Chinese startup DeepSeek has sent shock waves through the artificial intelligence world and created a headache for the United States.
Its AI assistant overtook Western rival ChatGPT on January 27 to become the top-rated free app on Apple‘s App Store in the U.S., delivering a trillion-dollar blow to U.S. tech firms on the stock market.
The OpenAI rival sent a sobering message to both Washington and Silicon Valley, showcasing China’s erosion of the U.S. lead in a critical tech battleground. Newsweek contacted DeepSeek, OpenAI and the U.S.’s Bureau of Industry and Security via email for comment.
Moor Studio/Getty
DeepSeek, a little-known company in Hangzhou, China, is the brainchild of Liang Wenfeng, a former hedge fund manager who transitioned to AI development in 2023. Liang told Asia Tech Review: “This is about democratizing innovation.”
His platform’s flagship model, DeepSeek-R1, sparked the largest single-day loss in stock market history, wiping billions off the valuations of U.S. tech giants like OpenAI minority owner Microsoft and chipmakers Nvidia and Broadcom last month. The brutal selloff stemmed from concerns that DeepSeek, and thus China, had caught up with American firms at the forefront of generative AI—at a fraction of the cost.
DeepSeek claims that it trained its models in two months for $5.6 million and using fewer chips than typical AI models. U.S.-based OpenAI was reported to have spent around $100 million to develop GPT-4. Gary Marcus, a cognitive scientist and an author of Rebooting.AI, told Newsweek: “Nobody has landed on the moon yet, or will they soon, but China has basically caught up to the U.S. in the flawed and faddish techniques of generative AI.”
While DeepSeek’s budget claim has been disputed by some in the AI world, who generally argue that it used existing technology and open source code, others disagree.
Nabil Jawdat Sarhan, associate professor of Electrical & Computer Engineering at Wayne State University and director of Wayne State Computer Systems and Deep Learning Research Lab, told Newsweek: “I have no reason to doubt the credibility of this information. China has invested heavily in nurturing talent in artificial intelligence, computer science, computer engineering and other fields.
“As a researcher in AI, I’m astonished by the massive volume of Chinese publications in top research journals and conferences in the field. Moreover, Chinese companies have been successful in making competitive products at much lower costs than in the U.S. The technology gap between the USA and China has substantially narrowed over the years, and China has started to gain an edge in some areas.”
VCG/VCG/AP
‘Significant Impact’
Kristian Hammond, a professor of computer science at Northwestern University, where he directs the Center for Advancing Safety of Machine Intelligence, told Newsweek: “If they are telling the truth and the system can be built on and run on much less expensive hardware, DeepSeek will have a significant impact.
“Nvidia currently is so valuable because everyone thinks that we cannot build AI without them. But DeepSeek demonstrates that we can. Similarly, OpenAI has ‘argued’ that we cannot support their software locally, but DeepSeek, again, demonstrates that we can.
“And the fact that DeepSeek could be built for less money, less computation and less time and can be run locally on less expensive machines, argues that as everyone was racing towards bigger and bigger, we missed the opportunity to build smarter and smaller. If they open the door to their techniques, it will be a disruptor as everyone will rush towards small.”
Former President Joe Biden‘s restrictions on foreign AI development may have pushed China to innovate. To keep a steady lead in the AI industry, Biden’s 2022 CHIPS Act increased semiconductor manufacturing in the U.S., while also clamping down on chip firms investing in China and Russia, the biggest competitors in the industry.
He later granted $6.6 billion to Taiwan Semiconductor Manufacturing Company for construction of advanced chip factories in Arizona, reducing the risk of supply chain disruptions in the event a conflict over Taiwan. However, DeepSeek had stockpiled 10,000 of Nvidia’s H100 chips and used this to continue work, though export controls remain a challenge, according to Liang.
Greg Baker/AFP via Getty Images
Alexandra Mousavizadeh, creator of the Global AI Index and CEO of Evident, told Newsweek that Biden’s policies may have forced China to innovate with fewer resources. “Seeing DeepSeek, [Biden’s regulations] don’t seem to have done a huge amount…. I think first of all we’re definitely seeing that they’re not that successful in containing chips getting into the Chinese markets.”
Even if the U.S. clamps down harder, “it’s going to be almost impossible to completely ring-fence China out of high-end chips,” she said. “The other dynamic is that even if it was successful in containing…you’re pushing China to develop pretty sophisticated chips.”
