Geopolitics & Sovereign Positioning

The US Department of Defense has formally declared its intent to operate as an 'AI-first' fighting force, signing eight new contracts with major technology companies as part of this reorientation. This is an enacted policy commitment, not aspirational language — contracts represent binding procurement instruments that will shape military AI integration over multi-year periods. The move signals that the Pentagon is accelerating beyond pilot programmes toward systematic embedding of AI in command, logistics, intelligence, and potentially lethal decision systems. The concurrent appointment of the former head of the Pentagon's think tank to Anthropic's leadership, described by the hire as a 'civilizational' adaptation challenge, underscores the deepening revolving door between national security establishments and frontier AI labs — a relationship that shapes both capability development and governance norms. BBC Defense One

For allied militaries, this doctrine shift creates both an interoperability imperative and a dependency risk: allies using US-origin AI systems gain capability but cede architectural autonomy. For China and Russia, it accelerates the timeline pressure to match AI-enabled military decision-making. The strategic stability implications are significant — AI-accelerated military cycles compress decision windows in crisis scenarios, increasing escalation risk in theatres such as Taiwan or the South China Sea.

Chatham House's assessment — published as the US House Foreign Affairs Committee advanced new export control legislation and the Chip Security Act continued its legislative path — argues that hardware controls alone are insufficient to constrain Chinese AI development. This is not new commentary, but the timing matters: it arrives alongside hard evidence of Chinese domestic semiconductor firms filling the gap. Cambricon Technologies, dubbed 'China's little Nvidia,' reported 160% revenue growth and 185% profit growth in Q1 2026, with its market capitalisation making it mainland China's most expensive listed stock. Peer firm MetaX reported comparable growth. Chatham House South China Morning Post

The geopolitical logic is becoming clearer: export controls have created a protected domestic market in China by eliminating Nvidia competition, channelling Chinese AI demand toward Cambricon, Huawei Ascend, and similar firms. The question for US policymakers is whether this acceleration of Chinese domestic capability represents a worse long-term outcome than the alternative. War on the Rocks analysis on economic statecraft notes that corporate risk calculations in adversarial environments require rethinking — firms that assumed compliance with export controls would secure their market position are now facing a China that is structurally incentivised to build around them. War on the Rocks

DeepSeek's addition of multimodal capabilities — image and video processing — to its V4 model marks a functional convergence with leading US frontier models including GPT-4o and Gemini. The Council on Foreign Relations frames this as signalling a new phase in the US-China AI rivalry: one in which China does not need to win the performance benchmarks race to reshape geopolitical outcomes. The strategic logic is that DeepSeek's cost efficiency, open-weight model releases, and rapid capability additions make it a globally competitive alternative for governments and enterprises that cannot afford or are restricted from accessing US-origin models. CFR South China Morning Post

For Global South governments weighing AI partnerships, DeepSeek's trajectory is directly relevant: a capable, low-cost, multimodal Chinese AI product creates a genuine alternative to US-aligned providers, complicating Washington's efforts to shape AI governance norms and technology standards globally. The simultaneous appearance of Alibaba, ByteDance, and Zhipu AI on Time's inaugural AI A-list alongside six US firms and France's Mistral underscores that the global AI landscape is no longer a US monoculture — and that the narrative of Chinese AI competitiveness is now entering mainstream Western framing. South China Morning Post

Samsung and SK Hynix — which together dominate global HBM and advanced DRAM supply — are reporting record-low order fulfilment rates and multi-year forward booking by customers anticipating prolonged shortages. This is an enacted supply constraint, not a forecast: customers are already pre-booking capacity for 2027 and beyond, suggesting the industry itself does not expect relief in the near term. Both firms have simultaneously disclosed increased investment in their China-based wafer fabs to meet AI demand, creating a tension between their role as critical suppliers to US AI infrastructure and their commercial exposure to the Chinese market. South China Morning Post

For US AI dominance strategy, South Korean memory supply is a strategic dependency that has received insufficient attention relative to logic chip geopolitics. The CHIPS Act addressed leading-edge logic fabrication but memory was largely left to market forces and allied relationships. A sustained HBM shortage constrains the $700 billion US hyperscaler capex programme — not because money is unavailable, but because physical supply cannot be manufactured fast enough. For China, the same shortage affects domestic AI scaling, but also creates pressure to develop indigenous HBM capacity, where firms like CXMT are making incremental but confirmed progress.