Geopolitics & Sovereign Positioning

Huawei launched the Atlas 350 AI accelerator card for inference workloads, powered by its latest Ascend 950PR chip, with claims of superior performance to Nvidia's H20 chip — the downgraded variant Nvidia designed specifically to comply with US export controls. Vice-president Ma Haixu positioned the card for enhanced computing power and storage in AI inferencing applications. South China Morning Post The announcement follows years of escalating US semiconductor restrictions intended to constrain China's access to advanced AI chips.

The strategic significance extends beyond the hardware itself. If Huawei's performance claims withstand independent verification, it demonstrates that export controls are accelerating rather than preventing China's development of competitive alternatives. The Ascend ecosystem — spanning training and inference chips, software frameworks, and cloud infrastructure — represents China's most credible effort to build a vertically integrated AI stack independent of US technology. The Atlas 350's focus on inference is particularly telling: inference workloads consume the majority of deployed AI computing, meaning competitive inference chips directly enable operational AI systems at scale, not just research prototypes.

African governments have invested over USD 2 billion in AI-powered surveillance technology supplied and financed primarily by Chinese firms and banks, according to new research. Rest of World The infrastructure includes facial recognition systems, smart city monitoring networks, and integrated public security platforms deployed across multiple countries. Chinese financing — often through policy banks or tied to Belt and Road Initiative agreements — has enabled purchases that would otherwise exceed the budgets of recipient governments.

This creates a network effect that extends well beyond individual contracts. Once a government adopts Chinese surveillance infrastructure, operational dependencies lock in: maintenance contracts, software updates, training, and data integration all flow through the original supplier. The systems generate massive datasets on African populations, raising questions about data sovereignty and potential access by Chinese state entities. For Beijing, this represents deployed AI capability that shapes governance models, creates technical dependencies, and positions Chinese firms as default providers for future expansions — all while Western competitors face higher political costs from civil society scrutiny of surveillance exports.

Nvidia CEO Jensen Huang described China as 'formidable' in robotics, citing the country's dominance in microelectronics, motors, rare earth materials, and magnets — all foundational components for physical AI systems. South China Morning Post Huang's comments, made during a Silicon Valley podcast, acknowledged that US robotics firms will remain dependent on Chinese supply chains despite American pioneering of the robotics market. The statement comes as Nvidia pivots toward 'physical AI' and seeks to restore access to the Chinese market following years of export restrictions.

The admission reveals a critical asymmetry in the US-China technology competition. While export controls target discrete high-end components like advanced chips, China controls integrated supply chains for the physical substrates of robotics and automation — advantages built through decades of industrial policy and unlikely to be replicated quickly. This creates a structural dependency: even US firms with superior AI software and chip design rely on Chinese manufacturing for embodied AI systems. For Beijing, this represents leverage that cannot be easily sanctioned away, and positions China to dominate physical AI deployment even if it lags in frontier model development.

ByteDance sold gaming studio Moonton to Savvy Games Group, owned by Saudi Arabia's Public Investment Fund, for over USD 6 billion. South China Morning Post Moonton, developer of Mobile Legends: Bang Bang, represents established technical capability in real-time multiplayer systems and user engagement — capabilities adjacent to AI development. ByteDance CEO Zhang Yunfan confirmed the sale in an internal memo. The transaction follows ByteDance's broader pivot toward AI, suggesting capital reallocation from mature gaming assets toward core AI research and infrastructure.

The deal illustrates converging strategies: Gulf states acquiring proven technical capabilities and talent rather than building from scratch, while Chinese firms divest non-core assets to fund AI competition. Saudi Arabia's sovereign wealth has emerged as a major buyer of AI infrastructure and capabilities, from data centres to partnerships with US firms — a strategy that positions the Gulf as a critical swing region in technology competition. For China, sales like this provide capital and reduce exposure to US secondary sanctions risk, while maintaining commercial relationships that complicate US efforts at comprehensive technology decoupling.