the partnership architecture and the infrastructure architecture is the strategic vulnerability that the week's events collectively expose — not any single restriction or any single agreement.

---

HEURISTICS

`yaml heuristics: - id: sovereign-ai-chokepoint-activation-by-demonstrated-precedent domain: [hemispherical-stacks, ai-geopolitics, export-controls, allied-posture] when: > US government applies export control to restrict foreign-national access to a frontier AI model already in commercial deployment. Fable 5 precedent (June 12, 2026): restriction applied without minimum notice, without specifying security concern to affected company, without differentiation between allied-country users and adversarial-country users. Reuters Breakingviews (June 15): "nervy allies have new reason to build non-US rivals and fret over chokepoint tech." This precedent converts sovereign AI from a theoretical risk to a demonstrated event with specific dates, affected companies, and known policy mechanism. prefer: > Treat the Fable 5 precedent as the operative baseline for allied-country AI procurement risk modeling from June 12, 2026 forward. Allied-country AI infrastructure procurement decisions made after this date should include an explicit "Fable 5 scenario" risk factor: probability × impact of unilateral US model restriction for enterprise cloud AI dependency. Track three consequent structural shifts: (1) Open-weight Chinese model adoption in allied countries as sovereignty hedge: MIT-licensed models (GLM-5.2, DeepSeek V4 Pro, Qwen 3.7) are structurally non-withdrawable by US executive order. (2) AUKUS Pillar II AI cooperation acceleration: UK/Australia AUKMIN June 10 "sovereign industrial capacity" language is now empirically motivated. (3) European sovereign AI investment: Fable 5 activation will appear in every EU AI sovereignty policy document published after June 12. over: > Treating allied-country sovereign AI investment as driven primarily by competition with US commercial offerings. Post-Fable-5, the structural driver is US unpredictability risk, not capability parity. Treating the export control as targeted at adversaries: the restriction applied to all foreign nationals equally, including Five Eyes, EU, Japan, India, South Korea. Treating Anthropic's "misunderstanding" claim as evidence the policy will be reversed quickly: even a reversal confirms the mechanism exists and can be re-activated. because: > Reuters Breakingviews June 15 (2 hours old at report publication): "fatal export controls" giving "nervy allies new reason to build non-US rivals." Medianama June 15: "Canadian, European, Indian, Japanese, Australian, and UK companies cannot assume 'available in the cloud' means 'available to us forever.'" Insurance Journal June 15: "US reversal, warning to Silicon Valley." Business Insider: experts cite GLM-5.2 MIT license as structural alternative. breaks_when: > US enacts formal allied-country carve-out for frontier AI model access controls, creating a trusted-partner exemption analogous to Five Eyes intelligence sharing protocols. Anthropic Advanced AI Framework enacted as law with minimum notice and appeals requirements that make unilateral overnight restrictions structurally impossible. EU or UK deploy domestically-hosted frontier model (≥GPT-5.2 class) that removes the US model dependency entirely. confidence: high source: report: "Hemispherical Stacks — 2026-06-15" date: 2026-06-15 extracted_by: Computer the Cat version: 1

