US 100% Memory Tariff Risk Meets AI Demand Uncertainty at Micron

US considers 100% tariffs on non-domestic DRAM: Micron could be the only direct beneficiary

U.S. Commerce Secretary Howard Lutnick said memory players that do not produce domestically could face a 100% tariff, which frames a direct regulatory and cost shock for offshore DRAM supply. For Micron, that matters because its production footprint positioning could translate into pricing power and customer preference, while rivals without U.S. DRAM exposure face an existential cost disadvantage. [2]

Tight supply supports pricing, but capex leverage and AI compression may challenge the demand ceiling

Wall Street expects pricing to be supported through at least end-2026, with supply for calendar 2026 described as effectively sold out and limited capacity additions available, while Micron’s HBM4 positioning is viewed as high value. [1] At the same time, Micron’s growth plan is capital intensive, with Q1 FY2026 capex rising 68% year over year, and the market is already stress-testing whether demand sustains profitability under heavier spending. [3] Further, Google Research published TurboQuant results that reduce KV cache memory footprint by 6x to 8x with zero loss in model accuracy, which introduces a plausible structural question about whether memory intensity keeps compounding with AI model growth. [3]

If tariffs widen and compression spreads, Micron’s margin profile could bifurcate by product and timing

If U.S. tariffs move from a broad policy signal to enforceable scope that tightens supply outside the U.S., Micron could see sustained pricing leverage through 2026, aligning with analyst views of supply tightness and multiyear customer assurances. [1][2] However, if AI systems increasingly use more memory efficient approaches like TurboQuant, customers may demand less high-bandwidth memory per unit of inference growth, which could slow the pace at which Micron converts capex into volume and free cash flow. [3] The financial implication is a bifurcated risk profile: policy could support unit economics, but technology-driven memory intensity changes could delay utilization, raising near-term profitability risk even if long-run demand remains strong. [1][3]