January 31, 2025
SEOUL – What some are calling “AI’s Sputnik moment” slammed the US tech sector this week, as a small Chinese firm demonstrated a dramatic breakthrough that raised questions about the need for massive amounts of investment in chatbots, data centers and specialized AI chips.
On Monday, Nvidia, which rocketed to the lead in the global AI industry with its AI chips, saw its shares tumble by 17 percent, wiping $600 billion in its market value. This stock market rout that also hit other US tech firms came as investors suddenly realized that large language models, or LLMs, developed by a Chinese startup called DeepSeek can be as good as those of US tech giants like Google and OpenAI — for a fraction of the cost.
There is one immediate implication: DeepSeek’s generative AI innovation contradicts the long-held industry consensus that advanced AI systems require a huge amount of investment. US tech companies such as Microsoft, Google and Meta have so far been pouring money into their AI projects in a bid to make them more sophisticated and commercially successful. Suddenly, however, investors and experts alike begin to wonder whether such gargantuan investments are really necessary.
DeepSeek’s release has drawn keen interest among users, as well. Within days of its release, DeepSeek’s chatbot ranked as the most downloaded app on Apple’s AppStore, outsmarting OpenAI’s ChatGPT.
Of course, it is too early to say that DeepSeek has outpaced its bigger and better-funded US competitors in the intensifying global competition for AI solutions. But the Chinese company’s feat, building advanced AI models with fewer Nvidia chips and less money, means that China is fast closing the gap in the cutthroat race to build AI systems. This is all the more surprising in that the US government is fiercely imposing restrictions on the export of AI and chip technologies to China.
There are some disputes about how DeepSeek’s AI models match the level of AI products made by the US tech companies. DeepSeek has reportedly benefited from cutting-edge AI technology publicly shared by Meta and other solutions freely available on the internet. Meta’s decision to give away its AI technology two years ago was based on the notion that the software development method called “open source” is a far better way to spread and advance AI solutions, drawing resources and talents from around the world.
But some hold a skeptical view about the way DeepSeek has shocked Silicon Valley with its breakthrough AI technology. OpenAI, which released a global hit chatbot service ChatGPT, is now questioning whether DeepSeek may have inappropriately used data generated by OpenAI to build its own systems — a technique called “distillation.”
But the idea of distillation itself can be debatable since most AI companies have relied heavily on the open source software code and vast volumes of data freely available on the web. OpenAI itself has been sued for allegedly using copyrighted online data to train its AI systems.
Aside from the dispute, there is a big lesson for South Korea’s policymakers and tech companies. The country, which lags behind in AI solutions, may stumble into a fresh opportunity to catch up if — and it’s a big if — the government and conglomerates realize it’s not too late to invest in making inroads into the AI battlefield, as demonstrated by DeepSeek.
South Korea has global tech companies such as Samsung Electronics, LG Electronics and SK hynix. It is notable that SK hynix is a major suppler of high bandwidth memory chips to Nvidia, but the country still lacks local AI pioneers that can compete with global AI players on an equal footing.
DeepSeek’s groundbreaking success suggests that small startups can upend the global AI industry with innovative techniques at a lower cost. South Korea’s policymakers need to find ways to spur similar innovations in the AI field, where so much is at stake.