The invisible fabric of AI: chips aren't a US-China war, but a 30-country chain (zoopa.es)

A recent analysis highlights the intricate global supply chain that underpins AI chip manufacturing, debunking the oversimplified narrative framing the competition as a US-China dichotomy. Every AI chip from companies like OpenAI and Nvidia relies on critical components sourced from over 30 countries, with dependencies on specialized producers of silicon wafers, photoresist, and packaging technologies, primarily found in Japan and Taiwan. This nuanced understanding reveals that any disruptions in this interconnected web could significantly impact AI development and economic stability. The analysis also identifies key bottlenecks affecting AI capabilities, including the limited availability of packaging (specifically TSMC's CoWoS techniques) and high bandwidth memory (HBM), dominated by firms like SK Hynix and Micron. Additionally, Nvidia’s CUDA platform retains a commanding position in AI software, making it challenging for alternatives to gain traction. With China holding substantial control over rare earth elements and the rising energy demands of AI infrastructures, the implications for future AI growth are profound. Recognizing the collective responsibility of international stakeholders, especially Europe, is crucial for fostering a balanced and influential role in the evolution of artificial intelligence in the coming decade.