PwC expects US government efforts to regulate AI will focus on tangible, physical infrastructure like computer chips and energy production. This approach is more practical and easier to enforce than attempting to regulate software and models themselves, which are harder to track and control globally.

The primary constraint on AI development is not software or algorithms but the physical infrastructure required to support it: power, data centers, and supply chains. Policy will focus on this area regardless of election outcomes, though the specific approach may differ.

While the West obsesses over algorithmic superiority, the true AI battlefield is physical infrastructure. China's dominance in manufacturing data center components and its potential to compromise the power grid represent a more fundamental strategic threat than model capabilities.

Nvidia CEO Jensen Huang states AI growth is constrained by much more than just chips. The entire physical supply chain—including land, power, construction workers, photonics, and connectors—is a bottleneck. This indicates the next wave of investment and risk will focus on these fundamental, non-digital infrastructure components.

Software-based AI capability leaps matter, but the real structural inflection arrives when models integrate with physical robotics to perform manual labor and manufacturing. This physical frontier remains severely neglected in policy discussions despite impending industrial supply chain consequences.

The seemingly immaterial world of AI is entirely dependent on a vast physical system. Beyond electricity, AI's expansion drives demand for industrial commodities like copper and aluminum for grids, refined fuels for transport, and robust shipping infrastructure. This links digital growth directly to global commodities and logistics markets.

The abstract race for AI superiority is now grounded in physical reality. Control over electricity grids, cooling, and land for data centers has become as strategically important as semiconductor supply chains, shaping who can scale frontier AI.

A global pause on frontier AI development is presented as technologically feasible. Training superintelligence requires massive, city-scale data centers using advanced chips from a handful of suppliers. This creates a verifiable chokepoint for international monitoring and control, countering the "unstoppable progress" narrative.

A true, self-sustaining intelligence explosion requires more than AI automating its own software R&D. Ajeya Cotra emphasizes it must also automate the entire physical stack—from designing robots to fabricating chips and mining raw materials. This physical feedback loop is a critical, often overlooked bottleneck.