Bank of America reaffirmed its positive stance on Micron, maintaining a “Buy” rating and a $1,550 target price despite a recent drop in the stock. The bank’s analysts described the selloff as an opportunity for investors, citing ongoing strength in demand from cloud and artificial intelligence infrastructure.
BofA stated that end-user demand, especially from hyperscaler and AI-related sectors, remains firm. The firm anticipates a normalization of memory pricing beginning in 2027 and into 2028, but maintains that Micron’s fundamentals are robust.
The analysts pointed out several drivers supporting their forecast, including earnings potential that could stay well above prior cycle peaks, an increased adoption of long-term supply agreements, sustained utilization of both GPUs and high-bandwidth memory, and consistent hyperscaler investment despite increasing component costs.
According to the bank’s estimates, average selling prices of DRAM and NAND could decline by 10% and 18% year-over-year, respectively, in 2028. Even with these decreases, Micron’s earnings per share (EPS) are projected to reach approximately $150. In a more pessimistic scenario, where DRAM and NAND pricing each fall by 30% and 40%, EPS could still reach close to $100. This remains significantly above a peak EPS of roughly $12 in 2018.
If memory prices remain stable through 2028, the bank sees the potential for EPS to reach $175. The adoption of long-term agreements, which Micron and Samsung suggest could eventually cover 50% to 70% of capacity, may create greater supply and demand stability, reducing the chance of sharp price swings. However, the analyst noted that such agreements would not fully shield the sector from cyclical downturns.
BofA’s report also highlighted that CXMT, a leading Chinese DRAM producer, targets consumer and commodity memory rather than the advanced memory products used in AI, resulting in limited competitive pressure in this segment.
The ongoing strength in GPU rental rates, including for Nvidia’s A100, H100, and H200 models during July and August, was seen as further evidence of elevated AI computing demand. Memory integrated within these GPUs, particularly high-bandwidth memory used for AI inference tasks, continues to attract significant interest.
The analyst also commented on the latest quarterly updates from the four largest hyperscalers, noting that although higher component costs were discussed, only AWS specifically cited memory-related expenses as a factor in adjusting its capital expenditure expectations. No major cloud platform indicated that memory supply was limiting AI deployment capacity.
