Investing.com - Shares in artificial intelligence-darling Nvidia (NASDAQ:NVDA) tumbled nearly 17% Monday, erasing $589 billion in market value and weighing on the tech-heavy Nasdaq Composite, as a new model from Chinese start-up DeepSeek cast doubt over heavy recent investments in AI infrastructure. Despite the massive sell-off, NVIDIA praised DeepSeek, calling it an "excellent AI advancement and a perfect example of Test Time Scaling."
Nvidia closed down 16.86% and lost more market value in one day than any other company in history.
"For investors holding Nvidia or related stocks, the critical question is whether [capital expenditure] growth in 2026 can maintain its trajectory, especially in a market priced for perfection," US Tiger Research analysts wrote in a client note.
"Until there’s clarity, Nvidia’s increased uncertainty will pressure its stock price."
Broader Wall Street indices also fell sharply, after DeepSeek launched a set of open-source models over the past week claiming to match offerings from rivals such as OpenAI's ChatGPT in performance for a fraction of the cost.
The release and testing of the company’s flagship model -- DeepSeek R1 -- sparked questions over a surge in spending on building out AI infrastructure by major US tech companies.
Bernstein analysts noted that the new models were impressive, especially in their ability to compete with similar products from OpenAI and Meta Platforms (NASDAQ:META). However, they said the broader market reaction to DeepSeek's models appears to be overblown.
“If we acknowledge that DeepSeek may have reduced costs of achieving equivalent model performance by, say, 10 [times], we also note that current model cost trajectories are increasing by about that much every year anyway [...] which can’t continue forever,” Bernstein analysts wrote in a note.
They added that the AI industry needed innovations like DeepSeek to keep progressing, because it could help squeeze more out of existing hardware.
The brokerage noted that the need for increased AI infrastructure remains, saying that any fresh computing capacity was likely to get absorbed by a jump in AI demand.
Elsewhere, market watchers questioned if DeepSeek is using just 10,000 A100 NVIDIA chips as the company claims. Alexandr Wang, the billionaire CEO of Scale AI, thinks it is closer to 50,000 more-powerful NVIDIA Hopper GPUs but believes the company can't disclose the truth due to U.S. export controls on AI chips. Tesla (NASDAQ:TSLA) CEO Elon Musk agrees with Wang's assessment.
NVIDIA, for its part, praised DeepSeek, suggesting it is just the latest way to scale AI powered by NVIDIA chips.
"DeepSeek is an excellent AI advancement and a perfect example of Test Time Scaling," a NVIDIA spokesperson told Investing.com Monday afternoon. "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. Inference requires significant numbers of NVIDIA GPUs and high-performance networking. We now have three scaling laws: pre-training and post-training, which continue, and new test-time scaling.”
(Ambar Warrick, Senad Karaahmetovic, and Scott Kanowsly contributed to the reporting.)
Comments