Nine U.S. trade groups — spanning automotive, medical devices, retail, telecom, and other manufacturing and consumer sectors — sent a joint letter Wednesday to the Treasury and Commerce secretaries, warning that AI data centers are consuming a "disproportionate" share of memory chip capacity, driving up prices and squeezing supplies for other industries. The letter is blunt: this is not a temporary supply chain hiccup but a structural reallocation of capacity.
HBM Eats Into DRAM Production, Squeezing Standard Memory
Memory chips were long treated as commoditized components: standardized specs, ample supply, stable pricing. That assumption broke with the rise of AI chip designers like Nvidia and AMD. They require HBM (High Bandwidth Memory), a specialized structure that vertically stacks multiple memory dies and places them close to the AI processor, delivering vastly faster data transfer — essential hardware for training large language models.
To produce more profitable HBM, the Big Three memory makers — Samsung, SK Hynix, and Micron — have reassigned wafer capacity originally used for standard DRAM. DRAM (Dynamic Random Access Memory) is the chip that temporarily stores data in laptops, phones, car computers, and medical imaging devices. As HBM consumes more wafers, the share for standard DRAM shrinks. Market estimates suggest HBM already accounts for about 25% of total DRAM wafer capacity, growing at nearly 70% annually.
The result: AI giants get the parts they need; everyone else waits in line.
Price Hikes Ripple Through Telecom, Auto, Medical — Consumers Pay the Tab
According to Bloomberg, the letter catalogs a wider-than-expected impact. Consumer electronics prices are the most direct effect. Costs for network and telecom infrastructure are also rising. Manufacturing goods such as vehicles and medical devices face potential supply interruptions. Even federal contractors, especially small and medium-sized ones, are struggling to meet delivery deadlines.
This is not a sporadic shortage; it is a structural reallocation of production capacity.
The groups ask the government to: work with memory makers and downstream buyers to expand capacity in the U.S. and friendly nations; leverage trade agreements or the CHIPS Act framework to secure supply chains; and explicitly ensure that policy coverage includes consumer and manufacturing markets, not just AI-related clients.
The CHIPS Act is a 2022 U.S. law that uses subsidies to bring semiconductor manufacturing back to the States (the same law used to pressure TSMC to build fabs in America). Industry groups want to borrow that subsidy and government procurement mechanism to tilt policy force toward downstream consumer and manufacturing sectors, not just AI computing power.
Memory Giants Rake It In, But New Capacity Is Years Away
The shortage is a windfall for the Big Three. Micron and SK Hynix both crossed the trillion-dollar market cap mark last month; their shares have surged more than 240% year-to-date, a boom rare in decades.
But capacity cannot be turned on overnight. Micron is building new plants in New York and Idaho, but Bloomberg notes most of that capacity will not come online for years. U.S. Commerce Secretary Howard Lutnick is pressing Samsung and SK Hynix to set up factories on American soil. Samsung is building a logic chip fab in Austin, Texas; SK Hynix plans a packaging facility in Indiana — but neither is a memory wafer fab.
Currently, only Micron produces memory wafers inside the U.S.; Samsung and SK Hynix keep their core memory wafer capacity firmly in Korea. Even if the onshoring push succeeds, real supply relief is years away. In the gap, automakers will keep bidding against AI giants for parts, medical device makers will keep raising prices, and consumers will keep paying the bill.
Memory scarcity is no longer a semiconductor cycle issue — it is a political choice that prioritizes AI. And the cost of that choice is being systematically passed on to those not at the decision table.