In This Article:
Investing.com - The emergence of Chinese start-up DeepSeek's cut-price artificial intelligence model suggests the expenses related to training similar models is set to decline "substantially", according to analysts at Morgan Stanley (NYSE:MS).
Relatively unknown until Monday, Hangzhou-based DeepSeek roiled global markets after it introduced a free new chatbot it says can rival the likes of OpenAI's ChatGPT, albeit with less data and at a fraction of the cost.
The firm said it took a mere two months and under $6 million to build out its AI model on Nvidia (NASDAQ:NVDA)'s lower-capability H800 chips. DeepSeek's assistant run on the AI model became the top-rated app on Apple (NASDAQ:AAPL)'s App Store in the US on Monday, overtaking OpenAI's ChatGPT.
Although doubts still surround DeepSeek's claims, the statement fueled worries that massive spending on AI by the US tech industry's biggest names was unwarranted. U.S. President Donald Trump said DeepSeek was a "wake-up call" for America's tech companies.
These concerns fueled a slump in the tech-heavy Nasdaq Composite on Monday, with shares in AI-darling Nvidia in particular posting a record market cap loss of $593 billion. A broader index of semiconductor stocks also slipped to its largest single-day percentage dip since March 2020.
The sell-off extended beyond chipmakers and into AI-adjacent stocks such as power utilities Vistra and Constellation Energy (NASDAQ:CEG), both of whom have been bolstered in recent months by expectations that they would see a spike in demand as companies race to power the data centers that underpin AI applications. Data center infrastructure builders fell too, with Vertiv Holdings (NYSE:VRT) decreasing by over 30%.
Even still, Nasdaq futures were higher on Tuesday, while Nvidia's stock price pared back some of its steep losses in premarket U.S. trading.
In a note to clients, the Morgan Stanley analysts led by Brian Nowak argued the DeepSeek model "suggests the cost of training [generative AI] models is set to decline substantially [...] which we see leading to faster [generative AI] product innovation and availability from our US tech leaders".
"[T]he history of computing suggests lower costs accelerate usage -- and demand," they wrote.
A drop in computing expenses and increased adoption should drive up returns on AI investments for both the internet and software sectors, led by mega-cap players like Google-owner Alphabet (NASDAQ:GOOGL) and Facebook-parent Meta Platforms (NASDAQ:META), the analysts added.