Last week, semiconductor stocks like Nvidia (NVDA +2.13%), Advanced Micro Devices (AMD -10.43%), and Micron Technology (MU -0.21%) saw significant declines due to the revelation that a Chinese startup, DeepSeek, has developed a cost-effective method for training artificial intelligence (AI) models.

Initially, these drops reflected investor concerns that DeepSeek’s technology might create a sudden drop in demand for graphics processors (GPUs) and other data center components crucial for AI development. However, these fears may prove to be exaggerated.

DeepSeek’s Breakthrough

DeepSeek, a subsidiary of the successful Chinese hedge fund High-Flyer, has made waves by announcing its V3 large language model (LLM). This model matches the performance of OpenAI’s leading GPT-4 series at a fraction of the cost.

Notably, DeepSeek managed to train V3 with just $5.6 million, compared to the $20 billion that OpenAI has spent since 2015. This significant disparity raises questions about the future of AI training costs and the need for expensive hardware.

DeepSeek faces restrictions as the U.S. government bans recent Nvidia GPUs from being sold to Chinese firms. Despite this obstacle, the company used older GPU generations—H100 and H800—to achieve its breakthrough. This outcome suggests that advanced AI models can be developed without access to the latest technology.

To compensate for the lack of computational power, DeepSeek optimized its training methods. By employing more efficient algorithms and data input strategies, along with model distillation, DeepSeek successfully accelerated the training process significantly.

The Impact on Semiconductor Stocks

DeepSeek’s achievement has raised alarms among semiconductor investors. Their primary concern is whether other AI firms will adopt DeepSeek’s low-cost training methods, potentially reducing the demand for GPUs from Nvidia and AMD. Furthermore, less demand for GPUs could also affect Micron’s market for data center memory solutions.

Image source: Getty Images.

Meta’s Position

Hidden in the turbulence, Meta Platforms’ stance on AI chip investments offers some reassurance. The tech giant, a major Nvidia and AMD buyer, revealed its budget for chips and data center infrastructure in 2024, totaling $39.2 billion, and plans to spend up to $65 billion in 2025.

Mark Zuckerberg, Meta’s CEO, addressed concerns about DeepSeek during a conference call, asserting that it’s too early to predict the impact on capital investments in chips and data centers. He emphasized that even if training demands decrease, companies will still require high-performance computing resources for inference tasks.

Zuckerberg highlighted the shift from training to inference, a critical process where AI models generate responses based on user inputs. This move towards “reasoning” or test-time scaling, which involves methods like model distillation, requires more inference compute power. Thus, Zuckerberg believes that the demand for robust data center infrastructure will remain significant.

Investor Perspective

For investors, these insights provide a more balanced view of the impact of DeepSeek’s breakthrough. While cost-efficiency in AI training is a positive development, Meta’s perspective suggests that the semiconductor industry will continue to thrive.

Nvidia’s financial performance further supports this optimism. Despite the recent stock Price drop, Nvidia’s revenue in fiscal year 2025 is expected to exceed $128.6 billion, with approximately 88% of it coming from data center sales, largely driven by GPU sales. This robust revenue stream suggests that semiconductor companies like Nvidia remain integral to the future of AI.

AMD and Micron also play vital roles. AMD has emerged as a formidable competitor with its upcoming MI350 GPU, set to rival Nvidia’s latest models. Meanwhile, Micron remains a leader in high-bandwidth memory, critical for modern AI applications.

Conclusion

While DeepSeek’s breakthrough challenges the current landscape of AI development, it does not imply the decline of semiconductor companies. Meta’s optimistic stance on AI investments and the ongoing advancements in the field suggest that demand for GPUs and related components will persist.

Investors can look forward to continued growth in this sector, driven by the need for sophisticated AI infrastructure. The semiconductor industry appears well-equipped to adapt and innovate, ensuring its relevance and importance in the AI revolution.

To stay updated and informed about these developments, subscribe to our newsletter for exclusive insights and analysis. Share your thoughts on how DeepSeek’s breakthrough might affect the industry, and join the conversation below.

Feel free to comment, subscribe, or share this article on social media to spark further discussions and stay connected with our community.