Bank of America: Chip stocks "summer correction, autumn rebound"; memory "undervalued, should expand"

```

The latest report from Bank of America Securities believes that the current adjustment in the semiconductor industry is a healthy reset rather than a structural shift, and gives a clear endorsement of the long-term investment value of memory chips.

According to Wind Trading Desk, Bank of America analyst Vivek Arya pointed out in a July 6 report that after the Philadelphia Semiconductor Index surged 88% in the second quarter, it has corrected by 11% in the third quarter, which highly aligns with the historical pattern of the seasonally weakest periods. He characterizes this correction as a "summer reset" and anticipates a rebound in the fall. In the memory sector, the report reiterates a buy rating for Micron Technology with a target price of $1,550, naming it as the preferred pick.

These assessments are directly relevant to current market sentiment. Against the backdrop of the rapid rise of open-source models in China and doubts about the sustainability of AI capital spending, this report provides a systematic defense of the demand logic for the semiconductor industry and gives a clear prediction for the valuation recovery path of memory chips.

Seasonal correction does not change the trend; AI cycle foundation is solid

Bank of America believes that the 11% correction in the SOX index during the third quarter is consistent with historical seasonal patterns and should not be interpreted as a structural deterioration of AI demand. The report points out that after a consolidation period, momentum often regathers, driven by investors’ restored confidence in the next round of earnings growth and capital expenditure cycles.

Global cloud and AI infrastructure capital expenditure will approach $1.5 trillion by 2027, increasing another 40% to 50% from current levels. Supporting factors include the continued expansion of computing power demand, accelerated implementation of AI agents, and structural supply-side constraints.

The report emphasizes that the strategic focus of hyperscale cloud providers remains on maximizing utilization and business growth, rather than optimizing depreciation and amortization, meaning capital spending is highly rigid. As visibility of cloud capital spending in 2027 improves gradually in the second half of 2026, it is expected that storage (MU), computing (AMD, INTC), semiconductor equipment (AMAT, LRCX, KLAC, TER), optics (MTSI), and networking (CRDO, MRVL) sub-sectors will regain market leadership.

Chinese open-source models impact software profits, but benefit semiconductor demand

The report responds positively to concerns about Chinese AI models. Chinese open-weight models such as GLM, Kimi, DeepSeek, and Qwen have quickly narrowed the gap with leading US frontier labs and are competing with significantly lower inference costs. The latest third-party benchmark ranking as of July 4 shows that US models from Anthropic and OpenAI still lead, but Chinese models occupy 8 of the top 16 spots, with the highest being GLM 5.2 from Zhipu (Z.ai)—an open-weight model featuring 750 billion parameters and a million-context window.

The rise of Chinese models poses real pressure on AI software profit margins, but is actually positive for semiconductor demand. Low-cost intelligence will expand use cases and deployment scope, ultimately driving overall demand for computing power, storage, networking, and power infrastructure. The report makes clear that greater risk lies in model economics rather than the underlying semiconductor demand.

Additionally, the report notes NVIDIA is actively engaging in open-source community development, which not only improves its hardware ecosystem but also helps bring small and medium-scale AI adopters into its ecosystem—these customers often lack direct access to frontier labs.

Memory valuation is significantly underestimated, multiple factors support multiple expansion

The report's judgment on memory chips is its most assertive. The report states that memory currently accounts for about 35% to 40% of cloud AI capital expenditure, which is 2 to 3 times historical levels, but memory stocks trade at just around a 10x forward P/E, which is clearly undervalued.

The market has underestimated the industry's shift toward long-term agreements and more predictable pricing models. Investors' concerns about pricing sustainability, new supply, and customer concentration have long pressured valuations, but this is a misjudgment. As memory evolves from a cyclical commodity to strategic AI infrastructure, valuation multiples should be repaired.

Based on this logic, Bank of America reiterates a buy rating for Micron Technology, with a target price of $1,550, listing it as the top pick in the semiconductor sector.

~~~~~~~~~~~~~~~~~~~~~~~~

The above exciting content is from Wind Trading Desk.

For more detailed interpretation, including real-time insights and frontline research, please join the 【Wind Trading Desk ▪ Annual Membership】

Risk Warning and DisclaimerThe market involves risk and investment should be cautious. This article does not constitute personal investment advice and does not take into account the specific investment objectives, financial situation, or needs of individual users. Users should consider whether any opinions, views, or conclusions in this article are suitable for their particular situation. Investment based on this information is at your own risk. ```