Morgan Stanley's September 7 report suggests that OpenAI's GPT-6 Astra model represents a substantial leap in AI capability breadth and interoperability, shifting market focus from "how much infrastructure is needed to serve known AI demands" to "how many new workloads will become economically viable as model intelligence improves." The report points out that physical supply constraints such as AI hashrate, power, materials, and labor will become bottlenecks. Specifically, ABF substrates are expected to face shortages starting in 2027, HBM4E back-end processing complexity is increasing, and the U.S. faces a 38-gigawatt power deficit. The firm recommends investment priorities in AI hashrate, networking, memory, and analog chips.