How AI’s Global Supply Chain Is Reshaping Power

Summary

Artificial intelligence is presented not as a contest controlled by one country, but as a globally distributed system whose critical layers are concentrated in different places. The visible products, including ChatGPT and Claude, depend on chips, data, energy, models and applications that must work together. NVIDIA designs leading AI processors but relies on manufacturers such as TSMC, while TSMC depends on Dutch lithography equipment, Japanese chemicals and South Korean memory. This division of labour has made advanced AI both highly specialised and vulnerable to disruption: disputes between Japan and South Korea affected chip materials in 2019, while Russia’s invasion of Ukraine disrupted neon supplies in 2022. Data centres add further dependencies because training and running models requires processors, networking, cooling, land, water and large amounts of electricity. The article contrasts the United States’ strengths in frontier models, private investment and data centres with China’s advantages in electricity generation, robotics, research output and manufacturing, while Taiwan remains central to advanced-chip production. US export controls introduced in 2022 were intended to slow China’s AI development, but experts cited in the article say they also encouraged efficiency and domestic alternatives. China’s control of more than 90 percent of rare-earth refining adds another source of leverage. The resulting interdependence crosses political alliances: South Korea supplies semiconductor technology while remaining economically tied to China, and Singapore and Australia’s ambitions to become data hubs depend on foreign technology and sufficient energy. For Australia, the article argues that technological self-sufficiency is unrealistic. Its choices include building strengths in energy, critical minerals and sectors such as mining and agriculture, or becoming mainly a foreign-funded data-centre base. Experts warn that the latter could create an infrastructure enclave that uses domestic resources without delivering comparable intellectual property or strategic influence. The central policy question is therefore where Australia can become indispensable while managing the risks created by dependence, transparency gaps and the costs of securing the AI stack.