DeepSeek’s Arrival: A New Player in the AI Game
A surge of interest in DeepSeek, an artificial intelligence chatbot developed by the Chinese tech startup, has sent ripples through the stock market, igniting discussions on US-China economic and geopolitical competition in AI development.
The AI assistant became the No. 1 downloaded free app on Apple’s iPhone store, driven by curiosity about the ChatGPT competitor.
This development is causing concern among some US tech industry observers. They worry that DeepSeek may have reached a similar level of generative AI capabilities to American companies but at a potentially much lower cost. This situation could call into question the large-scale investments planned by US tech companies in data centers and computer chips, essential for further AI progress.
US President Donald Trump hailed DeepSeek AI as a positive step, suggesting that if accurate, lower costs could yield the same results, and a “wake-up call for our industries that we need to be laser focused on competing to win.” He signed an order on his first day in office that said his administration would “identify and eliminate loopholes in existing export controls.”
Hype and misconceptions about DeepSeek’s technological advancements also fueled some confusion.
“The models they built are fantastic, but they aren’t miracles either,” commented Bernstein analyst Stacy Rasgon, who focuses on the semiconductor industry. He described Wall Street’s response as “overblown.” He further clarified that DeepSeek isn’t using any unprecedented innovations; rather, they are employing techniques that are widely experimented with.
What is DeepSeek?
DeepSeek was established in Hangzhou, China, in 2023. It launched its initial AI large language model later that year. Its CEO, Liang Wenfeng, previously co-founded High-Flyer, a top Chinese hedge fund that specializes in AI-driven quantitative trading. By 2022, the fund had amassed a considerable cluster of Nvidia’s high-performance A100 graphics processor chips. These chips, based in California, are used to build and manage AI systems. Afterward, the US restricted the sales of those chips to China.
DeepSeek has stated that its recent models were built with Nvidia’s H800 chips, which are not banned in China. This indicates that cutting-edge AI research may not necessarily require the most advanced hardware.
DeepSeek’s Recent Progress
DeepSeek began to garner increased attention last month after releasing an AI model. It claimed to be on par with similar models from US companies like OpenAI, the creator of ChatGPT. The company also said that the model was more cost-effective in utilizing expensive Nvidia chips for training on vast amounts of data. The chatbot became more broadly available when it appeared on the Apple and Google app stores early this year. A research paper released last week spotlighted another DeepSeek AI model, R1. This model demonstrated advanced “reasoning” skills, such as the ability to adjust its approach to a math problem. It was also significantly cheaper than an equivalent model sold by OpenAI, called o1.
“What their economics look like, I have no idea,” Rasgon said. “But I think the price points freaked people out.”
The ‘Sputnik’ Backdrop and Geopolitical Implications
The discussions around DeepSeek’s technical capabilities are accompanied by an existing debate within the US over how to best compete with China in AI.
“Deepseek R1 is AI’s Sputnik moment,” stated venture capitalist Marc Andreessen in a post on social platform X. He was referencing the 1957 satellite launch that marked the beginning of the Cold War space race. Andreessen, who has advised Trump on tech policy, expressed a warning. He believes overregulation of the AI industry by the US government will hinder American companies and give China an advantage.
However, the attention on DeepSeek also puts at risk a central theme on the US policy to restrict the selling of the American-designed AI semiconductors to China. Some experts in US-China relations don’t think that is an accident, and they see a political motivation behind DeepSeek’s release.
“The technology innovation is real, but the timing of the release is political in nature,” said Gregory Allen, director of the Wadhwani AI Center at the Center for Strategic and International Studies. Allen noted a comparison to Huawei, a Chinese company sanctioned by the US. Huawei released a new phone during discussions regarding Biden administration export controls in 2023.
“Trying to show that the export controls are futile or counterproductive is a really important goal of Chinese foreign policy right now,” Allen explained.
Nvidia’s stock dropped 17% Monday, but the company applauded DeepSeek’s work in a statement. Nvidia called it “an excellent AI advancement” that used “widely-available models and compute that is fully export control compliant.”
DeepSeek’s Differentiators
DeepSeek’s models are “open source,” unlike other competitors such as OpenAI. Key components are freely accessible and modifiable, although the company hasn’t revealed the data it used in training. DeepSeek’s R1 model has garnered much admiration, particularly with Nvidia’s “perfect example of Test Time Scaling.” This means the AI models show how they think and then apply that knowledge for further training instead of requiring new data sources.
“It’s just thinking out loud, basically,” said Lennart Heim, a researcher at Rand Corp.
Heim noted that OpenAI’s reasoning models do the same. He suggested that other US-based contenders, such as Anthropic and Google, might have similar capabilities that have not been released. “It’s the first time that we see a Chinese company being that close within a relatively short time period. I think that’s why a lot of people pay attention to it,” Heim said. “I used to believe OpenAI was the leader, the king of the hill, and that nobody could catch up. Turns out this is not completely the case.”
