DeepSeek threatens monopoly of tech giants in 'arms race' to leading AI, expert says

An expert pointed out that the Chinese company DeepSeek's development of lower-cost models using lower-quality systems in the field of artificial intelligence (AI) poses a threat to the monopoly of tech giants in the AI "arms race." This move by DeepSeek has sparked a rethinking of the landscape of AI technology development.

Following DeepSeek's release of its AI model named R1, Wall Street suffered its largest single-day loss in history, with a market value evaporation of $593 billion (approximately A$943 billion). The stock price of Nvidia, the American tech giant that provides chips for AI systems, fell by one-sixth, and the stock prices of other AI-related companies, such as Google and Microsoft, also declined. This event highlights DeepSeek's potential influence in the AI field.

According to DeepSeek, the cost of R1's API is $0.88 per million input tokens and $3.49 per million output tokens, while OpenAI's o1 API costs $23.92 and $95.70 respectively, meaning DeepSeek's costs are 20 to 50 times lower than its competitors. Furthermore, DeepSeek's spending on AI power is only $8.9 million. Geoff Webb, Professor and Research Director of the Department of Data Science and Artificial Intelligence at Monash University, stated that while developing core AI systems previously required billions of dollars in investment, DeepSeek's achievement indicates that core technology may be more accessible than previously thought.

Webb likened the fierce competition between tech giants to develop the most powerful AI to an "arms race" and believes DeepSeek's achievement is credible. He pointed out that all companies participating in this "arms race" keep the internal details, including the input used for training systems and the specific processes, strictly confidential. Nevertheless, he believes DeepSeek may have found a more economical way to achieve AI development and benefited from the previous experiences of American companies and the research community. DeepSeek CEO Liang Wenfeng is also the co-founder of the hedge fund and AI company High-Flyer, which had stockpiled 10,000 Nvidia A100 graphics processor chips, although the US later banned the supply of these chips to China.

DeepSeek stated that its latest model was built using Nvidia's lower-performing H800 chips, which are not banned in China. DeepSeek claims that despite the lower cost and the use of lower-quality chips, R1's capabilities are comparable to OpenAI's ChatGPT o1 model. Webb believes this poses a "very real threat" to Nvidia's profitability, noting that it means AI technology companies no longer have the exclusive control over technology that they previously thought they had, their monopoly is being lost, and costs are also significantly reduced. This event also raises questions about the investment and equipment required to develop AI models. Microsoft signed a 20-year agreement last year to restart the Three Mile Island nuclear power plant to meet the huge demand for electricity from its AI data centers, while OpenAI has raised $10.5 billion from investors. Meta recently also announced plans to invest up to $103 billion to expand its AI infrastructure.

Webb stated that tech giants will be looking at how to reduce their own costs. He believes that although DeepSeek has not yet reached the level of leading AI companies, it shows that the gap between them is not as large as previously thought. R1 has been hailed by tech investor Marc Andreessen as the industry's "Sputnik moment." Andreessen, who has advised US President Donald Trump on technology policy, has warned that excessive US regulation of AI will allow China to take the lead in the AI field. Trump told members of Congress yesterday that the release of DeepSeek's R1 should serve as a "wake-up call" for the industry. He emphasized that the US needs to focus on competing and winning because the US has the best scientists in the world.