When the upstart Chinese firm DeepSeek revealed its latest AI model in January, Silicon Valley was impressed. The engineers had used fewer chips, and less money, than most in the industry thought possible. Wall Street panicked and tech stocks dropped. Washington worried that it was losing ground in a vital strategic sector. Beijing and its supporters concurred: “DeepSeek has shaken the myth of the invincibility of U.S. high technology,” one nationalist commentator, Hu Xijin, crowed on Chinese social media.
Then, however, OpenAI, which operates ChatGPT, revealed that it was investigating DeepSeek for having allegedly trained its chatbot using ChatGPT. China’s Silicon Valley–slayer may have mooched off Silicon Valley after all.
Which DeepSeek is the real DeepSeek? The plucky innovator or the unethical swindler? The answer is both.
Chinese companies have proved to be skillful inventors, capable of competing with the world’s best, including Apple and Tesla. And they have also proved adept at copying and stealing technology they don’t have, then turning it against the rivals that created it. Making a product on the cheap is much easier when you don’t have to invest in developing it from scratch.
“The old narrative was that China cannot innovate but can only copy,” Gregory Allen, the director of the Wadhwani AI Center at the Center for Strategic and International Studies, told me. “Now China can both innovate and copy, and benefit from both.”
DeepSeek’s model has genuinely inventive elements, some of which Silicon Valley engineers will surely study for features to adopt. The Chinese company has wrung new efficiencies and lower costs from available technologies—something China has done in other fields. Jimmy Goodrich, a senior adviser to the Rand Corporation who specializes in Chinese technology, told me that the same country that has excelled at reducing manufacturing costs and improving factory operations, in other words, has in this case figured out “how to crunch a lot of data at faster speeds with a smaller number of computers.”
But then DeepSeek may have gone a step further, engaging in a process known as “distillation.” In essence, the firm allegedly bombarded ChatGPT with questions, tracked the answers, and used those results to train its own models. When asked “What model are you?” DeepSeek’s recently released chatbot at first answered “ChatGPT” (but it no longer seems to share that highly suspicious response). What DeepSeek is accused of doing is nothing like hacking, but it’s still a violation of OpenAI’s terms of service. And if DeepSeek did indeed do this, it helped the firm to create a competitive AI model at a much lower cost than OpenAI. DeepSeek did not respond to a request for comment.
The implications of what DeepSeek has done could ripple through the industry. What’s the point of investing tens of millions in an AI model if a competitor (Chinese or otherwise) can simply rip it off?
But the story of DeepSeek also reveals just how much Chinese technological development continues to depend on the United States. DeepSeek used chips from the U.S. giant Nvidia to create its model, and, as it turns out, may have also tapped American data to train it. Washington can capitalize on that advantage to choke off Chinese tech firms.
Denying China the fruits of the most cutting-edge American research has been at the core of U.S. competitive strategy. Beginning in late 2022, for instance, the Biden administration effectively barred U.S. companies from selling Chinese firms the most advanced chips. DeepSeek acquired its chips before the controls kicked in. (U.S. officials are also investigating whether the company may have bought banned chips through third parties in Singapore.) In an interview last year, DeepSeek’s founder, Liang Wenfeng, admitted that “the problem we face has never been money, but the embargo on high-end chips.” The firm limited new users last week because, it said, of the threat of hacking—but the system also may not have the capacity to handle a deluge of curious customers.
China’s government and chip industry are racing to replace barred U.S. semiconductors with homegrown alternatives, but the technology is complicated and they are struggling to catch up. That’s why China’s leader, Xi Jinping, personally pressed President Joe Biden for relief from the controls. Now DeepSeek’s success may frighten Washington into tightening restrictions even further. Members of Congress have already called for an expansion of the chip ban to encompass a wider range of technologies.
But as much as the story of DeepSeek exposes the dependence of Chinese technology on American advances, it also suggests that stopping the transnational flow of technological goods and know-how may take more than export restrictions. DeepSeek’s engineers found ways to overcome Washington’s efforts to stymie them and showed that they could and would do more with less, compensating for scarcity with creativity—and by any means necessary.
The implication for the United States, Weifeng Zhong, a senior adviser at the America First Policy Institute, told me, is that “you really have to run much faster, because blocking may not always work to prevent China from catching up.” That could mean securing semiconductor supply chains, cultivating talent through education, and wooing foreign experts through targeted immigration programs. Because the tech war is, at its heart, a talent contest, Washington might even consider awarding green cards to Chinese engineers who graduate from U.S. universities, so as to get them working for Silicon Valley companies rather than DeepSeek.
President Donald Trump may be heading in a different direction. He has sharply criticized the CHIPS Act, passed in 2022, which provides government financial support for strengthening the semiconductor industry in the United States, and instead favors slapping tariffs on chips from Taiwan. He has threatened to close the Department of Education. And a recent spat between Tesla’s founder, Elon Musk, and MAGA loyalists over visas for foreign specialists showed that elements of the Republican coalition are too opposed to immigrants to attract the talent that Silicon Valley requires. On this one, Trump took Musk’s side in favor of the visa program.
Whatever the United States chooses to do with its talent and technology, DeepSeek has shown that Chinese entrepreneurs and engineers are ready to compete by any and all means, including invention, evasion, and emulation. Maybe DeepSeek is the great Chinese tech disrupter it has been touted to be. Or maybe that will be the next big Chinese tech company, or the next one.
The post The Real Lesson of DeepSeek appeared first on The Atlantic.
