Geopolitics & Sovereign Positioning

A Public First survey of over 18,000 respondents across 15 countries finds that majorities in 11 nations — including Canada, the UK, and France — believe China is leading the global AI race, according to South China Morning Post. This perception gap between acknowledged capability and trust deficits is geopolitically significant: it means Chinese AI models are winning the competition for adoption among cost-sensitive users even as they lose the competition for institutional confidence. A separate Rest of World report confirms this dynamic domestically — American developers are choosing DeepSeek for its price-performance ratio, rationalising the security trade-off with pragmatic framing.

The trust deficit is China's primary limiting factor in converting perceived capability leadership into geopolitical leverage. Governments and regulated enterprises remain reluctant to build critical infrastructure on Chinese AI. But in the Global South, among SMEs, and among cost-constrained developers globally, the calculus is different — and that is where Chinese models are accumulating deployment scale that will matter for long-run ecosystem influence. This is a direct second-order consequence of US export controls: by constraining Chinese access to frontier chips, Washington accelerated China's incentive to compete on price and openness, producing a competitive model that undercuts the US in exactly the markets America's export control regime cannot reach.

The Shanghai Stock Exchange's Star Market has issued clarified listing standards for unprofitable AI model developers, requiring a minimum anticipated market cap of 4 billion yuan ($591 million) alongside criteria tied to market potential rather than profitability, per South China Morning Post. This is a confirmed policy action — not a proposal — and it directly addresses the capital disadvantage Chinese LLM labs face relative to US counterparts flush with venture and hyperscaler investment. It also reduces exposure to US capital markets at a moment when geopolitical risk makes that dependency strategically uncomfortable for Beijing.

Simultaneously, ByteDance and Tencent have launched AI pricing offensives, deepening a price war that analysts describe as reflecting limited capability gaps across incumbent models, per South China Morning Post. Alibaba has extended its AI push into embodied intelligence with the Qwen Robot Suite, targeting physical-world AI deployment in robotics and autonomous systems — a domain where South China Morning Post analysis suggests China holds a deployment-scale lead in world models due to its broader rollout of robotics and autonomous vehicles. Taken together, these moves describe a coordinated domestic ecosystem push: commoditised AI inference at the software layer, sovereign capital formation at the financial layer, and physical-world AI at the hardware-application frontier.

A South China Morning Post analysis of five Chinese AI models trained on domestic chips confirms the critical asymmetry in China's AI hardware position: domestic chips are now widely used for inference but none of China's leading models are known to have been pre-trained on homegrown silicon. This is the most concrete evidence that US export controls are achieving their primary objective — constraining China's ability to train frontier-scale foundation models — even as they fail to prevent China from deploying and commercialising models at inference scale.

CFMEE's $410 million Hong Kong IPO for lithography and chip-manufacturing equipment, reported by South China Morning Post, is one node in a broader capital mobilisation for semiconductor self-sufficiency. The Infineon GaN patent ban by China's Supreme People's Court — upholding an injunction against the German chipmaker selling GaN products in China — signals Beijing is also deploying IP and court mechanisms to create domestic market space for third-generation chip firms, per South China Morning Post. The pre-training gap is real and persistent, but the trajectory is toward closure — and the timeline matters enormously for strategic planning.

Anthropic is again in conflict with the Trump administration, specifically over its involvement with Secretary Hegseth's Pentagon through projects linked to Mythos and Fable 5, per Foreign Policy. The specifics of the dispute are not fully detailed in available reporting, but the recurrence of this conflict reveals a structural tension in US AI governance: the government's most capable safety-focused AI developer is in a fractious relationship with the defence establishment at precisely the moment the US needs coherent civil-military AI integration to maintain strategic advantage.

This matters beyond the domestic politics. US adversaries and allies are both watching whether Washington can translate its frontier model advantage into integrated military AI capability. A safety-first developer unwilling or unable to work within the current administration's defence framework creates a gap that either goes unfilled — weakening US military AI — or gets filled by less safety-conscious alternatives. The Foreign Policy framing suggests this is an ongoing conflict, not a resolved one, making it an active variable in US AI power projection.

Christian Noyer, a founding ECB vice-president and lead author of a major European capital markets integration proposal, has publicly advocated welcoming Chinese investment and deepening China industrial partnerships as Europe seeks to reduce US technology dependence, per South China Morning Post. This is a notable signal from a senior European financial architect — not a binding policy position, but a credible indicator of the direction serious European strategic thinking is moving under US pressure. It sits in direct tension with EU technology sovereignty ambitions and with member state commitments to US-aligned export control regimes.

The Diplomat's analysis that Asia is accelerating on AI while Europe remains structurally behind, per The Diplomat, reinforces the urgency. The plurilateralism piece in Foreign Policy frames coalitions of the willing on AI, climate, and critical minerals as the operative multilateral architecture replacing formal institutions — which means Europe's AI alignment will be determined by which coalitions it joins, not which rules it writes. The trilemma is real: deeper US alignment preserves security relationships but locks in technology dependence; deeper China engagement reduces that dependence but risks security exposure and secondary sanctions; autonomous capacity requires capital and time Europe does not clearly have.