Micron’s HBM Push for AI Could Pressure Its More Profitable DRAM Business
High-bandwidth memory, better known as HBM, has become one of the most important components in the artificial intelligence hardware boom. Data centers, AI accelerators, and advanced GPUs rely on HBM because it delivers the speed and capacity needed for large-scale AI inference and training workloads. As demand for AI computing continues to grow, memory makers are racing to expand HBM production.
However, new market data suggests that Micron’s aggressive move into HBM may come with a costly trade-off. While HBM is a critical technology for the future of AI infrastructure, general-purpose DRAM is currently delivering stronger gross margins for the company.
According to recent research, Micron’s gross margin for general-purpose DRAM has climbed sharply. In 2025, margins reportedly increased from 44 percent to 50 percent. The biggest jump is expected in early 2026, when the figure could rise to around 80 percent. With data center memory demand still extremely strong, Micron’s DRAM margins may continue moving higher, potentially reaching 95 percent in 2027 and staying near 93 percent through much of 2028.
That creates an interesting dilemma for Micron. On one hand, HBM is essential for AI chips and future data center platforms. On the other hand, standard DRAM appears to offer better near-term profitability.
The challenge comes from how HBM is manufactured. Unlike conventional DRAM, HBM uses multiple DRAM dies stacked vertically, which increases complexity, raises production costs, and creates yield challenges. Because HBM consumes more DRAM die output from the same wafer input, expanding HBM production can reduce the amount of general-purpose DRAM available for shipment.
This means that every increase in HBM capacity may limit Micron’s ability to take full advantage of the highly profitable DRAM market. In simple terms, producing more HBM can eat into the supply of a product that currently generates better margins.
HBM margins are improving, but they are still not expected to match general-purpose DRAM in the near future. Estimates suggest that HBM gross margins could reach around 75 percent to 78 percent in 2027. While that is strong by most industry standards, it remains below the projected profitability of conventional DRAM during the same period.
Even so, Micron is unlikely to slow down its HBM ambitions. Competitors in the memory industry are moving quickly to develop next-generation HBM products, including HBM4, and Micron cannot afford to fall behind in a market that is becoming central to AI computing. HBM is expected to remain a key memory technology for advanced GPUs, AI servers, and high-performance computing systems.
Research also indicates that Micron’s HBM shipments are set to grow rapidly. Quarterly shipments are projected to rise from around 0.1 exabytes in early 2025 to approximately 0.43 exabytes by late 2027. That expansion shows Micron is building a stronger position in the AI memory supply chain, even if it temporarily pressures its DRAM business.
The company may also be thinking beyond short-term margins. Current DRAM profitability is being supported by unusually strong demand and tight supply conditions, especially from data center customers. Over time, memory prices are expected to stabilize. When that happens, having a larger HBM business could become strategically important.
Memory manufacturers are also reportedly holding tens of billions of dollars in advance customer payments, giving them significant leverage for now. But that advantage may begin to weaken around 2029 as market conditions shift more in favor of buyers. This gives Micron a limited window to expand HBM capacity, secure long-term customers, and strengthen its role in the AI hardware ecosystem.
For now, Micron faces a balancing act. General-purpose DRAM is delivering exceptional margins and could remain highly profitable over the next several years. At the same time, HBM is becoming indispensable for AI infrastructure and may define the future of high-end memory.
The company’s decision to continue investing in HBM suggests it is prioritizing long-term relevance in artificial intelligence and data center computing, even if that means sacrificing some near-term DRAM profitability. As AI demand grows, Micron’s success may depend on how well it manages this trade-off between today’s strongest earnings opportunity and tomorrow’s most important memory market.
