Samsung and SK Hynix, which together control roughly two-thirds of global DRAM production, warned that memory shortages will intensify through 2027 and persist into 2028. Samsung's memory chief told analysts that "significant shortages" across memory products are expected to continue through at least 2027, with conditions expected to worsen rather than improve. AI demand has consumed 70% of all memory chips produced globally.
Samsung's Q2 operating profit jumped more than 250-fold year-over-year to 89.2 trillion won ($61.7 billion), driven primarily by AI memory demand. The company has signed long-term supply agreements with the top five global data center firms and is close to deals with five more, covering 60-70% of capacity over multi-year terms with upfront payments and floor pricing. Samsung is shifting production capacity away from consumer memory toward high-bandwidth memory (HBM), which yields three to five times higher revenue per wafer than conventional DRAM.
The shortage has forced Apple, Samsung's rival, to raise MacBook, Mac, and iPad prices last month. NVIDIA is expected to raise consumer GPU prices by 20-30%. Samsung's own smartphone and TV divisions posted losses as higher memory component costs couldn't be fully passed to consumers. Memory fabrication capacity constraints and the fact that new fabs cost $15-20 billion and take years to build mean supply cannot respond quickly to demand, even with aggressive investment.
For architects, the extension to 2028 signals that HBM availability will remain the primary constraint on AI data center scaling, not GPU chip design or NVIDIA supply. The multi-year contracts between chipmakers and hyperscalers lock in pricing and prioritize committed customers, putting smaller operators and consumer electronics at the back of the queue—effectively creating a multi-tier supply economy tied to AI infrastructure capital commitments.