KATHMANDU: A frenzy surrounding an artificial intelligence chatbot developed by Chinese tech startup DeepSeek was disrupting stock markets on Monday, intensifying debates over economic and geopolitical competition between the U.S. and China in AI technology development.
DeepSeek’s AI assistant soared to the top of Apple’s iPhone store as the most downloaded free app, driven by curiosity about the ChatGPT competitor. U.S. tech industry observers expressed concern over the possibility that the Chinese startup had matched American companies in generative AI capabilities at a fraction of the cost.
If accurate, this challenges the massive investments U.S. tech companies claim are necessary for the data centers and advanced computer chips required to push AI advancements further.
Amid the hype, misconceptions about DeepSeek’s technological strides also created confusion.
“The models they built are fantastic, but they aren’t miracles either,” said Stacy Rasgon, a Bernstein analyst covering the semiconductor industry. Rasgon, along with other stock analysts, described Wall Street’s reaction as overblown.
“They’re not using any innovations that are unknown or secret,” Rasgon added. “These are things that everybody’s experimenting with.”
What is DeepSeek?
Founded in 2023 in Hangzhou, China, DeepSeek introduced its first AI large language model later the same year. The company’s CEO, Liang Wenfeng, previously co-founded High-Flyer, one of China’s top hedge funds specializing in AI-driven quantitative trading. By 2022, High-Flyer had reportedly acquired 10,000 of Nvidia’s high-performance A100 graphics processor chips to build and operate AI systems, according to a post on Chinese social media platform WeChat. However, the U.S. soon restricted sales of those chips to China.
DeepSeek has since claimed that its recent models were developed using Nvidia’s H800 chips, which are not restricted in China. This suggests that high-end hardware may not always be necessary for cutting-edge AI research.
DeepSeek gained attention within the AI industry last month with the release of a new AI model. The company claimed it was comparable to similar models from U.S. firms, including OpenAI, while being more cost-efficient in utilizing Nvidia chips to train the system on vast datasets. The chatbot became widely available after appearing on Apple and Google app stores earlier this year.
However, the real stir came after a research paper published last week detailed a new DeepSeek AI model, R1, showcasing advanced reasoning skills, such as rethinking approaches to mathematical problems. The model was reportedly much cheaper than OpenAI’s comparable o1 system.
“What their economics look like, I have no idea,” Rasgon said. “But I think the price points freaked people out.”
The ‘Sputnik’ Moment
The DeepSeek breakthrough has sparked a larger debate in the U.S. about how best to compete with China in AI development.
“DeepSeek R1 is AI’s Sputnik moment,” said venture capitalist Marc Andreessen on social platform X, alluding to the 1957 Soviet satellite launch that ignited the Cold War space race.
Andreessen, who has advised Trump on tech policy, cautioned that overregulating AI in the U.S. could hinder American companies and give China an advantage.
DeepSeek’s advancement has also put a spotlight on U.S. export controls. Some experts argue that the timing of the announcement serves political purposes.
“The technology innovation is real, but the timing is political,” said Gregory Allen, director of the Wadhwani AI Center at the Center for Strategic and International Studies. Allen compared DeepSeek’s announcement to Huawei’s release of a new phone amid U.S.-China diplomatic negotiations over export controls in 2023.
“China is trying to show that export controls are ineffective,” Allen said.
On Monday, Trump called DeepSeek’s progress “good because you don’t have to spend this much money.” Speaking to House Republicans in Miami, Trump described the news as a “wakeup call” for U.S. industries to focus on competition. Trump also signed an order last week directing his administration to identify and close loopholes in export controls.
DeepSeek’s cost-efficient progress could undermine the $500 billion AI investment commitments announced by OpenAI, Oracle, and SoftBank during Trump’s administration. Nvidia, whose stock dropped 17% on Monday, issued a statement praising DeepSeek’s work as “an excellent AI advancement” that utilized “widely available models and compute that complies with export controls.”
What Makes DeepSeek Different?
Unlike competitors such as OpenAI, DeepSeek’s models are open source, meaning key components are accessible for modification, though the company has not disclosed its training data sources.
The standout feature of DeepSeek’s R1 model is what Nvidia describes as a “perfect example of Test Time Scaling,” where the AI effectively shows its reasoning process and uses that to improve itself without additional external data.
“It’s just thinking out loud,” said Lennart Heim, a researcher at Rand Corp.
OpenAI’s o1 model and others from U.S.-based companies likely have similar capabilities, Heim said.
“But this is the first time a Chinese company has gotten this close within such a short time,” Heim noted. “It’s why people are paying attention.”
“I used to think OpenAI was uncatchable, the king of the hill. Turns out that’s not entirely true.”