Citi analysts have refreshed the firm's AI compute semiconductor stock rankings in the wake of the latest quarterly earnings season, reiterating that demand for AI compute capacity continues to outstrip available supply.
The note from Citi highlights a recent 20% price increase for AWS EC2 GPU instances and identifies DRAM memory shortages as the key bottleneck limiting compute availability. The bank points to multiple strategic customer agreements signed through 2030 by Micron as evidence of tight memory allocation on the part of major customers.
Reflecting that supply dynamic, Citi said its updated methodology now assigns the greatest weight to memory supply allocation among the factors it evaluates. Other components considered include accelerator exposure, CPU positioning and the breadth of diversified hyperscale sales exposure.
In Citi's revised mega-cap ranking, Nvidia occupies the top slot, with the bank citing the company's strong partnerships around high-bandwidth memory - HBM - as a competitive advantage. Broadcom holds second place, while Micron is ranked third.
Among the large-cap cohort, AMD is ranked first. Citi notes AMD's memory relationships include Samsung, Micron and CBRS, and highlights that CBRS employs a different on-chip memory approach - on-chip SRAM - versus HBM.
Following AMD in the large-cap list are CBRS, Intel and Marvell Technology. Qualcomm appears at the bottom of Citi's large-cap ranking, despite a recent high bandwidth compute announcement; Citi explains that the Qualcomm design "uses a stack of LPDDR memory due to limited allocation until 2029."
The updated rankings and the emphasis on memory allocation reflect Citi's view that memory shortages are a pivotal constraint for AI compute expansion. By prioritizing supply allocation metrics, the bank aims to better capture which suppliers and vendors are likely to convert AI demand into deployable compute capacity.
Investors tracking AI compute exposure and hyperscale demand will likely watch memory allocations and multi-year supplier agreements closely, as these arrangements appear to be influencing ranking positions across both mega- and large-cap semiconductors.
Contextual note - Citi's note ties product pricing, supplier commitments and memory architecture choices to relative stock rankings, without proposing specific market outcomes beyond its updated internal weighting of supply factors.