China’s DeepSeek sparked a historic AI stock selloff

DeepSeek is driving US markets to seek deep for answers.

AI infrastructure stocks fell across the board yesterday, as investors digested news that upstart Chinese AI firm DeepSeek had developed a pair of top-performing models at a fraction of the cost compared to Silicon Valley giants like OpenAI, Google, and Meta.

Why it’s a big deal: Prior to DeepSeek’s latest models – which have displaced ChatGPT at the top of Apple’s App Store – the conventional thinking was that AI companies required expensive, high-end computer chips to train the best systems. This belief served as justification for major US tech firms to invest hundreds of billions of dollars towards building, training, and supporting AI models.

• However, DeepSeek didn’t have anywhere close to that amount of financial power – one of its two top models reportedly cost just $5.6 million to train using reduced-capability chips.
• Despite this, DeepSeek’s AI models are still roughly on par with top US rivals, largely thanks to the company’s innovative training techniques.

The market reacts

Nvidia, whose high-end chips have been used by the tens of thousands to power many leading AI models, saw its shares fall ~17% yesterday. The move wiped out ~$593 billion from the company’s market value – the biggest single-day figure in history by more than 2x.

Other companies with AI infrastructure interests suffered double-digit losses, including Oracle (-14%), chipmakers Broadcom (-17%) and Micron (-12%), server maker Super Micro Computer (-13%), and power provider Constellation Energy (-21%).

• Overall, Monday’s selloff removed ~$1 trillion from the US stock market’s value, per Dow Jones Market Data.

👀 Looking ahead…While some analysts say it’s too soon to determine whether DeepSeek has upended the AI infrastructure industry, others predict DeepSeek’s success means AI models will soon require fewer chips and energy.