Geopolitics & Sovereign Positioning

OpenAI has paused negotiations for a significant UK data centre investment, citing prohibitive energy costs and regulatory uncertainty, according to BBC News. The project was central to British government claims that the UK could establish itself as an AI superpower through foreign tech investment. The pause underscores how nations competing for AI leadership face hard infrastructure constraints — energy pricing, grid capacity, and regulatory clarity — that cannot be overcome through policy aspiration alone.

This development highlights a strategic vulnerability for countries attempting to attract rather than build sovereign AI capacity. The UK's approach depended on creating conditions attractive enough for US companies to place compute infrastructure domestically, but OpenAI's calculus reveals that operational economics trump political commitments when margins are tight. Meanwhile, nations investing directly in state-backed infrastructure — or those with cheaper energy and lighter regulatory frameworks — gain relative advantage in hosting the physical layer of AI capability.

Alibaba's Qwen model family has captured over 50% of global open-source AI model downloads as of March 2026, reaching nearly 1 billion cumulative downloads and far surpassing Meta's Llama and other competitors, according to a report cited by South China Morning Post. Separately, Zhipu AI open-sourced its flagship GLM-5.1 model while simultaneously raising API prices by 10%, marking a strategic shift toward monetising advanced capabilities as SCMP reports.

This distribution dominance represents a strategic layer beneath headline frontier model competition. While US firms lead in closed, cutting-edge systems, Chinese companies are establishing themselves as the default infrastructure for developers globally who cannot afford proprietary API access or require local deployment. This creates dependency pathways that may prove more durable than temporary technological leads — developers building on Qwen gain familiarity with Chinese tooling, alignment approaches, and ecosystems. The simultaneous open-sourcing and price increases by Chinese firms suggest a maturing strategy: use open models to capture distribution and mind-share, while monetising premium tiers for revenue sustainability.

The ongoing Iran conflict has featured extensive attacks on AI infrastructure, representing the first war where artificial intelligence systems and their physical compute foundations have been primary military targets, according to analysis in Foreign Policy. The article argues this has exposed critical geopolitical miscalculations in the current technology race, revealing vulnerabilities in concentrated AI compute assets that were not adequately factored into strategic planning by Gulf states and their partners.

This development marks a threshold shift in how AI capability translates to strategic vulnerability. Nations with advanced AI infrastructure — data centres, training clusters, and inference systems — now face concentrated attack surfaces that adversaries can target to degrade military and economic capability. The conflict demonstrates that AI dominance creates new fragility: the physical assets enabling AI are large, expensive, energy-intensive, and difficult to harden or distribute. For countries racing to build sovereign AI capacity through massive data centre investments, this introduces a strategic dilemma between concentration for efficiency and distribution for resilience.

A new Atlantic Council report warns that the United States and European Union risk drifting further apart on AI policy despite AI's increasing centrality to national security, with the report calling for more structured cooperation mechanisms, according to Atlantic Council. The analysis identifies divergence on regulatory approaches, export controls, and military AI applications as creating friction that undermines collective Western positioning.

This assessment captures a fundamental tension in the geopolitical AI competition: the US and EU share strategic interests in maintaining democratic leadership in AI development and deployment, but pursue increasingly incompatible policy frameworks. The US prioritises speed, commercial leadership, and military integration, while the EU emphasises risk regulation, fundamental rights, and civilian oversight. Without structured coordination, this divergence creates exploitable gaps — for example, Chinese firms may access European compute or research partnerships foreclosed in the US, or US military AI applications may face European diplomatic opposition. The report's call for structured cooperation suggests current ad hoc mechanisms are insufficient.

A new analysis in Rest of World examines how Indian AI startups Sarvam AI and Krutrim are developing resource-constrained models that operate effectively despite limited compute and infrastructure, positioning India's approach as a potential blueprint for other developing nations seeking AI sovereignty without massive capital investment. The approach emphasises efficiency, local language capabilities, and deployment on modest hardware.

This frugal innovation model represents a third path in the geopolitical AI competition beyond US capital-intensive frontier racing and China's state-directed industrial scaling. For Global South nations, replicating either the American or Chinese approach requires resources most do not possess. India's emphasis on building capable-enough systems with limited resources offers a viable sovereignty strategy for countries that will never compete at the frontier but need functional AI systems aligned with local needs and languages. If this model proves successful, it could establish India as a technology provider and standard-setter for dozens of developing nations, shifting influence patterns independent of US-China competition.