Here's why Bank of America is not worried about DeepSeek and AI semiconductors

Market Reaction to DeepSeek - DeepSeek's R1 model has gained attention for matching the performance of leading Western AI models like OpenAI's ChatGPT at a fraction of the cost, developed using older Nvidia H800 chips for less than $6 million [2] - The emergence of DeepSeek has raised doubts about the sustainability of the multibillion-dollar AI spending spree in the US, leading to a 10.14% drop in Nvidia's share price in pre-market trading and a 15% decline on January 27 to $120.97 [3] Bank of America's Perspective - Bank of America remains optimistic about Nvidia and the semiconductor industry, stating concerns over DeepSeek's impact are overstated [1] - The firm emphasizes that cost-efficient AI models like DeepSeek's R1 still rely on foundational large language models (LLMs) such as Meta's open-source Llama, which require significant infrastructure investment [4] - Meta's plan to increase 2025 capital expenditures by over 56% to $60–$65 billion highlights the ongoing need for advanced technology to support foundational AI models [5] Semiconductor Industry Outlook - Bank of America maintains a bullish outlook on key semiconductor players, reiterating 'Buy' ratings on Broadcom and Marvell Technology for their critical roles in supplying AI infrastructure [6] - Citigroup analyst Atif Malik reaffirmed a 'Buy' rating on Nvidia with a $175 price target, expressing skepticism about DeepSeek's cost advantage and noting the likely use of advanced GPUs in model development [7] Industry Trends - Compute demand is expected to continue growing, driven by foundational models, derivative models using knowledge distillation and sparse attention, and large-scale inference applications across diverse industries [5]