A broad selloff saw tech stocks sinking deep on Monday after DeepSeek sent shockwaves through Silicon Valley as the Chinese artificial-intelligence startup said it trained high-performing AI models cheaply.
DeepSeek, a one-year-old startup, presented a ChatGPT-like AI model called R1, which has all the abilities of other models. However, it is operating at a fraction of the cost of OpenAI’s, Google’s and Meta’s popular AI models.
|
|
On Monday, DeepSeek’s AI assistant rapidly overtook ChatGPT as the most popular free app in Apple’s US and UK app stores, with attendant impact on the presumed dominance of the AI ecosystem by US companies.
US stocks dropped sharply by the close of business on Monday, with the tech-heavy Nasdaq dropping 3% and chipmaker Nvidia losing nearly $589 billion—or 17%, considered the biggest daily decline in dollars for a publicly listed company—in market value, as the frenzy generated by the Chinese startup reached its crescendo.
|
|
|
|---|
Deepseek said it had spent just $5.6 million on computing power for its base model, compared with the billions of dollars US companies spend on their AI technologies.
After the R1 was launched earlier in the month, the company boasted of “performance on par with” one of OpenAI’s latest models when used for natural language reasoning, maths, among other tasks.
On Monday, DeepSeek said it would temporarily limit user registrations “due to large-scale malicious attacks” on its services, though existing users will be able to log in as usual.
On Tuesday afternoon, multiple attempts by our correspondent to log in on the service failed.
Rebound
On Tuesday, checks by PREMIUM TIMES showed that Nvidia (NVDA) and other artificial intelligence stocks are rebounding in premarket trading.
After tumbling 17% Monday, Nvidia is rising almost 2% in premarket trading. In its reaction to the market disruption, Nvidia described DeepSeek’s AI model as 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,” the statement said.
There are multiple estimates on exactly how much DeepSeek’s R1 costs, or how many graphics processing units went into it. But the startup’s models stunned analysts because they were built despite the US curbing chip exports to China three times in the last 36 months.
According to a research paper released last week, Deepseek’s development team said it spent less than $6m on computing power to train the model. That represents a fraction of the multibillion-dollar AI budgets enjoyed by US tech giants OpenAI and Google, creators of ChatGPT and Gemini.
Although US tech stocks seem to be recovering, Deepseek may have challenged widespread assumptions about US dominance in AI revolution amid uncertainties surrounding the market valuations of its top tech companies.
Last year, Sam Altman, CEO of OpenAI, said the AI industry would need trillions of dollars in investment to support the development of high-in-demand chips needed to power data centres that run the sector’s complex models.
What is Deepseek?
Based in Hangzhou, Deepseek was founded in late 2023 by Liang Wenfeng, a serial entrepreneur who also runs the hedge fund High-Flyer.
Little is known about the young innovator, as he has kept a rather low profile over the years. According to Chinese media outlet Sina Finance, in 2013, he co-founded Hangzhou Jacobi Investment Management, an investment firm that employed AI to implement trading strategies, along with a co-alumnus of Zhejiang University.
In 2023, in a rare interview with Chinese media outlet Waves, Mr Liang dismissed the suggestion that innovations around AI was costly for startups.
“Reproduction alone is relatively cheap — based on public papers and open-source code, minimal times of training, or even fine-tuning, suffices. Research, however, involves extensive experiments, comparisons, and higher computational and talent demands,” Liang said, according to a translation of his comments published by the ChinaTalk Substack.
Meanwhile, there are concerns that the R1 appears to censor answers to sensitive questions about China and its government.
For instance, Chinese generative AI must not contain content that violates the country’s “core socialist values”, according to a document published by the national cybersecurity standards committee. Such content are described as those that incite “to subvert state power and overthrow the socialist system”, or any content that “endangers national security and interests and damages the national image”.
Australia’s science minister, Ed Husic, on Tuesday also raised privacy concerns about DeepSeek AI model, which, like all Chinese tech inventions, have repeatedly been the subject of allegations and fears that they could lead to peoples’ data being harvested for intelligence purposes by the Chinese state.
Husic told ABC News on Tuesday there remained a lot of unanswered questions on the usage of Deepseek’s, especially on data and privacy management.
“I would be very careful about that, these types of issues need to be weighed up carefully,” he added.
What next?
Deepseek’s entry into the market has sent shockwaves through the US and beyond principally because AI is a cost-intensive technology. Companies like Meta and OpenAI have argued that the AI industry would need trillions of dollars in investment to support the development of high-in-demand chips needed to power AI data centres.
Last year, Google signed a deal to use small nuclear reactors to generate the vast amounts of energy needed to power its artificial intelligence (AI) data centres.The company said the agreement with Kairos Power will see it start using the first reactor this decade and bring more online by 2035.
On the contrary, Deepseek’s ability to replicate America’s powerful AI models with a small fraction of the cost — and on less capable chips — has changed the dynamics of conversations on how much investment is indeed needed in AI, to advance the global economy into a new era and open up new frontiers of knowlege and possibilities.
Marc Andresssen, co-founder of Silicon Valley venture capital firm Andreessen Horowitz, described R1 as an amazing gift to the world. “Deepseek R1 is one of the most amazing and impressive breakthroughs I’ve ever seen — and as open source, a profound gift to the world,” he tweeted.
In his reaction to the new development, US President Donald Trump described the rise of DeepSeek as “a wake-up call” for the US tech industry. “If you could do it cheaper, if you could do it [for] less [and] get to the same end result. I think that’s a good thing for us,” he told reporters on board Air Force One.
