AI and robotics upheaval: compute, supply chains, and major governance risks ahead

A long-form outlook argues that the “AI takeoff” is still early and that algorithmic progress may accelerate faster than markets, governments, and companies expect. It expects front‑line AI models to keep improving, aided by more autonomous research via agents and better use of tokens for experimentation. The piece also claims that “depth-learning engineering science” is about to arrive, helping scaling experiments convert single-GPU learnings into multi-GPU outcomes, potentially reducing the role of hard-to-verify techniques. On infrastructure, compute is forecast to remain the fiercest competition for years, followed by rapid “commoditization” of compute as more accelerators, fabs, and power capacity come online. The AI supply chain is expected to shift from lab-centric advantage toward wider, automated discovery—potentially supported by open-source—though compute costs and opportunity costs could still bottleneck academia and open communities. Robotics is presented as the next step: two rounds of “ChatGPT-like” breakthroughs are expected within a few years, but global scaling of humanoids likely takes until around 2030 or later due to manufacturing economics and real-world deployment constraints. Crucially for markets, the essay highlights governance and security risks: faster AI diffusion, possible zero-day vulnerabilities across cyber/bio/infrastructure, automation of military and law enforcement, and uncertainty over whether AI labs could be nationalized. It warns about destabilizing power dynamics even if traditional “MAD” assumptions may not hold automatically. For traders, the core signal is that AI-driven capex (chips, data centers, energy) and automation narratives may stay bid, but tail risks around regulation, security incidents, and governance shifts could raise volatility.