Morgan Stanley's AI warning

Morgan Stanley’s thematic note, led by Stephen Byrd, models a cumulative 44 GW shortfall in U.S. power for AI data-center demand through 2025–28 before “time‑to‑power” mitigations, while alternative scenarios in the same work show smaller net gaps in the single‑digit to tens‑of‑GW range. investing.com The bank’s labor economics projection estimates roughly $920 billion in annual net benefits for S&P 500 firms from full AI adoption and flags that AI could touch about 90% of occupations, framing job‑mix risk as a measurable financial lever. axios.com An executive‑update slide that maps Morgan Stanley’s scenarios into three columns—(A) “time‑to‑power” mitigated gap (~13 GW), (B) median buildout stress (9–18 GW range), and (C) cumulative shortfall tail (≈44 GW)—lets boards see probability‑weighted capex and schedule impacts tied to each scenario. investing.com Operational leadership reviews should surface a heatmap with PV of risk: current data‑center share of U.S. electricity (~4% in 2024), projected GW demand, estimated permitting/transmission lead times (up to a decade), and location concentration to prioritize latency and outage risk. pewresearch.org Translate workforce exposure into role‑level numbers using Morgan Stanley’s “90% of occupations touched” metric—e.g., a 1,000‑FTE org would have ~900 roles flagged for evaluation—then pair that headcount exposure with scenario timelines (months to years) and projected productivity/savings referenced by the $920B estimate. msn.com Frame resource asks in MW and dollars: JLL’s 2026 data‑center outlook gives an industry benchmark near $11.3M per MW of commissioned IT load, so a 100‑MW incremental request maps to roughly $1.13B of build cost and should be tied to the scenario column and the grid‑upgrade lead times Morgan Stanley flags. jll.com