This highlights how DeepSeek was able to launch with just 2,000 specialized Nvidia chips compared with the 16,000 or more required to train leading U.S. models; Chinese companies worked with what they had.
“What China just needs to do is develop chips that are good enough,” Mousavizadeh added. “They don’t have to be state of the art to get to something that’s highly competitive.”
Chatbot Rivalry
DeepSeek’s success in the U.S. has been attributed to its accessibility, cost-effectiveness and performance. DeepSeek described itself as “rivaling OpenAI’s model o1” in reasoning and mathematical problem-solving. OpenAI CEO Sam Altman called DeepSeek “an impressive model, particularly around what they’re able to deliver for the price.”
He added on X: “We will obviously deliver much better models and also it’s legit invigorating to have a new competitor! We will pull up some releases. But mostly we are excited to continue to execute on our research road map and believe more compute is more important now than ever before to succeed at our mission.”
Getty Images
Stuart Russell, a distinguished professor of computer science and holder of the Smith-Zadeh chair in engineering at the University of California, Berkeley, told Newsweek it’s “probably too early to say” if DeepSeek is as good as the latest systems from U.S. firms.
“Chatbot performance is a complex topic,” he said. “If the claims hold up, this would be another example of Chinese developers managing to roughly replicate U.S. systems a few months after their release. The general outlines of how OpenAI’s o1 works have been known for quite a while—even before it was released—so it’s not all that surprising that it can be roughly replicated.”
Russell said the numbers for the training cost and chips used were the biggest surprise and “have caused grief in the markets, because U.S. companies such as Microsoft, Meta and OpenAI are making huge investments in chips and data centers on the assumption that they will be needed for training and operating these new kinds of systems. If that assumption is false and it can be done much more cheaply, then those investments are mostly a waste of money, and future demand for Nvidia’s chips in particular will be much lower than predicted,” he said.
A Nvidia spokesperson told Newsweek: “DeepSeek is an excellent AI advancement and a perfect example of test-time scaling. DeepSeek’s work illustrates how new models can be created using that technique, leveraging widely available models and compute that is fully export control compliant. Inference requires significant numbers of Nvidia GPUs and high-performance networking. We now have three scaling laws: pre-training and post-training, which continue, and new test-time scaling.”
VCG/VCG via AP
AI’s Future in the U.S.
Fellow U.S.-based firms Microsoft, Meta, Amazon and Alphabet are expected to spend up to $300 billion on AI projects this year. On January 21, Donald Trump announced Stargate—a $500 billion investment in U.S. infrastructure tied to artificial intelligence by a new partnership formed by OpenAI, Oracle and SoftBank. The U.S. president said: “This is a resounding vote of confidence in America’s potential. It will ensure the future of AI technology and keep it in this country.”
However, DeepSeek’s success could undermine a key pillar of foreign policy—restricting the sale of U.S.-designed AI semiconductors to China. Mousavizadeh outlined Trump’s options. “You can continue to try and contain access to chips and close the walls off. While you’re doing that, you’re doubling down on investment into data infrastructure, supporting the development of AI in the U.S. and being first in that race,” she said.
“Or you open up completely and you say, ‘Look, it’s to the benefit of all that everyone has access to everything, because the collaboration between Europe, the U.S. and China in the past has been what has led to the ability to get to where we are today.’ So closing off will probably slow down overall global development, in my view. My understanding is that he [Trump] definitely would lean towards the former.
“He’s trying to close off access to chips in China and go full throttle on AI development, which is done with Stargate. He’s got it very much top of the agenda in this White House.”
Whoever wins the AI race, Russell has a warning for the industry. He said the race for larger data centers “had already started to look like the race between the U.S. and the Soviet Union in the 1960s to build and test larger and larger bombs: they got as far as 50 megatons before realizing it was all rather pointless, but in the process they wasted billions of dollars and dumped enough radioactivity into the atmosphere to kill 100,000 people.
“The ‘[AI] race’ between companies and between nations is somewhat similar, except worse: even the CEOs who are engaging in the race have stated that whoever wins has a significant probability of causing human extinction in the process, because we have no idea how to control systems more intelligent than ourselves.”
In other words, Russell, said, it is “a race towards the edge of a cliff.”
fairness meter
Click On Meter To Rate This Article
