•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
DeepSeek’s newly released AI model, V4, has drawn a relatively muted market response, contrasting with the earlier hype that followed the company’s low-cost training models. The subdued reaction to the V4 preview suggests that the factors that previously drove investor excitement around DeepSeek are no longer as influential.
DeepSeek’s earlier models—DeepSeek-V3 and DeepSeek-R1—sparked global enthusiasm by demonstrating that a powerful large language model (LLM) could be built with modest costs. The introduction of DeepSeek-V3 at the end of January 2025 was particularly disruptive, triggering a wave of tech-stock selling worldwide. Nvidia’s market value fell by nearly $600 billion, and Meta reportedly set up a “war room” to understand how the model could reshape the competitive landscape in China.
Analysts described that moment as a “black swan” for the AI development cycle. The shock forced markets to reassess assumptions about AI costs, competition, and the pace of technological innovation in China under US chip restrictions.
Omdia’s chief analyst, Lian Jye Su, said the market reaction is “fairly predictable,” noting that investors have grown accustomed to low-cost, high-performance models developed under constrained computing resources.
Artificial Analysis data suggests that DeepSeek-V4 Pro delivers clear performance improvements over earlier generations, but does not create a significant overall gap versus competitors.
Analysts also point to the changing market context. Su said the earlier boom around DeepSeek reflected converging factors: inflated valuations of US tech firms, expectations that a small number of “giants” would dominate, and the sudden emergence of a little-known Chinese startup producing results beyond expectations. Those conditions, Su said, are no longer present, and the market is now more realistic about both the potential and limits of AI.
Domestic competition in China has also intensified. Numerous firms continue to release increasingly powerful models, which analysts say has eroded DeepSeek’s relative advantage.
Despite the tempered reaction to V4, the wider AI landscape remains vibrant. Earlier this week, stock markets in Korea and Taiwan reached fresh highs, supported by optimism around AI-related equities.
According to Alfredo Montufar-Helu, CEO of Ankura China Advisors, the importance of V4 is less about winning a US-China AI race and more about how the model affects market dynamics. He said DeepSeek optimized V4 to run most efficiently on Huawei chips, in the context of US export controls designed to limit China’s access to advanced chips used to train large AI models.
“The factors that impressed were present last year and are already reflected in prices. What matters now is whether China can sustain AI development and whether it can achieve chip autonomy. These are geopolitically meaningful considerations,”
Experts broadly agree that DeepSeek has moved into a steadier phase rather than delivering the kind of breakthrough initially hoped for. Even so, DeepSeek’s contribution is still viewed as highly significant, with analysts noting that its progress has encouraged similar AI models in China and advanced work across research, engineering, and applications—citing examples including Alibaba’s Qwen, ByteDance’s Moonshot AI, and Zhipu AI’s GLM/ChatGLM.
Premium gym chains are entering a “golden era” that is ending or already in decline, as rising operating costs collide with shifting consumer preferences toward more flexible, community-based ways to exercise. Long-term memberships are shrinking, margins are pressured by higher rents and facility expenses, and competition from smaller, more personalized…
