The artificial intelligence boom is creating an unexpected bottleneck: memory.
While investors have spent much of the past several years focusing on GPUs, data centers and AI software, the explosive growth of AI models is also driving enormous demand for DRAM, high-bandwidth memory (HBM) and NAND flash storage.
That has put memory manufacturers and suppliers at the center of an increasingly tight supply chain.
In a recent discussion, market commentators argued that Wall Street may be underestimating the severity of the memory shortage. They highlighted three companies they believe could benefit from continued AI infrastructure spending: Micron Technology (MU), SanDisk (SNDK) and Silicon Motion (SIMO).
The three companies occupy different positions within the memory and storage ecosystem, ranging from producing memory itself to storing AI-generated data and managing the flow of information through flash storage.
Why AI Is Creating a Memory Bottleneck
The argument starts with DRAM, the short-term working memory used by computers, smartphones and data-center systems.
DRAM is also a critical building block for HBM, a high-performance type of memory increasingly used alongside GPUs for AI workloads.
HBM stacks multiple DRAM chips and connects them to provide extremely fast data access. However, producing HBM consumes significant manufacturing capacity.
According to the discussion, roughly 30% of global DRAM capacity can be associated with HBM production while delivering a much smaller proportion of the overall DRAM output because of the complexity and intensity of manufacturing.
The result is what the speakers describe as a “bottleneck within a bottleneck.”
Building additional semiconductor manufacturing capacity can take years. Meanwhile, AI models are becoming larger and more computationally demanding, potentially increasing memory requirements before new factories are ready.
That dynamic is forming the foundation of the bullish thesis for memory stocks.
1. Micron Technology (MU)
Micron Technology is presented as the most direct way to gain exposure to the memory shortage.
Micron participates across multiple parts of the memory market, including DRAM, HBM and NAND flash.
DRAM provides short-term working memory, while HBM is designed to deliver extremely high bandwidth for demanding applications such as AI. NAND flash, meanwhile, provides persistent storage that retains information even after a device is powered off.
That puts Micron in an important position as AI data-center infrastructure expands.
The speakers highlighted a sharp decline in Micron’s share price from its recent highs and pointed to a forward price-to-earnings ratio of roughly six times based on the figures discussed in the interview.
They also cited analyst earnings expectations that showed substantial projected EPS growth over the coming years.
According to the discussion, estimated earnings per share were approximately $73 for 2026, rising to around $156 in 2027 and $171 in 2028.
The consensus price target cited during the discussion was approximately $1,575.
The fundamental argument is straightforward: if AI companies continue expanding data-center capacity and demand for advanced memory continues rising faster than supply, memory manufacturers could gain significant pricing power.
Micron therefore represents the memory manufacturing side of the AI opportunity.
2. SanDisk (SNDK)
The second company highlighted is SanDisk, which represents the storage side of the AI infrastructure equation.
If Micron supplies memory used for computing, SanDisk’s products are positioned to benefit from the enormous amount of data being generated and stored by AI applications.
Every AI interaction, image, video, dataset and model requires storage somewhere.
SanDisk specializes in NAND flash-based storage, the technology behind many solid-state drives.
Unlike DRAM, NAND retains information when the power is switched off, making it suitable for persistent data storage.
The speakers argued that the rapid expansion of AI is creating a structural need for more storage capacity as companies accumulate increasingly large datasets.
SanDisk was cited as trading at approximately seven times forward earnings based on the figures presented in the discussion.
The speakers also highlighted a PEG ratio of around 0.42 and cited a consensus price target of approximately $2,263.
They suggested that earnings estimates could continue rising if the memory shortage persists and AI infrastructure spending remains strong.
SanDisk therefore offers a somewhat different exposure from Micron: rather than primarily focusing on the production of working memory, the investment thesis centers on the long-term storage requirements created by AI.
3. Silicon Motion (SIMO)
The third name is significantly smaller than Micron and SanDisk: Silicon Motion.
With an approximately $8 billion market capitalization according to the discussion, Silicon Motion is a more specialized company operating within the NAND ecosystem.
Its role can be thought of as the controller layer for flash storage.
Silicon Motion manufactures flash controllers that manage how information moves into and out of NAND memory.
The speakers compared the company to a traffic controller, directing data to the appropriate location and managing memory operations.
That function becomes increasingly important as flash storage grows faster and more sophisticated.
The company’s recent financial performance was another major part of the bullish argument.
The discussion cited second-quarter revenue of approximately $451 million, compared with an analyst estimate of $403 million.
Even more notable was the company’s third-quarter guidance, which was cited at approximately $530 million against a Street expectation of roughly $430 million.
The speakers also highlighted projected EPS of approximately $11.19 for 2026, more than $15 for 2027 and nearly $20 for 2028.
At a forward P/E of approximately 16.6 times and a PEG ratio of roughly 0.31 based on the figures presented, Silicon Motion was characterized as comparatively inexpensive relative to its projected earnings growth.
The consensus price target cited in the discussion was approximately $369.
Because Silicon Motion is considerably smaller than the other two companies, however, its shares can also experience greater volatility.