Stay informed with free updates
Simply sign up to the Artificial intelligence myFT Digest — delivered directly to your inbox.
Triumphalist glee lit up the Chinese internet this week. Just as Google DeepMind’s victory over China’s strongest Go player in 2017 showcased western brilliance in artificial intelligence, so DeepSeek’s release of a world-beating AI reasoning model has this month been celebrated as a stunning success in China.
DeepSeek’s smarter and cheaper AI model was a “scientific and technological achievement that shapes our national destiny”, said one Chinese tech executive. The start-up had become a key player in the “Chinese Large-Model Technology Avengers Team” that would counter US AI dominance, said another.
China’s delight, however, spelled pain for several giant US technology companies as investors questioned whether DeepSeek’s breakthrough undermined the case for their colossal spending on AI infrastructure. US tech and energy stocks lost $1tn of their market value on Monday, although they regained some ground later in the week.
The stereotypical image of China abroad may still be that of a state-subsidised, capital-intensive manufacturing economy that excels at churning out impressive low-cost hardware, such as smartphones, solar panels and electric vehicles. But, in truth, China long ago emerged as a global software superpower, outstripping the west in ecommerce and digital financial services, and it has invested massively in AI, too.
DeepSeek’s emergence confounds many of the outworn prejudices about Chinese innovation, although it is far from a typical Chinese company. It certainly invalidates the old saw that while the US innovates, China imitates and Europe regulates. In several ways, DeepSeek resembles a bootstrapped Silicon Valley start-up, even if it was not founded in a garage. Launched in 2023, the company has the same high-flown ambition as OpenAI and Google DeepMind to attain human-level AI, or artificial general intelligence (AGI). But its founder Liang Wenfeng runs one of China’s leading hedge funds, meaning the company has not had to raise external financing.
In an interview republished in the China Talk newsletter, Liang explained that DeepSeek operated more as a research lab than a commercial enterprise. When recruiting, it prioritised capabilities over credentials, hiring young Chinese-educated researchers. Liang said these people were given the space to explore and the freedom to make mistakes. “Innovation often arises naturally — it’s not something that can be deliberately planned or taught,” he said.
DeepSeek relies on open-source AI models, such as Meta’s Llama, in contrast to the proprietary models favoured by OpenAI and Google. It also focuses narrowly on language in its quest to reach AGI rather than attempting to go multimodal and incorporating images, audio and video. “What you think of as ‘thinking’ might actually be your brain weaving language. This suggests that humanlike AGI could potentially emerge from language models,” he said.
DeepSeek’s focused approach has enabled it to develop a compelling reasoning model without the need for extraordinary computing power and seemingly at a fraction of the cost of its US competitors. As with other Chinese apps, US politicians have been quick to raise security and privacy concerns about DeepSeek. And OpenAI has even accused the Chinese company of possible breaches of intellectual property rights. Given the cases against OpenAI for infringing others’ copyright, though, that might strike some as rich.
While some big US tech companies responded to DeepSeek’s model with disguised alarm, many developers were quick to pounce on the opportunities the technology might generate. The capabilities and cheapness of DeepSeek’s reasoning model may allow them to deploy it for an ever-expanding number of uses. On Monday, DeepSeek was the most downloaded free app on the US Apple App Store.
Ironically, that may yet enable the US to benefit more from DeepSeek’s breakthrough than China. Over the past few years, China has been throttling its own private sector as the state has exerted tighter control. The number of start-ups launched in China has plummeted since 2018. According to PitchBook, venture capital funding in China fell 37 per cent to $40.2bn last year while rising strongly in the US.
DeepSeek has punctured the hubris of the US tech oligarchs. It has intensified global competition and will accelerate the adoption of AI tools. Temporarily this could be a case of China innovating and the US imitating. But is it just a spectacular blip or the start of a long-term trend?
Read the full article here
