AI's Memory Hunger Grows: Munster Identifies 4 Key Beneficiaries Ahead of 2027 Hardware Upgrade Wave

Gene Munster, Managing Partner at Deepwater Management, stated that as generative AI tools become increasingly sophisticated, demand for memory capacity is rapidly rising. He expects both enterprise and consumer markets to enter a new hardware upgrade cycle starting in 2027. Munster believes hardware and storage equipment manufacturers such as Apple (AAPL-US), HP (HPQ-US), SanDisk (SNDK-US), and Western Digital (WDC-US) will be key beneficiaries of the AI wave.

Munster explained that this assessment stems from the team's recent practical testing of AI development tools. Doug Clinton, Deepwater's internal AI research lead, recently conducted extended development tests using Grok Build (a command-line version of Claude Code and OpenAI Codex). While he gave positive feedback on overall functionality, he also found that AI programs consumed excessive memory, even causing laptops to shut down due to high load.

Munster described such command-line AI tools as "memory hogs," indicating that current hardware configurations of commercially available laptops are already struggling to smoothly run large AI models and local AI workflows.

He emphasized that as enterprise and consumer AI models continue to evolve, the memory capacity of existing devices will increasingly become a system bottleneck. If hardware specifications do not advance in parallel, it could not only lead to system instability but also fail to support future, more demanding AI computing needs.

"The trend is very clear," said Munster, predicting that starting in 2027, a new wave of AI-driven hardware refresh cycles will begin in both enterprise and consumer markets, with increased unified memory (Unified Memory) capacity being one of the main upgrade directions. He forecasts that as businesses and consumers realize their current laptops cannot handle AI applications, market demand will naturally shift toward next-generation high-memory-capacity devices, ushering the PC, storage, and memory industries into a new growth phase.

In recent years, platforms such as Apple Intelligence, Microsoft Windows AI PC, Qualcomm, AMD, and Intel have actively promoted the development of AI PCs. OEMs are also progressively increasing the minimum memory configurations for AI PCs from 16GB to 24GB and even 32GB to meet the computational demands of on-device large language models and AI agents. Market research firms IDC and Counterpoint Research also project that AI PC penetration will rise rapidly over the next few years, becoming a key driver for PC market replacements.

Meanwhile, the continuous expansion of AI infrastructure is simultaneously driving up demand for high-bandwidth memory (HBM) and DRAM. Companies including NVIDIA (NVDA-US), Micron (MU-US), SK Hynix, and Samsung Electronics are continuously benefiting from growing AI server demand. Multiple Wall Street institutions expect tight memory supply and demand conditions to persist for several years, supporting the profitability of related industries.

Munster also summarized the recent stock performance of the four companies he favors:

Company: Apple 1-month performance: 9.59% Year-to-date performance: 20.13% 6-month performance: 54.65%

Company: HP 1-month performance: 3.06% Year-to-date performance: 8.71% 6-month performance: -3.35%

Company: SanDisk 1-month performance: -36.33% Year-to-date performance: 485.96% 6-month performance: 3196.87%

Company: Western Digital 1-month performance: -34.68% Year-to-date performance: 182.94% 6-month performance: 616.79%