China’s DeepSeek could upend AI business models. Here’s why – National
Chinese startup DeepSeek sent shockwaves through financial markets Monday on claims that it could develop advanced artificial intelligence models using much cheaper semiconductors than previously thought possible.
Those implications sent shares of Nvidia, a chipmaker whose value surged in recent years on promises to power the AI revolution, down double digits as markets opened for the week.
Other tech stocks also took a hit Monday as investors grappled with what the new player’s emergence could mean for the future of the AI industry.
“This company may not be the winner at the end of the day, but it’s caused a lot of people to question the model that had been the expectation a few weeks ago,” Greg Taylor, chief investment officer at Purpose Investments, told Global News.
The S&P 500 was down 1.8 per cent in afternoon trading and heading for its worst day in more than a month. Big Tech stocks took some of the heaviest losses, with Nvidia down 17.4 per cent, and they dragged the Nasdaq composite down 3.5 per cent.
The S&P/TSX composite, Canada’s benchmark index, was down roughly one per cent by 2 p.m. Eastern.
The flurry of activity on Monday revolved around DeepSeek and its AI assistant, which as of Monday morning had unseated OpenAI as the top-downloaded free application on Apple’s App Store charts in the United States.
The startup DeepSeek was founded in 2023 in Hangzhou, China, and released its first AI large language model later that year. Its CEO Liang Wenfeng previously co-founded one of China’s top hedge funds, High-Flyer, which focuses on AI-driven quantitative trading.
The startup says its apps are on par with the kinds of advanced AI models that power popular chatbots such as Silicon Valley-based OpenAI’s ChatGPT.
They are also cheaper to use. The DeepSeek-R1, released last week, is 20 to 50 times cheaper to use than OpenAI o1 model, depending on the task, according to a post on DeepSeek’s official WeChat account.
These advancements are particularly notable because the U.S. has banned exports to China of the most advanced AI chips, such as the latest semiconductors from Nvidia. Some have expressed skepticism that DeepSeek has been able to make such progress on the backs of Nvidia’s older generation technology.
Get weekly money news
Get expert insights, Q&A on markets, housing, inflation, and personal finance information delivered to you every Saturday.
If a Chinese startup has been able to replicate U.S. firms’ work on the back of cheaper, less sophisticated chips, that could have widespread implications for the future of machine learning and the so-called AI arms race.
Valerie Wirtschafter, fellow in foreign policy and artificial intelligence at the Brookings Institute, tells Global News that DeepSeek’s emergence could represent a “Sputnik moment” for the U.S., where the Americans believed they had dominance in an emerging field, only to be surprised by innovation from a foreign rival, a-la Russia’s launch of the Sputnik satellite in 1957.
“At least in the United States, the thought was, if we prevent China from having the hardware, the chips to process this information, they won’t be able to compete with the United States in terms of developing these large language models,” she says.
“I think the incentives of not selling China the chips actually worked against the United States, because they innovated. So they were able to do more with less.”
Why are other AI tech stocks taking a hit?
AI-linked stocks took a hit on Monday because DeepSeek’s claims, if they hold, push back on the idea that the key to success in AI was pumping money into the most expensive chips and infrastructure like advanced data centres.
Major U.S. players recently doubled down on this approach with the announcement of the Stargate Project — US$500 billion in spending over four years aimed at bolstering the American AI industry with support from President Donald Trump’s second administration.
If similar advancements can be made with more affordable chips, however, the business model could change significantly.
Maura Grossman, computer science professor at the University of Waterloo, tells Global News that there will still be a need for chips from the likes of Nvidia, but its reputation as a go-to destination for AI software developers could be at risk with DeepSeek’s emergence.
“It’s like, you were selling shovels during the gold rush and now nobody needs shovels anymore — spoons work,” she says. “And what happens to you if you’re the one selling the shovels?”
Grossman notes that if companies can get the same AI output with less resource-intensive technologies, that could have a positive environmental impact, as the top-line data centres that power artificial intelligence applications have significant demands for electricity, water and critical minerals.
Taylor says that the current moment calls back to the dot-com bubble at the turn of the millennium, when firms such as Nortel and JDS were rushing to install as much fibre-optic cable as possible to meet other firms’ networking demands.
“And then suddenly we woke up one day and said, ‘We have way more fibre than we’ll ever need. And all these companies just crashed. And I think that’s the parallel that some investors are making this morning,” Taylor says.
“That’s the big wake-up we’ve got right now. And I think that’s why the market’s having a bit of a stumble.”
Taylor notes that it was never realistic for Nvidia to have a “winner-take-all” share of the AI chips market. He suggests that some of the hype around AI in the market might start to shift away from the chipmakers like Nvidia towards the users — software companies such as Apple who are developing applications.
Wirtschafter and Grossman both say that the DeepSeek developments might fuel even more demand for AI chips as the barriers to entry get lower for companies that want to start building that infrastructure.
Up until now, it has been prohibitively expensive to get into the AI game without significant capital available upfront. The cheaper it is to get involved, the more competition could come to the AI industry.
“There’ll still be a need for their chips,” Grossman says. “Maybe not at the same frenzied level that it has been up to this point, especially if people start to find ways to do the same quality work with lesser chips.”
Taylor says one of the biggest questions going forward is how Trump will react. AI has been one of the sectors he’s been keenest about championing under his America-first banner, Taylor says, and any perceived threat to that industrial dominance could provoke a swift response in the already frayed U.S.-China trade relationship.
Steep tariffs, as Trump has threatened to wield against Canada, Mexico and China already in the first week of his second presidential term, could play a role in the U.S. response to DeepSeek, Taylor suggests.
— with files from Global News’s Nathaniel Dove, Reuters and the Associated Press
