The day after Christmas, a small Chinese start-up called DeepSeek unveiled a new AI system that can match the capabilities of the latest chatbots from companies like OpenAI and Google.
That alone would be a milestone. But the team behind the system, called DeepSeek-V3, described an even bigger step. In a research paper explaining how they built the technology, DeepSeek engineers said they used only a fraction of the highly specialized computer chips that major AI companies rely on to train their systems.
These chips are at the center of a tense technology competition between the United States and China. As the US government works to maintain the country’s lead in the global AI race, it is trying to limit the number of powerful chips, such as those made by Silicon Valley firm Nvidia, that can be sold to China and other rivals.
But the performance of the DeepSeek model raises questions about the unintended consequences of the US government’s trade restrictions. The controls have forced researchers in China to get creative with a wide range of tools that are freely available online.
The DeepSeek chatbot answered questions, solved logic problems and wrote its own software as proficiently as anything already on the market, according to standard tests that US AI companies have used.
And it was built on the cheap, challenging the prevailing idea that only the tech industry’s biggest companies — all based in the United States — could afford to make the most advanced AI systems. Chinese engineers said they only needed about $6 million in raw computing power to build their new system. That’s about 10 times less than what tech giant Meta spent building its latest AI technology.
“The number of companies that have $6 million to spend is far greater than the number of companies that have $100 million or $1 billion to spend,” said Chris V. Nicholson, an investor with venture capital firm Page One. Ventures, which focuses on AI Technologies.
Since OpenAI sparked the AI boom in 2022 with the release of ChatGPT, many experts and investors had concluded that no company could compete with the market leaders without spending hundreds of millions of dollars on specialized chips.
The world’s leading AI companies train their chatbots using supercomputers that use up to 16,000 chips, if not more. DeepSeek engineers, on the other hand, said they only needed about 2,000 specialized computer chips from Nvidia.
Chip restrictions in China forced DeepSeek’s engineers to “train it more efficiently so it can still be competitive,” said Jeffrey Ding, an assistant professor at George Washington University who specializes in emerging technology and international relations.
Earlier this month, the Biden administration issued new rules aimed at preventing China from getting advanced AI chips through other countries. The rules build on multiple rounds of previous restrictions that prevented Chinese companies from being able to buy or manufacture advanced computer chips. President Trump has not yet indicated whether he will rescind the rules or rescind them.
The US government has tried to keep advanced chips out of the hands of Chinese companies because of concerns that they could be used for military purposes. In response, some firms in China have stockpiled thousands of chips, while others have sourced them from a thriving underground smugglers’ market.
DeepSeek is run by a quantitative stock trading firm called High Flyer. By 2021, it had channeled its profits into buying thousands of Nvidia chips, which it used to train its earlier models. The company, which did not respond to requests for comment, has become known in China for recruiting young talent from top universities with the promise of high salaries and the ability to pursue the research questions that most pique their interest.
Zihan Wang, a computer engineer who worked on an earlier model of DeepSeek, said the company also hires people without any computer science background to help the technology understand and be able to generate poems and questions on the notoriously difficult exam. of Chinese college admission.
DeepSeek does not make any products for consumers, leaving its engineers to focus entirely on research. That means its technology isn’t hemmed in by the strictest aspect of China’s AI regulations, which require consumer-facing technology to comply with government controls on information.
Major US companies continue to advance the state of the art in AI In December, OpenAI unveiled a new “reasoning” system called o3 that outperforms existing technologies, although it is not yet widely available outside the company. But DeepSeek continues to show that it’s not far behind. This month, she published her own impressive reasoning model.
(The New York Times has sued OpenAI and its partner, Microsoft, accusing them of copyright infringement of news content related to AI systems. OpenAI and Microsoft have denied the claims.)
An essential part of this rapidly changing global market is an old idea: open source software. Like many other companies, DeepSeek has open-sourced its latest AI system, meaning it has shared the underlying code with other businesses and researchers. This allows others to build and distribute their own products using the same technologies.
While employees at big Chinese tech companies are limited to collaborating with colleagues, “if you work with open source, you work with talent around the world,” said Yineng Zhang, principal software engineer at Baseten in San Francisco, who works on open source SGLang. the project. It helps other people and companies build products using DeepSeek’s system.
The open source ecosystem for AI gathered steam in 2023 when Meta freely shared an AI system called LLama. Many assumed that this community would only thrive if companies like Meta—tech giants with massive data centers filled with specialized chips—continued to open-source their technologies. But DeepSeek and others have shown that they can also extend the powers of open source technologies.”
Many executives and experts have argued that large American companies should not open source their technologies because they could be used to spread misinformation or cause other serious harm. Some US lawmakers have explored the possibility of preventing or curbing the practice.
But others argue that if regulators impede the progress of open source technology in the United States, China will gain a significant advantage. If the best open source technologies come from China, they argue, American developers will build their own systems on top of these technologies. In the long run, this could put China at the center of AI research and development.
“The center of gravity of the open source community has moved to China,” said Ion Stoica, a professor of computer science at the University of California, Berkeley. “This could be a big risk for the US because it allows China to accelerate the development of new technologies.
Hours after his inauguration, President Trump rescinded a Biden administration executive order that threatened to curb open source technologies.
Dr. Stoica and his students recently built an AI system called Sky-T1 that rivals the performance of the latest OpenAI system, called OpenAI o1, on several benchmark tests. They only needed $450 in computing power.
They did this by building on top of two open source technologies released by Chinese tech giant Alibaba.
Their $450 system isn’t as powerful as OpenAI’s technology or DeepSeek’s new system. And the techniques they used are unlikely to yield systems that outperform mainstream technologies. But the project showed that even operations with small resources can build competitive systems.
Reuven Cohen, a technology consultant in Toronto, has been using DeepSeek-V3 since late December. It says it’s comparable to the latest systems from OpenAI, Google and San Francisco startup Anthropic — and much cheaper to use.
“DeepSeek is a way for me to save money,” he said. “This is the kind of technology that someone like me wants to use.”