July 7, 2026 – Bank of America (BofA) released a research report stating that global capital expenditure on cloud computing and artificial intelligence infrastructure is expected to approach $1.5 trillion by 2027, growing at an annual rate of 40% to 50%.
The report, led by BofA analyst Vivek Arya and his team, points out that this growth is driven by multiple factors, including sustained token usage growth, rapid adoption of AI agents, and continued supply‑side infrastructure constraints.
Semiconductor Pullback Called a “Healthy Correction”
Addressing the recent pullback in semiconductor stocks, BofA explicitly characterized it as a normal and healthy correction, rather than a structural shift in AI demand.
The bank noted that the Philadelphia Semiconductor Index (SOX) surged 88% in the second quarter and then corrected about 11% in the third quarter, which aligns with its historically weakest seasonal pattern. Historical data suggests that consolidation periods are often followed by renewed momentum.
Memory Emerges as Core of AI Capex
BofA particularly emphasized that memory now accounts for 35%–40% of cloud AI capital expenditure—two to three times its historical share. Yet memory stocks remain undervalued, currently trading at only about 10x forward P/E.
The bank believes the market underestimates the industry’s transformation from a “cyclical commodity” to a “strategic core component of the AI era.” With the proliferation of long‑term supply agreements and more predictable pricing mechanisms, valuation multiples are expected to expand further.
BofA Names Beneficiary Sectors and Stocks
As visibility into 2027 cloud capex improves in the second half of 2026, BofA expects the following areas to regain market favor:
- Memory: Micron Technology – reiterated “Buy” with a price target of $1,550
- Compute chips: AMD, Intel
- Semiconductor equipment: Applied Materials, Lam Research, KLA, Teradyne
- Optics & networking: Macom Technology Solutions, Credo Technology, Marvell Technology
China’s Open‑Source AI Models Accelerate Adoption
BofA also noted that Chinese open‑source models such as Zhipu AI (GLM), Moonshot (Kimi), DeepSeek, and Alibaba (Qwen) have rapidly narrowed the gap with top U.S. labs while offering much lower inference costs. The bank views this as a positive catalyst for AI adoption—lower‑cost AI will expand usage scenarios, broaden deployment, and ultimately drive demand for compute, memory, networking, and power infrastructure.
BofA analysts stated bluntly: “The bigger risk lies in AI business models and software profit margins, not semiconductor demand itself. ” As the AI industry shifts focus from models themselves to AI agents, workflows, and enterprise automation, the core of future value creation will increasingly concentrate on the application layer and infrastructure.
For reference only, not for investment advice.