In the years 2023 and 2024, the global artificial intelligence landscape witnessed a major upheaval, primarily driven by advancements in large language models like GPT-3 and GPT-4. The surge in technological innovation and the corresponding influx of capital created an exhilarating atmosphere in the industryYet, amid this rapid evolution, an unexpected contender emerged on the scene, a newcomer named DeepSeek, which managed to disrupt the established norms of the AI domain and sent shockwaves through the international capital markets.

As the trading day closed on the eve of the "Year of the Dragon" in the Chinese lunar calendar, DeepSeek's influence transcended borders, penetrating the US stock marketsCompanies linked closely to AI computing power saw their stock prices plummet significantlyNotably, Broadcom (AVGO) and Taiwan Semiconductor Manufacturing Company (TSM) faced a staggering decline of 9% in pre-market trading, while NVIDIA (NVDA) and Constellation Energy's stock prices tumbled by over 11%.

At the heart of this tumult was the AI model DeepSeek, developed by Hangzhou’s Depth AI Research Co., LtdIt burst into the scene as a new star, boasting an astonishingly low training cost—less than 1/20th of that of its counterparts—and remarkable computational efficiency, achieved through the utilization of a cluster of 2048 H100 GPUs over a mere 53 daysThis radical shift in the traditional AI model-building paradigm swiftly attracted global attention.

One of DeepSeek’s most striking characteristics is its unparalleled cost-effectivenessBuilding and training traditional large AI models, such as GPT-4, often incurs costs in the range of hundreds of millions of dollarsIn stark contrast, DeepSeek has harnessed innovative technologies to slash these costs dramaticallyFor instance, its API pricing is significantly lower than that of its competitors, enabling developers to iterate on model features at a fraction of the usual expense.

Moreover, DeepSeek's open-source model serves as a potent weapon in its arsenal

Advertisements

Unlike traditional AI corporations, which often keep their technologies proprietary, DeepSeek not only shares its source code but also invites developers to engage with its platform by offering affordable API ratesThis approach has effectively built an efficient and cost-effective technological bridge for AI developers around the world, allowing DeepSeek to rapidly establish a foothold in the international AI supply chain.

The rise of DeepSeek has significantly increased the pressure on AI computing power-related enterprisesNVIDIA, long regarded as the leader in the global AI computing market, relies heavily on the sales of its GPUs for AI training and inferenceHowever, with DeepSeek's breakthrough in low-cost operations, demand for NVIDIA's chips is now under scrutinyAnalysts have drawn parallels between NVIDIA's current trajectory and Cisco's fortunes before the burst of the dot-com bubble in 2000, raising concerns that DeepSeek might eat into NVIDIA's market share.

Other players like Broadcom and Taiwan Semiconductor have not escaped unscathed eitherOnce hailed as titans in the AI field, Broadcom’s diverse hardware and software strategies now face skepticism, reflected in the dramatic stock price drop indicative of investor panicEven Constellation Energy, initially seen as a reliable provider of power for AI training, has witnessed a substantial drop in stock value as market reliance on nuclear power diminishes with DeepSeek's rise.

The emergence of DeepSeek has sparked a bifurcation in the capital marketsOn one hand, its low-cost, high-efficiency model has garnered significant investor enthusiasm, with Microsoft CEO Satya Nadella praising DeepSeek for its stellar inferential capabilities and supreme computational efficiencyInfluential investor Marc Andreessen also deemed it "one of the most astonishing breakthroughs."

Conversely, skepticism lingers in the market

Advertisements

Some investors suspect that DeepSeek's success might be ephemeral and view its low-cost training claims with cautionCertain Wall Street investors have even warned that DeepSeek could exert a "destructive" impact on the US stock market, potentially undermining investor confidence in the AI sector.

DeepSeek’s rise heralds a significant transformation within the AI industry, tearing down previous technological barriers and introducing profound changes to the landscapeIts cost-effective approach is likely to expedite the proliferation of AI applications, opening new doors for small and medium-sized enterprises that previously could not afford to engage in AI technologies.

Furthermore, its open-source paradigm promises to lead the AI industry toward a more transparent and accessible futureWith the participation of a broader range of developers, DeepSeek could evolve into a platform for tech sharing, accelerating innovationThis shift could signal a competitive landscape where AI dominance is determined by technological advancements and application capabilities rather than merely financial resources and computational power.

On January 26, the producer of the game "Black Myth: Wukong," Feng Ji, took to social media to laude DeepSeek as a "national-level technological achievement." His commendation derived from a series of breakthroughs within the realm of AI large models, which are critical to DeepSeek's rising prominence.

The unprecedented arrival of DeepSeek signifies a new phase in the AI industryIt challenges conventional paradigms with its low-cost and high-performance offerings, igniting tremors across the capital marketAlthough doubts linger regarding its sustainability, it is undeniable that DeepSeek presents fresh opportunities and challenges for the global AI industry.

As technology continues to advance, DeepSeek is poised to emerge as a key player propelling the AI sector into a realm characterized by openness, affordability, and efficiency.

Advertisements

Advertisements

Advertisements