Summary

Sean discussed Entext’s thesis that the AI cycle is shifting value from software towards hardware and the enabling industrial base, as physical supply becomes the binding constraint. He linked this to sustained hyperscaler capex and an “AI factory” framing where marginal cost of compute matters more. He highlighted power and supporting infrastructure as the core bottlenecks, citing rising data-centre energy density and long lead times across grid equipment, networking and advanced packaging. He argued this creates a mismatch between capital deployment and real-world build capacity, shifting pricing power towards parts of the Asian hardware supply chain while the US faces permitting and interconnection friction and China benefits from faster grid buildout and manufacturing scale. In the Q&A, discussion covered optical interconnects, the risk of vendor vertical integration, and a potential next leg of long-distance fibre investment as data centres link across regions and compute follows cheaper power. He also downplayed near-term disruption from emerging sponsor-backed approaches, before closing with a brief overview of Entext’s research offering and pricing.

Topics