In recent events, NVIDIA's stock has seen a significant downturn, primarily due to the emergence of a new AI model called DeepSeekThis model has garnered attention for its impressive capabilities, seemingly on par with well-established models like OpenAI's ChatGPT and Meta's Llama 3.1. Users have reported that DeepSeek handles conversational tasks and questions with remarkable ease.

Perhaps the most striking aspect of DeepSeek relates to its hardware configuration; it utilizes Huawei's Ascend 910C chips in partThis raises important questions about NVIDIA's future in a market that seems poised to shift in favor of alternatives.

Just how seismic is the impact DeepSeek is having on the AI landscape? Could it potentially dethrone NVIDIA from its dominant position?

The turmoil in the US stock market due to Chinese AI upheavals is hard to ignoreOn January 27, 2025, NVIDIA's share price plummeted by 17%, leading to a rapid evaporation of approximately $590 billion in market capitalization

Advertisements

This decline is roughly equivalent to the combined market value of Alibaba and half of Tencent or even the total value of AMD and Intel combinedSuch a loss is catastrophic for NVIDIA.

The fallout has positioned NVIDIA as the third-largest company globally by market capitalization, following in the footsteps of Apple and MicrosoftMeanwhile, CEO Jensen Huang saw a staggering personal loss of $20 billion in just one day.

Moreover, NVIDIA was not alone; many American companies felt the heat as wellMicrosoft, Google, Meta, and Amazon all experienced declines in stock value between 4-6%. The total market loss encompasses more than a trillion dollars, and the collective wealth of several individual billionaires shrunk by about $108 billion.

In light of these staggering losses, even the President of the United States commented that "the rise of DeepSeek should sound alarm bells for American businesses." This illustrates the considerable impact that this new AI model has had on the U.S

Advertisements

tech ecosystem.

But what exactly is propelling DeepSeek to such heights? How could a new entrant disrupt a well-established market?

DeepSeek is developed by a company named Deep Pursuit, which only came into existence less than a year agoAfter setting up operations in Hangzhou and developing its technology, they launched the DeepSeek model just half a month ago, and it quickly became a phenomenonParticularly in the U.SApp Store, DeepSeek surged to the top position, reminiscent of the explosive growth seen by earlier platforms like Xiaohongshu, or Little Red Book.

The reception from users has been overwhelmingly positive, largely because DeepSeek excels in addressing a myriad of issuesBeyond its capabilities in conversation and inquiry, the model has proven proficient in creative writing, including the continuation of popular Chinese literary works like "Dream of the Red Chamber" and "Sword Comes."

Some users have gone so far as to request that DeepSeek write narratives in the style of revered Chinese author Jin Yong, known for his complex martial arts stories

Advertisements

DeepSeek has impressively risen to challenge, managing to encompass vivid elements such as heroism, faction conflict, suspense, and nuanced character arcs in its creative outputsWhether crafting beginnings or endings, or building intricate world settings, it covers an extensive array of storytelling techniques, suggesting it has reached the competency level of a median-quality author.

So why is DeepSeek considered disruptive? At this juncture, one might wonder why a strong performance isn’t the end of the discussionThe crux of it lies in one word: cost-efficiencyThis cost advantage challenges the traditional paradigms associated with training AI models.

Historically, significant investment in numerous processing units was required to train AI systems effectivelyFor instance, models like GPT-4 produce hundreds of terabytes of data during their trainingConsequently, the demand for a plethora of high-performance chips was inevitable.

Elon Musk has publicly shared that his AI startup, xAI, used around 100,000 NVIDIA H100 chips for training their Grok-3 model

• Nvidia Loses Nearly $600 Billion in Market Value
• Surge in Over-the-Counter Trading in the U.S.
• Dollar Crisis Looming
• Europe Lowers Interest Rate Expectations
• Surge in Foreign Rare Earth Mineral Production!

This illustrates the necessity of substantial hardware resources in the AI fieldNVIDIA's H100 chips carry immense price tags, typically ranging from $30,000 to $40,000 each—an expense comparable to purchasing a carThis expense structure is one reason why training AI has historically been so expensiveFigures from OpenAI suggest that training GPT-4 incurred around $100 million in costs, while Google reportedly spent $190 million on their Gemini Ultra model.

Unsurprisingly, this substantial chip demand has made NVIDIA extremely wealthy, previously pushing its market cap to $3.6 trillion, making it the most valuable company globallyIt underscores the perception of AI as an insatiable financial behemoth.

Moreover, NVIDIA has maintained a dominant position among global chip suppliers, consequently dominating the AI marketplace while overshadowing competitors like Intel and AMD.

However, the onset of DeepSeek’s meteoric rise may signify a shift in NVIDIA's fortunes

Reports indicate that training DeepSeek requires only a few million dollars and only 1/20th of the computing power needed for GPT-4. A model achieving similar performance with such relatively low overhead raises serious questions about the economic viability of previous substantial investments in traditional chip-heavy platforms.

Clearly, the dynamics of AI development are shifting; the notion that endless chips must underpin AI advancements is being called into questionAs the pattern of chip consumption evolves, NVIDIA may face challenges in maintaining its high market value.

It’s essential to note that while the emergence of DeepSeek heralds significant changes in training efficiencies, NVIDIA’s decline is not certainThe demand for chips across various sectors will likely remain robustThis phenomenon recalls the historical improvements in internal combustion engines

Although the thermal efficiency of engines has vastly increased from a mere 4% to around 40%, this improved efficiency hasn't necessarily led to reduced oil consumption; rather, demand has continued to riseThus, improved efficiencies do not equate to reduced needs.

When applied to AI chips, the same logic applies: though efficiency has risen, the overarching demand for such chips will continue to escalate as AI technology advances.

To illustrate, the AI chip market in China is booming and projected to reach 144.7 billion yuan by the end of 2024. This places AI chips in a position of industry significance that rivals traditional automotive markets, asserting their status as emerging players.

The story does not end here, as China actively stakes its claim in this lucrative fieldDomestic companies like Huawei and Cambricon are offering robust alternatives to NVIDIA’s offerings, with experts suggesting that two Chinese AI chips can effectively replace one NVIDIA A100 chip.

A vital advantage of domestic AI chips is the accompanying real-time guidance and support from engineers during implementation—something DeepSeek is benefitting from with Huawei's engineers involved during its training.

Thus, the influence of Chinese AI continues to grow, with multiple competing models such as ByteDance's Doubao, Baidu's Wenxin Yiyan, and Alibaba’s Qwen2.5-Max emerging to challenge the current landscape

Each of these models not only strives to catch up in performance but also to reduce costs, which could foreseeably position Chinese AI as a dominant force globally.

Meanwhile, there's been an ongoing dialogue about lapses and ethical dilemmas in the U.SAI sceneRecent incidents involving Stanford University students who produced a model called Llama 3-V showcase some troubling tendencies for imitation in the fieldInitially hailed for competitive performance against GPT-4 with a paltry training expense of $500, the claims of Llama 3-V soon crumbled under scrutiny, revealing similarities to a Chinese product instead.

Ultimately, these instances illustrate a broader truth: even the titans of American industry occasionally falter, and the swift rise of Chinese AI models reinforces the rapid growth and innovation in that sectorAs the success of DeepSeek and similar products continues, the future of AI may bring even more surprises.