The tech giant\'s stock began the day trading at $140 but fell 16% and closed at $118 as news of DeepSeek’s low-cost AI developments and high-performance spread.
Its plunge reflects a broader trend, with the U.S. tech industry shedding an estimated $1 trillion in market capitalization and those concerns spreading to other areas of the economy, including crypto.
DeepSeek\'s ability to outperform OpenAI’s o1 while operating at a significantly lower cost, reportedly under $5 million, has rocked the tech industry, which has prioritized computational power over efficiency.
Its success has propelled DeepSeek to become the top free app in the U.S., according to Appfigures’ data.
Another factor fueling the sell-off is claims that Chinese developers are training their AI models on Nvidia H100 chips that the U.S. barred Nvidia from selling to them, raising questions about the effectiveness of U.S. export controls and China’s access to advanced hardware.
\"The Chinese labs, they have more H100s than people think, you know,\" Scale AI CEO Alexandr Wang told CNBC. \"My understanding is that DeepSeek has about 50,000 H100s, which they can\'t talk about, obviously, because it is against the export controls that the U.S. has put in place.\"
The U.S.\'s reluctance to embrace open-source AI development may have given Chinese companies an edge in the AI development field, according to Professor Ion Stoica, a computer scientist at UC Berkeley and co-founder of Databricks and Anyscale, citing regulatory concerns and national security fears.
“When I say open source, I mean open data, open training algorithms, open weights, and open evaluations—maximum visibility into how they\'re trained and what they\'re trained on,” Stoica told Decrypt. “Now we’re in a situation where Chinese companies bet on open source, unlike the U.S., and are clearly ahead.\"
While Stoica declined to speculate on the tech stock market, he emphasized the transformative potential of lower AI model costs.
“If the cost of building or serving these models drops by 10 or 100x, it could hurt these companies,” he said. “On the other hand, if it drives innovation and accelerates AI development by doing more with the same hardware, these companies could become even more valuable.
Edited by Sebastian Sinclair