The week has begun with the tech sector reeling. On Monday, Nvidia led a market slump—dropping as much as 17%—triggered by the strong performance of the low-cost generative AI assistant developed by Chinese company DeepSeek. According to experts, the emergence of a potentially more efficient approach to AI processing—reducing model training costs by 86%—raises questions about the necessity of the billions of dollars planned for infrastructure and intellectual property investment.
As a result, the S&P 500 index dropped 1.5%, and Nvidia‘s decline—the largest single-day market capitalization loss for the company at $589 billion—dragged the Nasdaq Composite down 3.1%. “The emergence of the new competitor has primarily impacted the entire data center value chain. This includes chip manufacturing equipment makers like ASML (-7.2%), high-performance chip manufacturers (Nvidia and Broadcom (-17.4%)), as well as companies specializing in energy infrastructure like Schneider Electric (-9.6%) or the real estate side of data centers like Digital Realty (-8.7%),” explain analysts at Banca March.
What explains these movements? In recent days, the generative AI assistant developed by DeepSeek has become the most downloaded app for iPhone, surpassing the popular ChatGPT application from OpenAI. Nvidia‘s drop has been the most visible consequence of this shift, driven by fears about the impact DeepSeek could have on the demand for high-end microchips.
“DeepSeek could be a seismic shift for the AI industry. If its advancements hold true, model training costs would drastically decrease, changing the game for everyone,” says Víctor Alvargonzález, founder of Nextep Finance. In his view, one of the main reasons behind Wall Street’s recent sell-off—particularly Nvidia‘s worst-ever trading session in U.S. stock market history—is DeepSeek‘s promise to reduce algorithm training costs. Estimates suggest that training costs could drop from the current $50 million per model to just $7 million or less, thanks to process simplification and a 75% reduction in memory requirements.
Amid these declines, Louise Dudley, portfolio manager of global equities at Federated Hermes Limited, believes there are still many questions left unanswered. “For Nvidia, as a key supplier of premium chips worldwide, the concern is whether companies will need fewer chips in the future. However, the company responded to the news by highlighting ‘excellent progress,’ signaling optimism about ongoing AI model developments, which are still in their relative infancy.
For companies involved in building data centers, the short-term impact is likely to be significant, as demand has been very strong. The new DeepSeek model code will be reviewed for potential performance improvements. Existing projects under development are at risk, and this will be a key focus for investors. This news will likely increase both corporate and consumer appetite for AI tools, given improved accessibility, leveraging this innovation and accelerating AI adoption timelines,” Dudley points out.
Market Reactions and Expert Insights
According to Hyunho Sohn, portfolio manager of the Fidelity Funds Global Technology Fund, Chinese AI startup DeepSeek has introduced AI models that perform comparably to OpenAI’s ChatGPT models while being significantly more cost-effective. “This efficiency advantage has raised a series of questions about the perceived ‘winners’ in the global AI ecosystem, the implications for hyperscaler capital expenditures, and the effectiveness of sanctions and export bans aimed at preventing high-level generative AI progress in China.
This is an evolving situation, and we may see short-term volatility until it becomes clear how much more efficient this technology really is. While broader implications must be assessed on a case-by-case basis, I generally believe this development will be deflationary,” Sohn states.
Despite the shockwaves, Fidelity’s portfolio manager believes this is ultimately beneficial for end-users and service providers, though it could have negative implications for hardware. “This is similar to what we saw in the early days of the internet when people vastly underestimated the scale of innovation, technological adoption, and service-based business potential, while greatly overestimating the total addressable market (TAM) for hardware,” explains Sohn.
In the view of Amadeo Alentorn, manager of the Global Equity Absolute Return fund and head of the systematic equity team at Jupiter AM, DeepSeek’s rise is part of a broader trend that has been developing for months. In recent times, there have been major advances in Small Language Models (SLM), which contrast with the large models used by companies like OpenAI. The central question has been whether it is possible to build more precise, specialized models that focus on specific areas, such as law or medicine, rather than encompassing all knowledge.
“So far, the rise of artificial intelligence has primarily benefited a small group of large companies. However, recent advancements suggest that we may be witnessing a paradigm shift, where smaller companies can also leverage this technology without needing massive infrastructure investments. Identifying which companies will lead this new AI phase is a complex task, but what is clear is that this evolution promotes diversification within the sector. AI could expand beyond tech giants and create new business opportunities across various industries,” Alentorn asserts.
High Valuations Under Scrutiny
In this context, Fidelity’s portfolio manager acknowledges that, as he has been saying for some time, many AI semiconductors are expensive, with sentiment, valuations, and momentum slowing down—“the most interesting opportunities lie within the services ecosystem.”
“It’s still early, but I would add that the rapid developments in generative AI highlight the need for proximity and connection throughout the tech ecosystem—something we are well-positioned for, given the breadth and depth of our research coverage,” Sohn adds.
For Oliver Blackbourn, portfolio manager in the Multi-Asset team at Janus Henderson, AI has long been considered a highly complex area of development, with industry leaders perceived as having technological advantages that would allow them to maintain rapid growth. In his view, the expectation of high earnings growth has been used to justify elevated valuations, making these stocks highly vulnerable to any disappointment.
“Competition always seemed like the biggest threat, but also the hardest to assess for investors. The market’s reaction to a perceived radical shift in the competitive landscape has been fierce. Before U.S. markets opened today, Nasdaq 100 futures had fallen 3.9%, and ASML—one of Europe’s companies most exposed to the AI theme—had dropped more than 10%,” Blackbourn notes.
In his opinion, while it is easy to get ahead of events, it is also important to remember that high expectations have driven up valuations across the U.S. stock market and, consequently, global equities. “If we start to see U.S. stock valuations drop significantly, there is a risk that this will spill over into other high-valuation areas in Europe and Asia.
Similarly, with U.S. consumers more exposed than ever to the stock market, there is a broader risk of negative feedback loops if consumer confidence is shaken. A significant tightening of financial conditions due to stock market losses could quickly change the Federal Reserve’s outlook,” Blackbourn concludes.
