Geopolitics & Sovereign Positioning

Huawei president He Tingbo unveiled the Tau Scaling Law at a Shanghai conference, presenting a chip architectural framework that the company claims will enable semiconductor performance equivalent to a 1.4nm process node by 2031 — without relying on ASML extreme ultraviolet lithography equipment that US export controls have effectively denied China. The approach leverages chip stacking, interconnect innovation, and system-level architectural optimisation to compensate for the transistor density advantages that EUV enables. Analysts cited by the South China Morning Post describe this as a potential 'DeepSeek moment' for hardware — a workaround that reframes the competitive terrain rather than competing head-on.

The strategic implication is significant but requires careful qualification. This is a roadmap announcement, not a demonstrated product. Analysts note that substantial manufacturing challenges remain, particularly in yield rates and heterogeneous integration at scale, and 2031 is a five-year horizon during which US controls could tighten further or Huawei's claims could prove overstated. However, the directional signal matters regardless: Washington's semiconductor containment strategy has been premised on the assumption that denying advanced lithography tools would cap China's chip ceiling. Huawei is now demonstrating institutional commitment to architectural innovation as an alternative vector. If even partially successful, this shifts the burden back onto US policymakers to identify new chokepoints — and the options narrow as Chinese firms reduce their dependency on each restricted input.

During Pakistani Prime Minister Shehbaz Sharif's visit to Alibaba's Hangzhou headquarters, Sharif issued a real-time request for a comprehensive strategic agreement, which Alibaba negotiated and delivered on-site — a deliberate demonstration of deployment speed that the South China Morning Post frames as matching Sharif's own reputation for rapid project execution. The deal positions Alibaba's Qwen ecosystem — including cloud infrastructure and AI models — as the backbone of Pakistan's digital economy acceleration programme. Pakistan is the world's fifth-most-populous country with a young demographic profile and significant unmet demand for digital infrastructure.

This deal is structurally representative of a broader pattern: Chinese AI and cloud firms are moving faster than US counterparts in formalising government-to-government AI partnerships with emerging economies. Alibaba's Qwen3.7-Max is now in the global top five for coding benchmarks, placing it ahead of OpenAI and Google on that dimension, which provides technical credibility to accompany the commercial and diplomatic speed advantage. The combination of competitive model performance, willingness to negotiate bespoke sovereign agreements, and absence of US export control complications gives Chinese AI platforms a structural advantage in Global South markets that Washington has been slow to contest through technology diplomacy or alternative partnership frameworks.

President Trump explicitly blocked an executive order that would have directed agencies to review advanced AI models for safety concerns, citing competition with China as the rationale. The South China Morning Post notes the irony: China has in fact moved faster on AI regulation in certain domains, including algorithmic recommendation rules and deepfake controls, while the US frames deregulation as the competitive response. The result is a regulatory vacuum where Washington is neither setting domestic safety standards nor leading international governance frameworks.

This divergence has direct alliance consequences. The EU's AI Act is in force. The UK has adopted a principles-based approach with active international engagement. Japan and South Korea are developing their own frameworks. The absence of a US federal AI governance position makes interoperability and mutual recognition agreements harder to negotiate, and it hands China a soft power argument — that Beijing is a more stable regulatory partner than a Washington that changes direction with each administration. The CFR analysis of AI accountability frameworks CFR highlights the governance vacuum explicitly, questioning whether the UN can fill a coordinating role that the US has vacated. The practical answer, absent US leadership, is fragmentation — which benefits the actor most capable of setting de facto standards through deployment scale.

The War on the Rocks analysis War on the Rocks centres on a pointed question Jensen Huang declined to answer directly: if US-made compute trains AI with demonstrated cyber-offensive capabilities and is sold to strategic adversaries, what is the seller's responsibility? The piece frames this as a failure of the defence-industrial habit of accounting for capability diffusion — a discipline that governed hardware exports during the Cold War but has been largely abandoned in the current AI race context. Anthropic's own Mythos Preview demonstrating serious cyber-offensive capabilities makes this a live operational question, not a theoretical one.

The War on the Rocks dual-use technology discussion War on the Rocks reinforces the point: AI's applicability to cyberattack, autonomous weapons, and biosecurity threats is not a future risk but a present-tense policy challenge. The structural problem is that the US commercial AI sector and the defence establishment are operating under fundamentally different incentive structures — one optimised for market capture and one theoretically optimised for strategic security — and current policy architecture does not resolve the tension. Huawei's chip progress, if it materialises, widens the window during which adversaries can access compute equivalent to current-generation US systems before controls tighten.