Can Policy Solve AI’s Chipflation?

Morgan Stanley's US Public Policy Strategist Ariana Salvatore examines 'chipflation' — the memory-chip cost surge driven by AI demand — and assesses whether policy tools can solve it. She concludes that policy can only help at the margin, because new fab capacity takes years, Chinese supply cannot close the high-bandwidth memory gap, and US export controls will remain restrictive. Investors should expect persistent tightness in AI strategic memory and only limited flexibility for commodity memory, keeping chipflation elevated near-term. - AI infrastructure is fueling surging memory costs, turning chips into an inflationary factor. - US policy tools (subsidies, tax credits, permitting) can support new fabs but face multi-year timelines. - China can add conventional memory supply but cannot solve the high-bandwidth memory bottleneck. - US export controls will stay restrictive, prioritizing supply-chain resilience over near-term price relief. - AI strategic memory (HBM/advanced DRAM) will see policy focus on allied trusted capacity. - Commodity memory may get targeted flexibility, but practical limits remain. - Net result: policy can mitigate, not end, chipflation near-term.