The news that the Chinese startup DeepSeek had made a breakthrough in artificial intelligence (AI) and had become the most popular AI model in the US, surpassing ChatGPT in App Store downloads, shook global markets and Western tech giants at the end of January. Source: Financial Times.
Additionally, the new DeepSeek language model, which is comparable in performance to OpenAI’s solutions, was developed at a significantly lower cost than anticipated. It emerged that a robust AI model does not necessitate an extensive quantity of Nvidia GPUs, which have become a scarce commodity in China as a result of US sanctions.
The news was sudden and unexpected: Nvidia, the most important Western chipmaker, experienced a nearly 18% decline in its stock price in a single day, resulting in a loss of over $600 billion in market capitalization.
That is nearly $150 billion more than the recent allocation of funds to OpenAI by the newly elected US President Donald Trump to counter China’s dominance in the AI market. Furthermore, Nvidia experienced the most significant single-day stock decline in history, falling from first to third place among the world’s most valuable companies.
The shockwave did not only affect Nvidia. Marvell Technology plummeted 14%, Super Micro Computer plummeted 9%, Broadcom lost 12%, and Palantir plummeted 6%. Additionally, shares of other AI-related tech companies decreased. The cryptocurrency market also began to experience a decline.
In a single day, the total market value of US and European tech titans decreased by over $1 trillion, a figure that exceeds Spain’s GDP and is nearly twice the size of the entire Russian stock market, according to Bloomberg. China had effectively transformed the market into a playground.
Investors are currently in the process of urgently reevaluating their multibillion-dollar investments in AI infrastructure. Marc Andreessen, a venture capitalist, referred to the event as a “AI Sputnik moment,” drawing a comparison to the astonishment that accompanied the Soviet Union’s launch of the first satellite. Indeed, DeepSeek is tens of times cheaper than OpenAI’s neural networks, despite the fact that it outperforms American AI models in every manner. Additionally, its R1 model, which is comparable to OpenAI’s most potent AI, is accessible for free under an open-source license.
The Chinese startup’s domino effect also affected traditional Western IT titans. Siemens Energy, a significant provider of AI infrastructure energy solutions, experienced a 20% decline, while Schneider Electric experienced a 9.2% decline. Many people compared the market turmoil to the dot-com collapse of the early 2000s.
The Nasdaq Composite experienced a 3.5% decline, the S&P 500 experienced a 1.9% decline, and Microsoft experienced a 3.6% decline. The stocks of European chipmakers, such as ASM International and Dutch behemoth ASML (the world’s largest manufacturer of lithography machines for processors), experienced a 7.5% and 12.5% decline, respectively.
In addition to disrupting the US and EU markets, DeepSeek’s success also impacted Asian competitors. The Financial Times reports that Japanese companies experienced the greatest losses, while Chinese equities made significant gains. The Nikkei index experienced a 1.7% decline at the opening of trading in Tokyo, but it subsequently recovered to a 1.4% loss. SoftBank’s shares experienced a 5.2% decline, which was associated with a 10% decrease in Arm Holdings’ stock. Over the course of two days, SoftBank’s shares lost 12% of their value.
In the interim, Chinese companies prospered on this side of the globe. The Hang Seng index in Hong Kong rose by 0.2%, with Tencent (+1.4%), Alibaba (+0.95%), and Baidu (+2.44%) contributing to the increase. Baidu’s Hong Kong shares increased by 4% and Alibaba’s by 3% as a result of the optimism surrounding DeepSeek.
Nvidia and Broadcom were the symbols of the AI processor race until recently, with Silicon Valley giants—including Elon Musk and Sam Altman—believing that scaling up hardware was the key to AI breakthroughs. Nevertheless, the emergence of DeepSeek has raised questions about that notion.
It is remarkable that a Chinese startup valued at less than $6 million has caused a stir among Western tech titans with a combined market capitalization of trillions of dollars. Despite the fact that Google and Microsoft invest billions in AI development, DeepSeek achieved comparable results for a mere $5.6 million. DeepSeek’s ability to compete with ChatGPT while operating on less sophisticated processors immediately cast doubt on the necessity of Nvidia’s costly hardware.
The Financial Times has questioned whether the United States can sustain its AI leadership by investing billions in processors in light of DeepSeek’s breakthrough. The startup asserts that it has trained its models with significantly fewer Nvidia processors than its American competitors. This raises concerns regarding the potential profitability of such expenditures and the future AI hardware investments in Silicon Valley.
DeepSeek’s model also questions the need for significant investments in Western AI hardware. Even a modest venture can now compete with the success of industry leaders such as Microsoft, Google, and OpenAI.
In essence, a small group of Chinese engineers, who are operating on a tight budget and using antiquated processors, have developed DeepSeek—a model that surpasses the most recent ChatGPT in terms of efficiency and speed while simultaneously circumventing the bugs that afflict their American counterparts. Top administrators at Meta and other IT giants are experiencing a state of panic. The cost of producing a comparable model in the United States ranges from $60 million to $6 billion. For the purpose of comparison:
OpenAI: $6 billion
LLaMA’s Meta: $60 million
DeepSeek: $5.6 million
Despite operating on outdated hardware, DeepSeek significantly outperforms OpenAI’s models powered by state-of-the-art technology. The Chinese have demonstrated that the process of training AI models is not as intricate as previously believed and does not necessitate exorbitant computational resources. This confronts the Western semiconductor industry’s monopoly and Taiwan’s purported indispensability in AI hardware.
The American AI models that are set to be released are already lagging behind DeepSeek’s R1, which is a sign of the current state of Silicon Valley. In the interim, it is purported that Trump is currently in the process of attempting to identify a legal justification for the ban of DeepSeek in the United States.
Nevertheless, China is not experiencing any deceleration. Just a few days ago, China released the new Qwen 2.5 model, another “ChatGPT killer.” China is also making significant investments in AI, with a $138 billion national plan that is expected to compete with the $500 billion US Stargate initiative. The state-sponsored initiative aims to aid Baidu, ByteDance, and Alibaba in the creation of advanced AI systems. The recent disclosure of DeepSeek’s open-source model has further exacerbated the competition with Western AI titans. China is accelerating the AI arms race by rapidly developing AI accelerators and finding methods to circumvent export restrictions, despite sanctions.
This month, China has also made significant strides in other technological disciplines, in addition to AI. Chang Guang Satellite Technology, China’s response to SpaceX’s Starlink, has achieved a record-breaking 100 Gbps satellite-to-Earth data transmission speed. At the same time, the EAST tokamak, which lacks an American equivalent, maintained temperatures exceeding 100 million degrees for 1,066 seconds, beating its own 2023 record of 403 seconds. Plasma power plants could potentially disrupt the demand for oil, gas, and even renewable energy as a result of this innovation.
Therefore, this event can be appropriately referred to as a “Chinese Black Swan” for Silicon Valley, as it signifies the start of an AI conflict between the United States and China.