DeepSeek releases new image model as U.S. tech stocks plummet
DeepSeek has unveiled a new image model in the wake of a significant decline in U.S. tech stocks.
TroibNews 
U.S. tech stocks experienced substantial declines on Monday as investors reevaluated the economic repercussions of the rapid advancements made by Chinese startup DeepSeek, particularly in relation to its newly launched Janus-Pro-7B image generator.
The company, based in Hangzhou, announced that its latest model surpasses OpenAI's DALL-E 3 in performance benchmarks. Moreover, its DeepSeek assistant app has surpassed ChatGPT in terms of downloads on the U.S. App Store. DeepSeek also disclosed that it had temporarily paused new user registrations following a cyberattack.
Analysts indicated that the selloff revealed deeper issues, such as inflated valuations, worries about excessive infrastructure spending hype, and increasing evidence that U.S. chip export restrictions are failing to impede Chinese advancements in AI.
Nvidia was particularly hard hit, plummeting nearly 17 percent and erasing almost $600 billion in market value, marking the largest single-day loss in Wall Street history. The Nasdaq dropped 3.1 percent, while the Philadelphia Semiconductor Index saw a significant decline of 9.2 percent, its worst performance since March 2020. Oracle, Broadcom, and data center company Vertiv Holdings each experienced declines exceeding 13 percent, and power utility companies tied to AI infrastructure demand, such as Vistra and Constellation Energy, fell by as much as 28 percent. The downturn had global repercussions, impacting Japan's SoftBank and Europe's ASML.
"If it's true that DeepSeek is the proverbial 'better mousetrap,' that could disrupt the entire AI narrative that has helped drive the markets over the last two years," said Annex Wealth Management's Brian Jacobsen. "It could mean less demand for chips, less need for a massive build-out of power production to fuel the models, and less need for large-scale data centers."
DeepSeek's recent launches—the Janus-Pro-7B and the DeepSeek-R1 less than a week ago—fueled existing skepticism. The Janus-Pro-7B, which improves on an image model introduced in December, utilizes 72 million synthetic images to achieve what its developers describe as "significant advancements" in multimodal understanding and text-to-image instruction capabilities. The sudden popularity of the DeepSeek-V3-powered assistant in the U.S. showcased China's emerging strength in consumer AI. Notably, researchers asserted that the V3 model was trained using only $6 million worth of Nvidia's limited H800 chips—a small fraction of what is typical in the industry—and that the API for the R1 model is priced at around 1/30th of OpenAI's offerings.
The situation drew attention from Silicon Valley figures, with venture capitalist Marc Andreessen calling DeepSeek's rise a "Sputnik moment," invoking the Soviet satellite that ignited the space race.
The stock selloff highlighted conflicting views on the future of AI. Daniel Morgan, senior portfolio manager at Synovus Trust Company—who oversees almost a million shares of Nvidia—described the market reaction as exaggerated, explaining that DeepSeek's AI model is geared toward mobile phones and PCs rather than data centers, which are where significant profits are typically found.
Nevertheless, even those with optimistic outlooks recognized the pressing questions raised, especially after former President Trump referred to DeepSeek as a "wake-up call" for U.S. competitiveness.
Established in 2023 by quantitative trading expert Liang Wenfeng, DeepSeek has rapidly transitioned from relative obscurity to garnering admiration in Silicon Valley within a matter of months. Its open-source approach and ability to amass chips have created a notable case study in China's capacity to innovate despite sanctions.
At CN Digital, we are dedicated to providing coverage on the latest developments in AI. We will continue to track these stories and offer insightful summaries to keep you informed. We also invite you to share your thoughts and engage in discussions about AI with our community—let’s keep the conversation going!
Anna Muller contributed to this report for TROIB News
The company, based in Hangzhou, announced that its latest model surpasses OpenAI's DALL-E 3 in performance benchmarks. Moreover, its DeepSeek assistant app has surpassed ChatGPT in terms of downloads on the U.S. App Store. DeepSeek also disclosed that it had temporarily paused new user registrations following a cyberattack.
Analysts indicated that the selloff revealed deeper issues, such as inflated valuations, worries about excessive infrastructure spending hype, and increasing evidence that U.S. chip export restrictions are failing to impede Chinese advancements in AI.
Nvidia was particularly hard hit, plummeting nearly 17 percent and erasing almost $600 billion in market value, marking the largest single-day loss in Wall Street history. The Nasdaq dropped 3.1 percent, while the Philadelphia Semiconductor Index saw a significant decline of 9.2 percent, its worst performance since March 2020. Oracle, Broadcom, and data center company Vertiv Holdings each experienced declines exceeding 13 percent, and power utility companies tied to AI infrastructure demand, such as Vistra and Constellation Energy, fell by as much as 28 percent. The downturn had global repercussions, impacting Japan's SoftBank and Europe's ASML.
"If it's true that DeepSeek is the proverbial 'better mousetrap,' that could disrupt the entire AI narrative that has helped drive the markets over the last two years," said Annex Wealth Management's Brian Jacobsen. "It could mean less demand for chips, less need for a massive build-out of power production to fuel the models, and less need for large-scale data centers."
DeepSeek's recent launches—the Janus-Pro-7B and the DeepSeek-R1 less than a week ago—fueled existing skepticism. The Janus-Pro-7B, which improves on an image model introduced in December, utilizes 72 million synthetic images to achieve what its developers describe as "significant advancements" in multimodal understanding and text-to-image instruction capabilities. The sudden popularity of the DeepSeek-V3-powered assistant in the U.S. showcased China's emerging strength in consumer AI. Notably, researchers asserted that the V3 model was trained using only $6 million worth of Nvidia's limited H800 chips—a small fraction of what is typical in the industry—and that the API for the R1 model is priced at around 1/30th of OpenAI's offerings.
The situation drew attention from Silicon Valley figures, with venture capitalist Marc Andreessen calling DeepSeek's rise a "Sputnik moment," invoking the Soviet satellite that ignited the space race.
The stock selloff highlighted conflicting views on the future of AI. Daniel Morgan, senior portfolio manager at Synovus Trust Company—who oversees almost a million shares of Nvidia—described the market reaction as exaggerated, explaining that DeepSeek's AI model is geared toward mobile phones and PCs rather than data centers, which are where significant profits are typically found.
Nevertheless, even those with optimistic outlooks recognized the pressing questions raised, especially after former President Trump referred to DeepSeek as a "wake-up call" for U.S. competitiveness.
Established in 2023 by quantitative trading expert Liang Wenfeng, DeepSeek has rapidly transitioned from relative obscurity to garnering admiration in Silicon Valley within a matter of months. Its open-source approach and ability to amass chips have created a notable case study in China's capacity to innovate despite sanctions.
At CN Digital, we are dedicated to providing coverage on the latest developments in AI. We will continue to track these stories and offer insightful summaries to keep you informed. We also invite you to share your thoughts and engage in discussions about AI with our community—let’s keep the conversation going!
Anna Muller contributed to this report for TROIB News
