Five architects of the AI economy explain where the wheels are coming off (techcrunch.com)

At the recent Milken Institute Global Conference, five key figures in the AI supply chain discussed pressing issues facing the industry, including chip shortages and the limitations of current AI architectures. Christophe Fouquet, CEO of ASML, emphasized that despite rapid advancements in chip manufacturing, there is likely to be a supply limitation for the next few years. Francis deSouza, COO of Google Cloud, highlighted the skyrocketing demand for AI infrastructure, with a backlog of nearly $460 billion in committed revenue. Qasar Younis pointed out that the greatest constraint isn't silicon, but the real-world data needed for training autonomy systems, illustrating the complex challenges AI developers face. The panel also explored innovative solutions to energy constraints. DeSouza mentioned Google's exploration of space-based data centers to access more abundant energy, while Eve Bodnia introduced energy-based models (EBMs) as an alternative approach that could foster a deeper understanding of data, differing significantly from traditional large language models. Dmitry Shevelenko discussed Perplexity's transformation into a "digital worker" that balances automation with human oversight, addressing the increasing need for security in AI-driven enterprise systems. The conversation underscored the intricate relationship between AI, data, and geopolitical concerns, suggesting that as physical AI technologies advance, they will raise new questions about safety, control, and national sovereignty.