US AI giants won’t be sunk by DeepSeek

DeepSeek is unlikely to undermine the dominance of American AI giants. Nvidia’s market capitalization experienced a significant increase due to soaring demand for AI chips and services; however, the company’s stock plummeted during a broader selloff in the semiconductor sector on Monday. While necessity is often heralded as the catalyst for innovation, the notion of triggering an unprecedented selloff appears rather far-fetched. On Monday morning, however, U.S. markets greeted the day with renewed apprehension regarding DeepSeek. A Chinese artificial-intelligence startup revealed a noteworthy advancement late last week, showcasing AI models that operate almost at the same level as sophisticated technology developed in the United States. The crux of the matter is that DeepSeek asserts it has trained one of its most recent models at a cost of $5.6 million in computing expenses—a mere fraction of the expenditures currently incurred on this side of the Pacific for similar endeavors.
The launch of OpenAI’s GPT-4 model in late 2023 has resulted in a significant increase in market values, driven by a burgeoning demand for AI chips and services. The elevated entry costs associated with AI, coupled with U.S. government sanctions restricting the sale of advanced AI chips to Chinese firms, effectively create a competitive barrier for major technology players like Microsoft, Amazon, Google, and Meta Platforms. They stand out as one of the limited number of firms possessing the requisite capital to develop extensive AI networks on a grand scale. Consequently, DeepSeek’s innovation appears to be especially detrimental for almost all firms with a market capitalization exceeding $1 trillion. Nvidia and Broadcom shares plummeted over 14% by Monday afternoon, spearheading a selloff in the semiconductor sector that eroded more than 8% from the PHLX Semiconductor Index. Microsoft and Alphabet, the parent company of Google, both recognized leaders in the provision of AI-driven cloud computing services, experienced declines of 3% during the morning trading session. The technology-focused Nasdaq experienced a decline of 3%, in contrast to a modest increase observed in the blue-chip Dow Jones Industrial Average.
The recent selloff appears to be overdone. A considerable amount remains uncertain regarding DeepSeek’s assertions, particularly concerning the types of chips the company was able to access in light of the sanctions imposed. On Monday, multiple chip analysts challenged the assertion that DeepSeek has developed a product comparable to sophisticated U.S.-based AI models at a remarkably low cost. “DeepSeek did not ‘build OpenAI for $5 million,’” stated Stacy Rasgon of Bernstein. “The ‘DeepSeek’ moment is prompting investors to act impulsively, prioritizing immediate action over subsequent inquiry,” noted Joshua Buchalter of TD Cowen. “Although DeepSeek’s accomplishment may be revolutionary, we raise doubts about the assertion that its successes were achieved without the application of sophisticated GPUs for refinement,” stated Atif Malik of Citigroup.
Moreover, this technical advancement is improbable to dampen the AI competition or diminish the financial resources being allocated to it. In discussing the parallels between DeepSeek and Sputnik, Edward Yang from Oppenheimer remarked that the Space Race did not lead to a reduction in financial outflows. “Heightened competition seldom leads to a decline in overall expenditure,” he noted in a communication to clients. Pierre Ferragu of New Street Research observed that more sophisticated “frontier models” will continue to advance the technical frontier and leverage the most cutting-edge computing resources, whereas smaller “lagging edge” models will strive to create more cost-effective AI functionalities. “DeepSeek does not represent a transformative shift; rather, it aligns seamlessly with the trajectory of industry developments observed over the past three years,” Ferragu stated.
DeepSeek’s innovation emerges at a time when investment in AI seems poised for a significant increase. On Friday, Meta CEO Mark Zuckerberg revealed intentions to increase capital expenditures once more, potentially reaching $65 billion in the current year. This development follows the Stargate Project, which encompasses collaborations between OpenAI, SoftBank, and Oracle, with ambitions to invest up to $500 billion in AI infrastructure. Microsoft, set to announce its quarterly results on Wednesday, stands as the inaugural major tech firm poised to indicate whether DeepSeek’s innovation will jeopardize its investment strategies. According to consensus estimates from Visible Alpha, the company is projected to reduce its capital expenditures by approximately $84 billion for the fiscal year concluding in June, followed by a further decline to $94 billion in the subsequent year. The company, at this juncture, is unequivocally not indicating any intention to scale back its aspirations. “As AI becomes increasingly efficient and accessible, its utilization is poised to surge, transforming it into a commodity that will be in high demand,” stated Microsoft CEO Satya Nadella in a post on X on Monday morning.
The competition for AI investment appears to be transitioning into a new stage.
Rachel Long
Rachel Long is our Desk Correspondent covering Stock Markets across the globe. She is based in New York