New Delhi: A new Cambridge report highlights that nearly one-third of the global AI supply chain lies in a “zone of invisibility” and receives no attention from national or international policymaking and regulation.
In the study, published by the AAAS, the team of researchers identified several “chokepoints” in the supply chain where nearly 137 elements are produced or put together in over 50 countries. The global narrative around AI often portrays the US and China as key players in the AI infrastructure, however, researchers found that other countries control materials and technology, in ways that perhaps the countries themselves do not understand yet.
“The US and China are central to the AI story, but this research shows us the extent to which critical technologies and materials across the AI stack have dependencies that extend far beyond their borders. It shows that no one country can control the AI supply chain alone,” Verity Harding, Director of the AI and Geopolitics project said in a statement.
One such country is Japan. The study found that nearly 95 per cent of the market share of a critical material needed to make AI chips is controlled by one Japanese firm. In the chip-making industry, a Dutch firm dominates EUV lithography, a manufacturing process used to print microscopic circuit patterns onto silicon wafers. This firm too relies on essential parts which are controlled by companies in Germany and the US.
“Most policy discussions reduce artificial intelligence to just a few inputs, such as compute, data and models. This is not an accurate understanding of the complexity of the AI supply chain,” lead author Siyi Liu, researcher at Cambridge’s Bennett School of Public Policy, said in a statement.
Also read: IIT Delhi leads Indian institutions in QS sustainability rankings, TISS ranked 59th nationally
‘Middle powers’ in the AI story
The study noted that countries such as Japan, Germany, and others referred to as “middle powers” — Belgium to South Korea” — could be controlling more of the AI production than they realise.
“Many governments are unaware of the strengths and vulnerabilities they hold at the heart of the global AI supply chain. Understanding this will be the basis of credible bargaining power in AI governance,” Liu added.
The researchers found that the UK is directly involved in producing 41 of the 137 supply-chain elements, but the country’s AI policies name only three, ignoring companies in the UK that focus on chip design and industrial gases. When they looked into 117 AI-related policy documents from nearly 43 countries and territories, they found that policies often focus on AI models, datacentres, funding and governance of AI. Factories, chemicals, chip design software, and even packaging technologies essential to the AI supply chain are completely overlooked.
When they broke down the supply chain into smaller elements they realised the importance of raw minerals, the ultra pure contaminant-free gases needed for semiconductors, and 351 “sub-elements” that are manufactured by nearly 800 firms across 56 countries.
One such “major blindspot” that researchers highlighted was advanced packaging, which connects several semiconductor chips into AI processors. This packaging is focused in countries such as Taiwan, Vietnam, and the Philippines.
“Policymakers focus on finished chips rather than the ecosystem behind them. They focus on GPUs but ignore the packaging operations needed to turn those chips into usable AI hardware,” said Liu.
