The Washington Post recently reported on China’s newest AI model from DeepSeek, addressing the claims surrounding its apparent humble budget and resources, and citing comments from a DSET research fellow.
The unveiling of DeepSeek’s new chatbot rivals ChatGPT, claiming comparative performance produced at a cheaper cost. In addition, DeepSeek made its code open source, allowing smaller competitors to use it and potentially further disrupt the market that China and the U.S. are competing to dominate.
Dr. Kai-Shen Huang, head of the Democratic Governance Research Program at DSET, was quoted in the article, saying that DeepSeek is a strong competitor in the AI market through both price and performance. “I think DeepSeek’s pricing strategy has the potential to disrupt the market globally,” he states.
Among the various claims about DeepSeek following its surprising rise in the AI industry, the company says that it spent only $5.6 million to train its AI model, significantly less than what U.S. companies have been spending to train its models. This claim challenged the prevailing notion that major financial investment is needed for AI development, causing the U.S. stock market to dip drastically and questions to arise about the future competitiveness of the AI industry.
In actuality, DeepSeek’s parent company, High-Flyer, has built a supercomputer which had cost $139 million to produce. Furthermore, additional investigation finds that DeepSeek spent nearly $1.3 billion in total, factoring the costs it took to amass the necessary materials and the research and development. The supercomputer gave DeepSeek the boost that it needed to succeed with the development of its AI assistant, signalling that the original claim of $5.6 million does not accurately factor in the total cost that went into achieving these results.
Additionally, skepticism have been raised over whether DeepSeek distilled models from OpenAI or found ways to circumvent the restrictions placed on Nvidia and its ability to sell more advanced chips to Chinese companies.This concern of looking for plausible explanations to attribute DeepSeek’s success reflects an underlying anxiety about how the AI market could change on a global scale.
Other AI experts shared their opinions, acknowledging how the rise of DeepSeek marks a notable advancement in the competition of the AI industry. While the U.S. has attempted to block China from purchasing the necessary computer chips in an effort to stall its AI development, in 2021 High-Flyer had purchased a bulk of the chips before the ban in 2022. This move proved significant in ensuring DeepSeek’s success.
Additionally, researchers contested the notion that this AI model came from seemingly out of nowhere. With access to High-Flyer’s supercomputer and teams of dedicated AI researchers, DeepSeek was able to benefit from the resources that were already long in production.
While some of DeepSeek’s claims about scarce resources and budget appear to be overstated, clearly its advancement marks a critical shift in AI development. While their budget may not have been as little as $5.6 million, it still stands that DeepSeek was up against strict regulations that restricted access to more advanced computer chips. Competing companies are challenged to reassess development strategies when competitors are able to innovate in different ways despite obstacles in resources.
Read more: https://www.washingtonpost.com/technology/2025/01/31/deepseek-ai-china-us-nvidia/