A Chinese AI firm, DeepSeek, is capturing attention in the U.S. and causing ripples in Wall Street due to its innovative technology, which some analysts suggest competes directly with OpenAI’s ChatGPT.
Investors are also taken aback by the surprisingly low development costs of DeepSeek’s AI application, estimated by Wedbush Securities analyst Dan Ives to be around $6 million. In contrast, major U.S. companies like OpenAI and Google are projected to invest approximately $1 trillion in AI over the next few years, as per Goldman Sachs.
On Monday, the launch of DeepSeek’s app sent shockwaves through the stock market, affecting established players like Nvidia, known for manufacturing advanced chips for AI, and ASML, another chip producer. Experts are questioning whether demand for Nvidia’s chips might decline and if tech stocks buoyed by AI projections—ranging from Meta to Microsoft—are overvalued.
“DeepSeek has disrupted the market by achieving more with less,” commented Giuseppe Sette, president of AI market research firm Reflexivity, via email. “This indicates that the AI landscape will continue to surprise us in the coming years.”
DeepSeek’s latest application was launched shortly after President Trump announced a new $500 billion initiative involving OpenAI, Softbank, and Oracle, which he described as pivotal for the “future of technology” in the U.S.
Shares of AI-related stocks faced significant downturns on Monday, with Nvidia’s value plunging 17%—a loss of $600 billion, marking the largest single-day drop for a company in stock market history, according to CNBC. ASML fell by 6%, while Broadcom, another semiconductor firm, also dropped 17%.
Additionally, some energy sector stocks declined sharply due to concerns that this new technology could lead to reduced energy requirements, indicating lower demand from the tech industry. GE Vernova, which manufactures wind and gas turbines, plummeted 21%, and electricity provider Vistra saw a 28% decrease.
The tech-focused Nasdaq index fell 3%, equating to a loss of 612 points, while the S&P 500 decreased by 1.5%. Meanwhile, the blue-chip Dow Jones Industrial Average saw a modest gain of 0.7%.
Despite the notable drop in the Nasdaq, it is not the most significant decline for the index in the past five years. The most considerable one-day decrease since January 27, 2020, occurred on March 16, 2020, when the index fell over 12% in response to the COVID-19 pandemic.
What is DeepSeek?
DeepSeek, established in July 2023 by Liang Wenfeng—a graduate of Zhejiang University, one of China’s premier universities—was exclusively funded through his hedge fund, as noted by the MIT Technology Review. Liang boasts assets worth about $8 billion, per Ives’ report from January 27.
Previously focused on applying AI for investment purposes, Liang acquired a “stockpile of Nvidia A100 chips,” which are now banned from export to China. These chips serve as the backbone of DeepSeek, according to the MIT publication.
Ben Reitzes, head of technology research at Melius, informed investors that DeepSeek is making substantial improvements as an AI tool, delivering enhanced learning capabilities and more efficient memory usage, though he expressed doubts regarding their “chip utilization.”
Is DeepSeek available in the U.S.?
The AI application is accessible through Apple’s App Store and on the company’s website. It is offered for free and became the top download on Apple’s store as of Monday morning, although some users experienced difficulties with the signup process.
On its Chinese platform, DeepSeek cited “large-scale malicious attacks” affecting its service, necessitating temporary restrictions on new registrations. “Current users can log in as usual,” the company stated in a message dated shortly after midnight on January 28 in China’s local timezone.
The company introduced its latest AI model on January 20, inciting Wall Street to reevaluate the AI industry.
“Last week, DeepSeek launched a model that stands up to OpenAI’s ChatGPT and Meta’s Llama 3.1 and reached #1 on the Apple App Store over the weekend,” wrote Wedbush’s Ives. “DeepSeek constructed this model using reduced capability chips from Nvidia, which is notable and has led to significant unrest among U.S. tech stocks, putting immense pressure on Nasdaq this morning.”
How is DeepSeek different from other AI apps?
DeepSeek is an open-source large language model that utilizes what is known as “inference-time computing,” which Sette described in simpler terms as “activating only the most pertinent sections of the model for each query, conserving both financial resources and computational power.”
Some experts have lauded DeepSeek’s capabilities, including prominent tech investor Marc Andreessen, who remarked on X on January 24, “DeepSeek R1 is one of the most extraordinary breakthroughs I’ve ever encountered — and as open source, it is a significant gift to the world.”
David Sacks, a venture capitalist appointed by Mr. Trump to help steer AI and cryptocurrency policy, commented on social media Monday that DeepSeek’s app “demonstrates that the AI competition will be fierce.”
However, Ives remains doubtful that this service will gain traction among leading U.S. corporations.
“No U.S. Global 2000 company will implement DeepSeek from this Chinese startup to establish their AI infrastructure and related use cases,” Ives stated. “Ultimately, there is only one chip manufacturer in the world developing autonomous robotics and broader AI applications, and that is Nvidia.”
Are there privacy issues with DeepSeek?
Concerns regarding privacy have been raised by some experts due to the personal information that DeepSeek gathers, as the company maintains user data—such as date of birth, keystrokes, text or audio entries, uploaded files, chat records, and other details—on servers positioned in China, as outlined in its privacy policy.
This situation mirrors concerns associated with TikTok, another Chinese-owned entity that raised alarms about potential risks related to its Chinese affiliations to national security. Last year, Congress enacted a law prohibiting TikTok in the U.S. as long as it remains under Chinese ownership, although this position is now subject to change following President Trump’s issuance of an executive order directing the Justice Department to pause enforcement of the ban for 75 days.
What does DeepSeek imply for Nvidia and other tech companies?
Wall Street is evaluating the long-term ramifications of a budget-friendly AI tool from China that stands against ChatGPT and other generative AI applications. This development prompts inquiries into whether Silicon Valley is overspending on technological advancements within the AI sphere, according to Angelo Zino, a senior equity analyst at CFRA Research, in an email.
“The premise that this technology is expected to consume less energy and be more cost-effective than U.S.-based models has stirred considerable anxiety among U.S. technology investors,” remarked Jay Woods, chief global strategist at Freedom Capital Markets.
Furthermore, the type of response or pushback from the White House remains uncertain, as Mr. Trump has hinted at the possibility of imposing new tariffs on Chinese imports, even though he previously granted TikTok, owned by a Chinese firm, a reprieve by instructing the Justice Department not to enforce a pending ban.
Some analysts on Wall Street believe the stock sell-off on Monday was excessive, maintaining that the strong demand for AI will continue to favor key players in the industry.
“While it’s one thing to train a [large language] model at a lower cost, meeting the enormous appetite for the deployment of AI technology will still necessitate substantial infrastructure,” stated Adam Crisafulli of VitalKnowledge in a report.
Bernstein Research analysts also commented that while DeepSeek’s technology shows promise, it likely does not represent a revolutionary shift, indicating that the investment rout on Monday could be overstated.
What is Nvidia saying about DeepSeek?
Nvidia has expressed its support for DeepSeek in a statement to CBS News.
“DeepSeek represents a significant advancement in AI and is a prime example of test-time scaling,” the company remarked via email. “DeepSeek’s achievements illustrate how new models can be developed using this technique by utilizing widely accessible models and compute that comply with export controls.”
However, Nvidia cautioned that AI inference, which involves employing AI models for decision-making or predictions, “demands a considerable number of NVIDIA GPUs and high-performance networking capabilities. We now operate under three scaling principles: pre-training and post-training continue on, alongside new test-time scaling.”
