China’s DeepSeek: Why is it disrupting the market and hurting Nvidia?

The fierce optimism of the markets at the start of 2025 has been shaken by an unforeseen factor — one that almost no one had anticipated. The rise of the Chinese artificial intelligence startup DeepSeek triggered a stock market earthquake on Monday, which could undermine the foundations of the robust rally that equities have enjoyed since 2023. This rally has been largely driven by the artificial intelligence (AI) boom, which has emerged as a major economic catalyst.

Dubbed the “next internet,” AI is poised to transform society while boosting corporate productivity and profitability. Goldman Sachs, for instance, estimated a 1.5% improvement in earnings per share for S&P 500 companies due to AI advancements. This optimism has been irresistible to investors, yet it has primarily materialized in the valuations of a select few companies: the tech giants with the financial muscle to invest heavily in AI development, and the chip designers, manufacturers, and assemblers, led by U.S.-based Nvidia.

DeepSeek’s ability to launch a free, open-source app that competes directly with OpenAI’s ChatGPT — but with significantly lower investment — has upended the prevailing market narrative. Until now, the AI race was seen as a battle of scale and brute force: massive investments to process vast amounts of data, with the expectation that results would inevitably follow. DeepSeek’s success challenges this assumption, raising questions about the sustainability of such a strategy.

Could DeepSeek be the black swan of the markets this year? While it’s still too early to tell, its launch has already had seismic effects. Nvidia, until recently the world’s most valuable company, saw $589 billion wiped off its market capitalization in a single trading session. The ripple effects have spread across global stock exchanges, impacting companies of all kinds.

Unlimited investment

Last year, investments in AI infrastructure — including cloud computing, chips, and related technologies — exceeded $250 billion. This year, the race has intensified even further: Microsoft has announced investments of $76 billion, Meta plans to allocate another $62 billion and Amazon and Google are committing $42 billion and $24 billion, respectively. Adding to this, a new alliance formed by OpenAI, Oracle, and SoftBank aims to invest up to $500 billion. This initiative, sponsored by the newly elected U.S. President Donald Trump, will contribute a fifth of the total investment this year alone. And this is just the beginning — BlackRock estimates that, in the long term, the investment will exceed $1 trillion. Meanwhile, DeepSeek has achieved remarkable efficiency, training its model in just two months at a cost of under $6 million.

Oliver Blackbourn from Janus Henderson points out that “the emergence of a potentially more efficient approach to AI processing calls into question the need for the billions of dollars being poured into infrastructure and intellectual property.” DeepSeek’s success challenges the paradigm of recent years, which has been built on massive investments by tech giants to control the advancement of artificial intelligence — and, ultimately, to reap the rewards of these investments.

Now, these companies face a double-edged sword: while AI development may become cheaper, it will also be harder for them to maintain their dominance in the market.

The Nvidia phenomenon

If any company has benefited from the AI boom, it is undoubtedly Nvidia. Initially designed for graphics cards aimed at video game users — and therefore optimized to handle multiple tasks in parallel — the company’s chips have proven to be ideal for training artificial intelligence models. This has made Nvidia an indispensable partner for major tech companies. According to Moody’s, last year Nvidia’s high-performance processing units (GPUs) were used 11 times more than chips from competitors like Intel or Advanced Micro Devices.

Nvidia’s stock price reflects this success. Starting 2023 at less than $15 per share, it soared to $150, becoming the most valuable company in the world and even surpassing Apple and Microsoft at times. This surge was fueled by an overwhelming backlog of orders. While the company closed 2022 with revenues of $26 billion, analysts project it will reach $60 billion by the end of 2024 and approach $200 billion by 2026.

Despite these impressive figures, analyst firms continue to raise their valuation targets for Nvidia, even as they scrutinize its financial performance more rigorously with each passing quarter. The company is scheduled to announce its next results on February 26.

Chipmakers on the rise

The surge in interest around Nvidia is far from isolated. Other chip and computer equipment manufacturers have also capitalized on the euphoria surrounding artificial intelligence. Supermicro, for instance, has seen its valuation skyrocket by 300% since 2023, peaking at a staggering 1,300% gain. AMD has risen 92%, while Broadcom has climbed 340%. However, not all companies have joined this rush; Intel, for example, has struggled, losing 54% of its market value over the past 12 months.

The driving force on the stock market

The launch of ChatGPT in November 2022 marked a turning point, with artificial intelligence becoming a dominant catalyst for the U.S. stock market and influencing global markets by extension. Over the past two years, the S&P 500 has climbed nearly 60%, fueled by the “Magnificent Seven” — seven leading tech giants: Apple, Microsoft, Alphabet, Amazon, Nvidia, Meta, and Tesla. These companies, all heavily invested in AI, now account for 34% of the index’s total weight.

In 2024, these seven stocks contributed half of the S&P 500′s rise, matching the combined contribution of the remaining 495 stocks. This strong performance has driven up valuations — the price paid for stocks relative to company earnings — yet U.S. equities remain a top choice for investors this year.

Blackbourn, portfolio manager at Janus Henderson, warns of potential risks: “If we start to see U.S. equity valuations fall significantly, there is a danger that this could spread to other high-valuation regions in Europe and Asia.” He further highlights the risk of a domino effect: “With U.S. consumers more exposed than ever to stock markets, a loss of consumer confidence could trigger broader negative feedback loops. A significant drop in financial indicators due to stock market losses might prompt the Federal Reserve to reassess its outlook more quickly.”

How will DeepSeek’s success impact the industry?

Analysts at investment firm Jefferies suggest that competitors in the AI space could adopt two potential strategies moving forward: pursue greater computing power to accelerate model improvements further, and focus on efficiency and return on investment (ROI), which could reduce the demand for computing power starting in 2026. That, they said, could put U.S. executives “under more pressure to justify raising AI capex.”

Moody’s recently highlighted a key challenge: “AI expertise will remain primarily in the hands of a few dominant players.” The massive infrastructure, vertical integration, and financial power of tech giants create significant barriers for smaller competitors and independent labs trying to enter the market. However, if a startup demonstrates OpenAI-level success with far less investment, these assumptions could be upended.

As Jefferies points out: “Re-evaluating computing power needs could cause 2026 Al capex to fall (or not grow).”

A new battlefront between the US and China

With Europe largely out of the race — represented only by France’s Mistral — the competition to dominate AI is now a two-way battle between the Silicon Valley-Washington axis and Beijing. Both the United States and China, as Moody’s notes, “view AI as a strategic asset and are investing heavily in research, infrastructure, and talent acquisition to secure technological dominance.” This rivalry has escalated to the point where the U.S. banned the export of advanced chips and chip manufacturing equipment to China.

According to the ratings agency, “by restricting global supply and demand, these bans have pushed companies to explore cost-effective AI solutions, such as smaller, optimized models that reduce reliance on expensive hardware. In China, DeepSeek’s V3 model was trained using surprisingly smaller hardware, proving that better software can outperform better hardware.”

Chinese strategy?

The recent launch of DeepSeek R1 is not an isolated event. In just a few weeks, two other Chinese AI projects have entered the market: Bytedance’s Doubao-1.5 Pro and Moonshot’s Kimi k1.5. All three offer performance comparable to their international counterparts but at a significantly lower computational cost. Analysts at Jefferies point out that China’s chip limitations have forced the country to prioritize efficiency. As a result, they suggest that Trump could “relax the AI diffusion policy” to make it easier for U.S. firms to compete.

Sign up for our weekly newsletter to get more English-language news coverage from EL PAÍS USA Edition