The emergence of DeepSeek has marked a significant turning point in the realm of artificial intelligence, primarily owing to its exceptional performance coupled with an economical priceAs of February 14, a plethora of enterprises has integrated the DeepSeek R1 model, showcasing a diverse range of utilizers from cloud service providers to applications developers and hardware manufacturersNotably, major tech companies that are pivotal players in the AI large model landscape are also jumping on the bandwagon—iFlytek has made the full suite of DeepSeek's models available on its open platform, while Baidu's Wenxin Yiyan has announced the integration with the DeepSeek R1 model, which will be freely accessible starting April 1. Furthermore, Tencent’s Yuanshi model now supports both its proprietary model and DeepSeek R1, and ByteDance's Feishu is also leveraging the capabilities of DeepSeek R1.

However, this surge of interest in DeepSeek is somewhat overshadowing the so-called "AI Six Dragons," a term used to refer to a cohort of standout startups that have flourished over the recent years due to the explosion of AI-generated technologies and models, including the likes of Zhipu AI, Kimi, Baichuan Intelligence, Jiyue Star, MiniMax, and Zero One UniverseThese companies, driven by esteemed backgrounds, have successfully captured the attention of venture capital firms and secured large clienteleWith market capitalizations surpassing 20 billion yuan after multiple rounds of funding in 2024, the pressure is now on these so-called "dragons" to reply to the shifting market dynamics and demonstrate how they will maintain their competitive edge amidst the growing prominent presence of DeepSeek.

Founded by the private quantitative investment firm Huansuan Quant in July 2023, DeepSeek initially flew under the radar, being recognized primarily for its investment toolsThe introduction of the DeepSeek V3 open-source model and the DeepSeek R1 model in December 2024 and January 2025, respectively, brought it into the limelight

Advertisements

The R1 model has shown impressive capabilities in mathematical problems, coding, and natural language inference, rivaling that of OpenAI's offerings, with the efficiency cost of training reduced by nearly 20 times compared to the GPT-4 seriesWhat sets DeepSeek apart from the Six Dragons, however, is its lack of dependency on external capital; the company has not undertaken any outside financing, which provides it a layer of security that these competing unicorns may not have.

According to Gong Bin, founding partner of the Wisdom Consulting Technology Think Tank, DeepSeek's market valuation could potentially eclipse the total worth of the AI Six Dragons collectivelyBloomberg Billionaires Index, based on various research data, suggests that DeepSeek's valuation ranges between 10 billion and 150 billion dollars, with numerous investment banks projecting figures that surpass 100 billion—peaking at the staggering figure of 150 billion, nearly half of OpenAI's latest valuationAs of February 8, DeepSeek boasted an impressive daily active user count of approximately 119 million.

This wave of new competition isn’t welcome news for the AI Six DragonsNoted economist Pan Helin has indicated that the free and open-source nature of DeepSeek, coupled with its OpenAI-like performance, significantly alters the existing business strategies in the model sectorWhat was once a hope to monetize through user subscriptions is now becoming an unattainable dream as companies feel compelled to switch to free models to remain competitiveAdditionally, more businesses are considering deploying the open-source model for self-development in AI, further illustrated by numerous application vendors and cloud service providers integrating DeepSeek's offerings.

The winds of change are undoubtedly blowing through the AI fieldDeepSeek's model represents a disruptive force, demonstrating the feasibility of utilizing pure reinforcement learning for model training, unlike the prevalent supervised fine-tuning practices

Advertisements

By setting reward mechanisms rather than depending on human-annotated data, the model evolves through continuous trial and error, developing critical capabilities such as reflective reasoning and decision-making over timeWith a reported 60% reduction in ineffective training and a 40% decrease in parallel processing demands compared to traditional architectures, DeepSeek holds promising possibilities for enhancing domestic chip performance and efficiencyWhile challenges in computational dependence and generalization remain, DeepSeek's innovative model is paving the way for a brighter future for domestic enterprises, possibly reducing reliance on imported chips.

In the wake of this 20 billion yuan valuation halo, the scramble among the AI Six Dragons to respond to the DeepSeek phenomenon has garnered much attentionWhen approached for insights regarding changes in product data and strategies in light of DeepSeek’s recent rise, several companies, including Baichuan Intelligence and Zhipu AI, declined interviews due to the sensitivity of the informationHowever, Kimi's co-founder, Zhang Yutong, revealed via social media that their new version saw remarkable user growth during the Spring Festival, reflecting the rapid pace of AI advancement during that period, although precise numbers remain undisclosed.

Insiders in the investment sector highlight the evolution of AI Six Dragons who have enjoyed substantial attention and support from capital markets over the past two yearsBy achieving an upward valuation trajectory past the 20 billion mark, they face a looming pressure to deliver solid returnsDue to high development costs and unclear paths to profitability, they might struggle to meet the capital markets’ expectationsDeepSeek's rapid rise with its economical and open-source approach creates an imperative for the Six Dragons to recalibrate strategies surrounding technology, market positioning, and fundraising.

Investigating recent financing events corroborated with data reviewed reveals that all of the AI Six Dragons have been active in securing funding throughout 2024. Zhipu AI secured three rounds of funding in June, September, and December, netting a total of 30 billion yuan, while Kimi saw funding rounds in February and August

Advertisements

Advertisements

Advertisements