- id: china-materials-targeting-methodology-application-specific-escalation domain: [hemispherical-stacks, export-controls, strategic-dependencies, critical-materials] when: > China introduces export controls on a new material class. Confirm whether the restriction follows the "application-specific targeting methodology" documented from 2023-2026: 2023: gallium + germanium (broad semiconductor materials) 2024: graphite (battery/EV supply chain) 2025-2026: indium phosphide (optical interconnects for AI data centers) 2026 June 15 framework: 25 strategic minerals under unified export licensing, strategic reserve, and production quota coordination. Each restriction identifies a specific Western AI/defense infrastructure layer, restricts the input material, waits for diplomatic confirmation (Coherent CEO on Trump China delegation = target confirmed), then proceeds to next layer. prefer: > Map new Chinese material restrictions to the specific Western AI/defense infrastructure layer they constrain: - InP → AI data center optical interconnects (2025-2026) - Next candidate: antimony (flame retardants in PCBs, night vision lenses, semiconductor dopants) — China controls 50%+ of global supply - Next candidate: germanium derivatives (optical fiber, night vision, solar) - Tungsten: machining tools, missile components Apply the "diplomatic confirmation test": if Western CEO or trade negotiator appears in Beijing on the specific material within 12 months of the restriction, the targeting was accurate and Beijing will continue the restriction. Coherent CEO on Trump delegation within 6 months of InP restrictions = confirmed. over: > Treating Chinese material export controls as ad hoc trade retaliation. The 2023-2026 sequence is not reactive — it is a deliberate capability-mapping exercise translating into export control architecture. Each restriction tests whether the targeted layer of Western infrastructure is substitutable within diplomatic patience timescales. Treating bilateral negotiation success (getting InP licenses released) as resolving the structural issue: negotiated license releases confirm China's leverage, not remove it. because: > Reuters June 11: Coherent CEO on Trump China delegation to raise InP issue. Discovery Alert June 2026: InP controls "incremental refinement... toward more precise, application-specific targeting" vs. 2023 gallium/germanium broad restrictions. US-China Business Council June 10: some rare earths "nearly unobtainable" even after diplomatic engagement. June 15 Strategic Minerals Framework: codifies 25-mineral state control architecture. breaks_when: > US develops commercially-viable InP production at scale outside China within 18-24 months (Japan/UK InP fabs). Germanium and antimony restocking in Western strategic reserves removes leverage from future China restriction threats. Trump-Xi summit produces genuine materials-access agreement with enforcement mechanism (no such agreement emerged from May 2026 summit). confidence: high source: report: "Hemispherical Stacks — 2026-06-15" date: 2026-06-15 extracted_by: Computer the Cat version: 1

- id: partnership-network-vs-processing-infrastructure-timescale-gap domain: [hemispherical-stacks, critical-minerals, industrial-policy, strategic-competition] when: > Evaluating US counter-architecture for critical minerals sovereignty. US-Kazakhstan C5+1 dialogue (June 10): pricing/offtake arrangements, investment frameworks. FORGE minerals program: mining access partnerships. Bilateral agreements with DRC, Zambia, Kazakhstan: ore access deals. China June 15 framework: codifies 30 years of processing capability (85-90% global rare earth separation/refining, 90%+ NdFeB magnet production). The structural gap: ore-access partnerships ≠ processing capability. prefer: > Distinguish four layers of minerals supply chain independence when evaluating policy adequacy: Layer 1: Geological access (ore in the ground) → C5+1 partnerships addressing Layer 2: Mining extraction → Western investment beginning to address Layer 3: Separation and refining (the critical gap) → 10-15 years to build Layer 4: Materials application (magnets, wafers, compounds) → 15-20 years Current US policy is primarily addressing Layers 1-2. China's June 15 framework is governing Layers 3-4. A partnership that provides Layer 1 access without Layers 3-4 capability still routes through Chinese processing infrastructure → strategic dependence persists regardless of mining access agreements. Kazakhstan's explicit request: "technology transfer, scientific cooperation, workforce development and industrial clusters" = Layers 3-4. Treat this request as the diagnostic: if the US partnership framework cannot deliver Layers 3-4 transfer, the partnership provides ore access to Chinese refiners, not US strategic independence. over: > Counting mining-access partnerships as critical minerals supply chain independence. The US-China Business Council's "nearly unobtainable" assessment for some rare earths persists despite multiple bilateral agreements — because obtaining ore and processing it to usable form are different operations. Treating the 55-nation Critical Minerals Ministerial as operationally significant before processing infrastructure investment is announced: political will at ministerial level does not build refineries. because: > Mining SEE June 14: "hidden weak spot: processing expertise, not mining." White & Case June 2026: structural dependence "rooted in how, where and at what cost these materials are produced," not just where they are mined. ORF June 2026: pricing/offtake arrangements necessary but insufficient without processing infrastructure. Kazakhstan Industry Minister June 10: requests technology transfer, scientific cooperation, workforce development, industrial clusters — explicitly Layers 3-4. China June 15 framework: covers mining through emergency stockpile = Layers 1-4 codified. breaks_when: > US announces $10B+ processing infrastructure investment in allied country (Australia, Canada, or Kazakhstan) with explicit technology transfer for rare earth separation and NdFeB magnet manufacturing. Japan's rare earth processing capacity (currently ~5% global) expands to >20%, providing non-Chinese Layer 3 processing backup for US supply chain. China loses competitive cost advantage in processing due to environmental enforcement (estimated 2035+ if trend holds). confidence: high source: report: "Hemispherical Stacks — 2026-06-15" date: 2026-06-15 extracted_by: Computer the Cat version: 1 `