Last week, semiconductor stocks like NVIDIA, AMD, and Micron saw declines amid concerns over a Chinese company, DeepSeek, which claims to have developed a cost-effective method for training AI models. Investors feared this innovation could reduce demand for GPUs and data center components. However, these fears may be overstated, particularly after Meta Platforms’ CEO Mark Zuckerberg reassured investors during a call, emphasizing that the demand for chips might not decrease even if training workloads are reduced.
DeepSeek, established in 2023, reportedly uses efficient software algorithms and older generation GPUs due to U.S. export bans on advanced hardware, managing to match performance benchmarks set by U.S. models with significantly lower training costs. This raised concerns among investors that other AI firms could follow suit, potentially decreasing purchases from NVIDIA, AMD, and Micron.
NVIDIA, a leader in the AI GPU market, and AMD, a competitor set to release new chips, are both crucial to the ongoing AI revolution. Micron also plays a significant role by providing high-performance memory solutions for data centers.
Zuckerberg highlighted that even if AI model training is less demanding, it does not necessarily equate to reduced chip demand, as the focus could shift from training to inference—where AI processes and responds to data input, requiring substantial data center infrastructure. Despite the uncertainties surrounding DeepSeek’s impact, Zuckerberg’s comments suggested that the long-term outlook for NVIDIA, AMD, and Micron remains robust, calming investor fears.
Source link
