Wall Street's focus on artificial intelligence is broadening, with Morgan Stanley advising investors to look beyond the dominant chipmakers like Nvidia and toward memory and CPU stocks. In a note on April 20, the bank's analysts argued that the rise of increasingly autonomous "agentic AI" systems will reshape data-center infrastructure, significantly boosting demand for central processing units (CPUs) and high-bandwidth memory (HBM).
Morgan Stanley estimates that agentic AI could add between $32.5 billion and $60 billion to a data-center CPU market already valued above $100 billion by 2030. This shift indicates the AI narrative is evolving from pure model training toward more persistent inference and orchestration workloads, which require more coordination and general-purpose compute. The bank sees CPUs acting as the critical control layer for these complex, multi-step tasks.
This development is crucial because it widens the capital expenditure cycle from a single-chip story into a full-stack infrastructure investment theme. "If AI data centers need more CPU throughput and more memory bandwidth, the capital-spending cycle widens," the note stated. Morgan Stanley believes memory demand is set to rise sharply, creating sustained pricing power for supply-constrained parts of the ecosystem, a condition it expects to persist through at least 2027.
The bank has consistently highlighted memory as a key bottleneck. In a March 26 note, it argued that memory had become the critical constraint for AI and next-generation CPU builds, with the strength in memory being more durable than investors anticipated. Morgan Stanley's preferred plays on this theme are memory stocks, specifically naming Micron and SanDisk, which it maintains at Overweight ratings. While the bank still sees upside in AI-adjacent compute names like Nvidia, AMD, Intel, and Arm, its highest-conviction trade remains in the memory sector due to constrained supply and increasingly structural demand.
This thesis is backed by explosive performance and fundamentals from market leader Micron Technology (MU). Micron's stock has surged over 500% in the past 12 months, including a 41% gain in April 2026 alone. The company's fiscal Q2 2026 revenue skyrocketed to $23.86 billion, up from $8.05 billion a year earlier, with non-GAAP EPS of $12.20. For Q3 2026, management guided for roughly $33.5 billion in revenue and about $19.15 in non-GAAP EPS.
A key differentiator in this cycle is the unprecedented demand visibility. Micron has locked in price and volume agreements for its entire 2026 supply of HBM, including its next-generation HBM4, which is already in high-volume production for Nvidia's upcoming Vera Rubin platform. The company estimates the HBM total addressable market could grow from around $35 billion in 2025 to roughly $100 billion by 2028.
Industry forecasts support the bullish outlook. Gartner expects global semiconductor revenue to jump 64% in 2026 to $1.32 trillion, with memory revenue tripling to $633.3 billion. It forecasts DRAM prices to rise 125% and NAND prices to surge 234% this year. Analysts have raised Micron price targets to $600 or more, implying roughly 40% upside from recent levels around $448.
Risks remain, however, as memory is a historically cyclical industry. Gartner warned of "memflation," where rising memory prices could delay non-AI demand. Supply will eventually catch up, and the Nasdaq PHLX Semiconductor Index's 14-day winning streak—its best since 2002—signals a potentially overheated short-term rally.
