Demand for high-bandwidth memory is tightening supply across the broader memory industry, driving higher pricing and strengthening earnings outlooks for leading manufacturers. While much of the market focus remains on AI chipmakers, the underlying bottleneck is increasingly rooted in memory capacity—especially HBM, which sits alongside AI processors and supplies the data that AI systems consume.
The imbalance is expected to persist for years as new manufacturing capacity takes time to build and HBM production requires substantially more factory space than conventional memory, leaving less capacity for other memory products.
Key takeaways
- Price move: Conventional memory prices rose 55% to 60% in a recent quarter, and some DRAM prices have more than doubled year over year.
- Catalyst: Exploding AI-linked demand for HBM is consuming capacity and spilling into shortages across the wider memory market.
- Who benefits: Manufacturers with leading positions in HBM, including SK Hynix, Micron Technology and Samsung Electronics, have pricing power.
- Implication for investors: The memory cycle remains highly volatile, and today’s scarcity could eventually reverse into oversupply.
- Macro angle: Tight semiconductor supply conditions may influence downstream device costs and constrain shipment growth in areas such as smartphones and PCs.
What drove the move
HBM is the critical memory type being pulled into the center of the AI supply chain. Data shows that each successive generation of AI hardware demands more memory bandwidth, and the manufacturing process for HBM is particularly capacity intensive.
According to the article, making a gigabyte of HBM consumes roughly three times as much factory space as producing an equivalent amount of ordinary memory. As a result, when major manufacturers prioritize HBM due to its high margins, they can indirectly starve supply for other memory categories, broadening the pricing impact beyond just HBM.
Data cited in the article also points to a multi-year imbalance. Analysts expect memory prices to keep rising into 2028, with at least one memory producer warning that the shortage could extend beyond 2030. The core issue is timing: new fabs take years to construct, and the capacity coming online is expected to barely keep pace with demand.
The shortage is spreading through the memory market
What starts as an HBM problem has expanded into the entire memory complex. The article states that conventional memory prices jumped 55% to 60% within a single quarter, while some DRAM prices have more than doubled compared with a year ago.
The pricing pressure is also showing up in company-level economics. The report said memory revenue is on track to reach roughly $200 billion this year, about a quarter of all semiconductor sales, underscoring how large the memory segment has become within the chip industry.
Beyond servers and AI accelerators, the memory shortage is expected to ripple into consumer technology. The article noted that memory has become a larger share of smartphone bill-of-materials costs, and it expects rising memory prices to weigh on both phone and PC shipment levels this year.
Who stands to gain—and why the cycle matters
In a shortage environment, sellers typically set pricing terms more effectively, and manufacturers can secure multiyear contracts at higher prices. According to the article, SK Hynix leads in HBM supply and is described as providing memory to major AI customers. Micron Technology is highlighted as having sold advanced memory well into the future, while Samsung Electronics is included as the third major participant in the dominant HBM group.
On the storage side, the article also ties the strength in the memory complex to flash-memory producers such as SanDisk, suggesting that broader memory pricing dynamics can lift parts of the NAND supply chain as well.
Still, the report emphasizes that memory remains among the most cyclical areas of semiconductors. Historically, shortages have ultimately resolved when new capacity arrives and prices fall sharply. Investors are therefore cautioned against treating this cycle as permanently different, especially after memory names have already surged and tend to trade with high volatility.
Bigger picture for AI investors
The AI-driven memory shortage offers a longer-duration tailwind, but it is not without risk. The article argues that investors who want direct exposure may prefer to own leaders with supply advantages in HBM. It also points to the option of holding a diversified basket through the Roundhill Memory ETF, framing it as a way to gain exposure across the sector while spreading company-specific risk.
For investors, the key is to match position sizing to the sector’s cyclicality. Even if demand stays strong, memory pricing can change quickly when capacity additions catch up or product mix shifts.
Looking ahead, investors will likely focus on production ramp timelines for memory fabs, the pace of capacity additions versus AI-related demand growth, and any guidance from major manufacturers on how long tightness in HBM and DRAM could last. With device demand and semiconductor supply chains still adjusting to higher memory costs, upcoming earnings updates from the industry’s top suppliers and continued tracking of memory pricing trends will be central to gauging whether the shortage persists into the next phase of the AI hardware cycle.
