ChainThink message, June 2: According to investment bank research reports, multiple institutions have released AI infrastructure-related research reports over the weekend, decomposing the expansion of AI compute power into a "two-stage bottleneck" ranking: storage/memory > SSD controller > data center power/liquid cooling > GPU/cloud platform > MLCC > AI security.

The demand for AI data center compute power has extended comprehensively from the GPU layer to backend areas such as storage, power supply, and thermal management. Supply-demand imbalances have prompted institutions to significantly upgrade ratings and target prices for related assets, a trend closely aligned with the recent surge in memory giants MU and SK Hynix entering the trillion-dollar market cap club and strong performance in after-hours tech stocks.

Specifically, Susquehanna and GF Securities assigned Positive/Buy ratings to SNDK (SanDisk) with target prices of $3,250 and $1,277 respectively, citing severe supply-demand imbalance in NAND and enterprise-grade SSDs, along with continuously upward revisions in cloud providers’ AI storage requirements.

Raymond James raised MU (Micron)’s target price from $530 to $1,100 on June 1, maintaining its Outperform rating, driven by robust AI memory demand, rising HBM/DRAM prices, and capacity expansion—MU is now approaching the $1,000 mark and has broken through the trillion-dollar market cap threshold.

Other rating adjustments include: BofA upgraded SIMO (Silicon Motion)’s target price from $320 to $450 while maintaining its Buy rating;

Goldman Sachs and multiple Seeking Alpha reports maintained Buy ratings on NVDA (Nvidia) with a target price of $285; Truist raised AMZN (Amazon)’s target price from $310 to $320, assigning a Buy rating;

Jefferies and Morgan Stanley assigned Buy ratings to VST (Vistra), while UBS and BofA assigned Buy ratings to ECL (Ecolab), with target price ranges between $325 and $345.