Chinese Open-Source AI Models Reinforce Memory Demand, Bank of America Says

The recent emergence of low-cost, open-source large language models (LLMs) from China reinforces, rather than threatens, the bullish outlook for memory chip demand, according to a client note from Bank of America. The firm reiterated its Buy rating on Micron, arguing that these AI models still require substantial hardware to operate effectively.

Pricing Reflects Business Models, Not Hardware Costs

Bank of America addressed concerns that aggressive Chinese API pricing, which it noted can be 5 to 350 times cheaper than Western models, could signal lower hardware needs. Analyst Vivek Arya wrote that this price difference reflects "business-model choices, not necessarily reflective of hardware costs."

While efficiency techniques can reduce the compute and GPU intensity of these models, the bank stated they "require the same or more memory as their model weights and active parameters increase." For example, the note highlighted that Moonshot's Kimi K3 model still needs approximately 1.4 TB of High-Bandwidth Memory (HBM) per serving instance, a requirement comparable to Western models when adjusted for parameter count.

'Open' Source Still Requires Hardware Investment

BofA clarified that the term "'Open' refers to weight distribution, not deployment cost." This means customers who adopt these open-source models must still purchase the necessary HBM, DRAM, and NAND memory to run them on their own hardware.

Ad

The firm attributed the pricing advantage of Chinese models to several factors, including:

• Architectural efficiency techniques
• Lower local costs for electricity, labor, and land
• Potential state-linked capital subsidies

Outlook for Micron

Bank of America sees limited competitive threat from China's domestic memory producer CXMT in the high-end AI market. The note stated that CXMT "primarily addresses the underserved consumer/commodity DRAM segment, not HBM3E/HBM4."

Based on its analysis, the bank maintained its positive outlook on Micron (MU), reiterating a $1,550 price target. The report also pointed to a potential future catalyst for the chipmaker: the expiration of CHIPS Act buyback restrictions around December 2026, which could potentially pave the way for $50 billion to $60 billion in annual share repurchases.