Did We Just Witness AI Black Monday? DeepSeek Shocks Tech StocksPanic sell, panic sell, panic sell! That’s basically how Monday went for Wall Street and those of you who hold Nvidia shares. Or just about any other tech stock — you name it, it likely fell nose first when a big and scary Chinese artificial intelligence startup unveiled its new AI model.
DeepSeek.
What in the world is DeepSeek and why do I hear about it now?
DeepSeek, a Chinese artificial intelligence startup, may have just stripped Nvidia of its untouchable status as the go-to company that develops expensive chips to train AI models. DeepSeek announced it had trained its latest model, a rival of ChatGPT, for the negligible $5.6 million in computing costs. The story gets even crazier: it did it with 2,048 Nvidia H800 GPUs (bought before the US rolled out export restrictions).
That’s a meager 5% of the $100 million OpenAI blew on training its GPT-4 model in late 2023. And, what’s even more, DeepSeek’s model, called R1, churns out responses that are scarily close to the advanced US-bred technology.
Oh, and it’s open source, unlike OpenAI, which was originally open source but shut its doors to the public. It’s also free to use, unlike ChatGPT, which offers a paid tier between $240 and $2,400 a year. DeepSeek’s R1 model is quickly gaining traction among users as it made its way to the top of Apple’s App Store rankings.
DeepSeek has factored in demand from corporations, too. While OpenAI hosts the model on its own platform, its Chinese rival allows you to host this beast on your own hardware, which is a big deal to lots of businesses that work with sensitive data.
The stock market was so shocked by the news that you can get pretty much the same result for a fraction of the cost (and give it to users for free), it ran for the hills. The aftermath — Monday saw more than $1 trillion washed out from the valuation of the Magnificent Seven club. One company specifically took the biggest blow.
Can DeepSeek deep-six Nvidia’s world dominance plans?
Have companies been overpaying for Nvidia’s $30,000 chips? And have investors been overpaying for Nvidia’s shares? Nvidia NVDA pulled in a record $35 billion in Q3 , 2024 and struck a gross margin of 75% and net income of $20 billion.
The Jensen Huang-led company on Monday showed it can also hit records in reverse. Closing down 17% for the cash session, it took the biggest L in history. This was the largest destruction of value for a single company ever — $589 billion . So why was Nvidia particularly hit by DeepSeek’s rise?
Nvidia has been the primary beneficiary of the vast amounts of cash companies spent on AI. Simply because Nvidia makes the semiconductors used to train AI models. But if the same result (or just about the same) could be achieved through far less expensive means, why bother propelling Nvidia to the top echelon of the world’s biggest companies ?
Nvidia has picked up roughly $131 billion over the past two years from the sale of data-center equipment, mostly AI chips. Its client list includes the biggest names in tech, such as Amazon AMZN , Microsoft MSFT , Meta META and Alphabet GOOGL . These four combined have shelled out $343 billion in AI-related capex (capital expenditures) over the past two years. Since the release of ChatGPT, Nvidia shares have surged more than 700%.
Could we be looking at the good old supply and demand equation in play? If DeepSeek’s claims are true, and other companies can do the same (it’s an open-source model), then the scales could turn from undersupply to oversupply.
Can we then see a market crash that’s beyond anything we’ve ever thought possible? Or is that freak-out an unjustified stretch? Share your thoughts in the comment section below.
