Grid and permits block AI data centers

National AI strategies can fail in practice because the real bottlenecks are local: timely electric‑grid hookups, planning approvals and energy rules (including Net Zero constraints on gas) determine whether large compute projects actually get built. The Nscale–Microsoft Loughton project shows how months or years of delay on connections and local approvals can erase commercial commitments and deter future investment. — If true broadly, this reframes AI industrial policy: governments must pair compute ambition with faster permitting and grid capacity planning or risk losing anchor investments and strategic leverage.