AI Investments and Market Volatility: The DeepSeek Effect

Equity Research Team,  Louis-Philippe Gagnon, Gabriel Houde, Zineb Bourchich,

Driven by the rise of artificial intelligence, the stock market continues to reach new all-time highs. Indeed, investors estimate AI is worth the hype, leading many stocks to trade at a premium. The anticipated massive future growth has prompted tech companies to invest heavily in AI development. OpenAI is currently leading the race with ChatGPT and recently launched a new project, StarGate, to build new AI infrastructure over the next 4 years at an estimated cost of $500 billion. Magnificent 7 companies are all involved in the AI revolution and deploy significant capital on AI as their future growth relies heavily on these investments.

However, since valuations are high because of massive potential growth related to AI and political uncertainties tied to Trump actions bring uncertainty in the market, a small bump in the road could disrupt investor trust and lead to a selloff.  In fact, we saw that reaction when a Chinese tech company, DeepSeek announced their development of a generative AI model for a fraction of the cost with less advanced chips. Deepseek’s app quickly became the number one download on the AppStore, raising questions about US tech dominance and the relevance of massive investments. This resulted in a shock for the market, a tech selloff and a sharp increase in volatility.

DeepSeek, a private Chinese based artificial intelligence software startup created in 2023, is owned and funded by Chinese hedge fund High-Flyer. The company quickly became competitive with the development of a generative AI model for much cheaper than its peers. Nvidia chips are considered the more powerful in the current market. However, since DeepSeek is a Chinese company, the most competitive chips are restricted by US regulation on exports. Nonetheless, they have currently access to less efficient Nvidia chips but still excellent chips that are fully export control compliant. The model is highly efficient because of open-source availability, knowledge distillation and lower memory usage, which enabled Deepseek to use less chips, and thus make it less costly.

According to an industry expert, the Chinese startup did not use innovations that are unknown or secret, their approach was something everyone has been experimenting with. It was a shock for the market simply because their software is allegedly low-cost while big tech companies are spending billions of dollars for a similar product. Although their financials are not available, if their claims are true, this would be a breakthrough for the AI industry.

On the day of the news, Nvidia stock soared 17% wiping nearly 600 billion dollars off its market capitalization, the largest market cap loss in US history for a single day of trading. Other tech companies were also affected, and the tech-heavy Nasdaq Composite closed 3,1% lower. Tech giants, mainly the magnificent 7, do not seem to be bothered by the low-cost model in DeepSeek. Meta, Microsoft, and Google parent Alphabet plan to invest even more in 2025, expecting a cumulative $325 billion in capital expenditures in 2025, driven by their investments in artificial intelligence infrastructure. That is a 46% increase from the roughly $223 billion those companies reported spending in 2024.

Innovation and efficiency have always been part of AI’s history, and DeepSeek’s breakthrough is no exception. Despite the market’s reaction, experts and stakeholders have nuanced views on the matter. The success of DeepSeek can lead to higher demand for AI services, a direct consequence of the cost decrease in their implementation, which in turn, can generate significantly more revenue for chip companies like NVIDIA, Broadcom, AMD, and Micron Technology. On the other hand, the implications raise concerns for data centers. The performance attained by DeepSeek implies increased efficiency in current data centers and a slowdown in their growth needs. This could lead to increasing borrowing costs for data centers, as lenders seek to protect themselves against the risk of obsolescence and an oversupply of assets. The private equity giant Blackstone with $80 billion invested in data center operators remains confident, claiming that its investments are tied to existing demand and that while the amount of power needed to answer a specific question may decrease, people will ask more questions.

Companies like Amazon and Apple are also likely to benefit from the low-cost, high-quality, open-source Chinese model in the long run, while Google may be in a more precarious position, as chatbots could potentially replace the Google search engine over time.

DeepSeek has also sparked security concerns around the globe; it could be used by the Chinese government to gather user data on Chinese servers or sway opinions through its outputs. The level of vigilance is especially high in North America, Taiwan, Italy and Australia. The Chinese chatbot is banned on some Canadian government mobile devices due to “serious privacy concerns,” while Washington is pushing a similar bill. This localized ban could quickly turn into a nationwide restriction in the U.S., similar to what is currently seen in Taiwan and Italy. Given the colossal political stakes related to the AI race, this is even more likely considering TikTok experienced a similar process before considerations for a countrywide ban in the U.S. However, due to the open-source nature of DeepSeek and the fact that the startup has published most of its breakthroughs online, some express pessimism regarding a significantly slower growth and popularity due to these restrictions.

As the market attest, DeepSeek’s debut marks a shift in the AI landscape, making China a more serious competitor than ever before and pushing the United States to reevaluate both the efficiency of its restrictions on China and the viability of the massive AI investments that have driven stock market performance over the past two years. Looking ahead, AI infrastructure spending will serve as a clear indicator of the response from tech giants. Additionally, several countries, including France and South Korea, are set to inquire further into DeepSeek’s management of personal data, while suspicions remain regarding the transparency of the Chinese startup’s statements on the true cost of training and intellectual property.

Equity Research Team,  Louis-Philippe Gagnon, Gabriel Houde, Zineb Bourchich