Expansion of the Ban on "Chinese AI" DeepCik Sparks Shockwaves in the Industry

 

The Impact of DeepCik on Global AI Paradigms and Semiconductor Industries

The ban on the low-cost, high-performance AI from China, known as DeepCik, is rapidly spreading across various industries, including major South Korean companies, government departments, and financial institutions. Despite these efforts to block its use, the shockwaves caused by DeepCik’s arrival are expected to reverberate through the industry for a long time.

As of February 8, 2025, major South Korean corporations such as Samsung Electronics, LG Electronics, and SK Hynix, have placed strict bans on using DeepCik’s AI technology. The LS Group has even issued a company-wide directive to proactively block the use of DeepCik products. This movement is not limited to South Korea, as government bodies, the IT sector, and the financial industry worldwide are increasingly limiting access to DeepCik’s services due to concerns about potential privacy violations. This growing backlash is not only seen in South Korea but also in countries like the United States, Japan, Australia, Italy, and Taiwan, where some DeepCik services have been blocked.

Despite the widespread pushback against DeepCik, the impact it has already made on the global semiconductor and power industries is undeniable. DeepCik’s emergence signals a possible shift in the AI model industry’s paradigm from "high-cost, high-performance" to "low-cost, high-performance" AI. The AI model developed by DeepCik, "R1," which was publicly introduced on January 20, has drawn significant attention due to its remarkably low development cost. DeepCik claimed that R1 was developed with only 5.6% of the cost of OpenAI’s GPT, amounting to approximately $5.57 million (around 80 billion KRW). The R1 training process utilized Nvidia’s low-specification AI accelerator chip, the H800, which is an older version of Nvidia's cutting-edge H100 chip.

The introduction of DeepCik has shaken the AI sector by demonstrating that high-performance AI can be achieved without the need for large investments in advanced technology or powerful GPUs. This revelation defies the established industry belief that greater investment in high-performance computing power (processing power) and data would naturally lead to better AI performance. Previously, large tech companies heavily invested in high-end GPUs due to the assumption that AI models would perform better with more computational resources and larger parameters.

However, DeepCik’s ability to deliver high-performance AI at a fraction of the cost has sent shockwaves through the global semiconductor market, driving down the stock prices of key players such as Nvidia, Samsung Electronics, and SK Hynix. Currently, SK Hynix provides cutting-edge high-bandwidth memory (HBM3E) for Nvidia, and concerns are rising that a potential reduction in demand for high-performance GPUs could lead to a decrease in demand for HBM as well. With DeepCik’s low-cost approach threatening to disrupt the AI semiconductor market, there are growing concerns over the potential decline of high-end AI chip demand.

The rise of this low-cost, high-performance AI model has raised doubts about the long-standing approach of large-scale investments in high-performance AI models, which are typically associated with high costs. The potential reduction in energy consumption that comes with DeepCik’s approach is also a key concern. DeepCik claims that R1 can reduce power consumption by 50-75% compared to GPT models, which could significantly impact future investments in AI infrastructure, including power generation and distribution networks.

Industry experts are closely monitoring the situation. A semiconductor industry source stated, “If the AI semiconductor market shifts toward low-cost, low-performance models, it will have an impact, but we need more information before making any definitive predictions.” Another source pointed out, “If it’s possible to implement AI models with fewer chips, then big tech companies’ investment strategies will likely shift from focusing on high investments to prioritizing efficient spending.”

The shift in the AI industry towards low-cost, high-performance AI models like DeepCik could profoundly reshape the way companies approach AI infrastructure and investments. While some are concerned about the decline of high-performance semiconductor markets, others see the potential for a more cost-efficient and energy-conscious approach to AI development. Regardless, the arrival of DeepCik is likely to continue influencing the global AI and semiconductor industries for the foreseeable future.