JAKARTA – Chinese AI startup DeepSeek is expediting the release of its next-generation artificial intelligence (AI) model, R2, earlier than expected. Initially slated for a May launch, sources familiar with the company’s operations indicate that the schedule has been moved forward, though an exact release date remains undisclosed.
This strategic acceleration follows the remarkable success of DeepSeek’s previous model, R1, which disrupted the global AI landscape. Despite utilizing Nvidia’s lower-tier chips, R1 outperformed several leading Western AI models. The model’s debut also contributed to a significant decline in technology sector market values, exceeding $1 trillion.
A Unique Approach in China’s AI Industry
Unlike mainstream Chinese tech firms such as Baidu and Tencent, DeepSeek has carved a distinct niche. Founded by Liang Wenfeng, a former manager of a closed quantitative fund, the company prioritizes research over commercial pursuits.
Liang, born in Guangdong in 1985 and a graduate of Zhejiang University in communication engineering, champions a decentralized management style. This contrasts with China’s traditional corporate hierarchy, which often adheres to the rigorous “996” work culture (9 AM to 9 PM, six days a week). At DeepSeek, employees—including fresh graduates—enjoy full autonomy over their projects and flexible working conditions.
“Liang gives us complete trust and treats us as experts in our respective fields,” said Benjamin Liu, a 26-year-old researcher at DeepSeek. “I have full freedom in developing key AI components.”
Focus on AI Model Quality Over Applications
Rather than investing in AI-based applications like its competitors, DeepSeek concentrates on refining AI model quality. This strategy has yielded impressive results, enabling R1 to rival world-class AI systems at a fraction of the cost.
DeepSeek’s growth has been fueled by substantial financial backing from High-Flyer, Liang’s quantitative fund company. Since 2020, High-Flyer has allocated approximately 70% of its revenue to AI research, constructing two supercomputer clusters. One of these, Fire-Flyer II, features 10,000 Nvidia A100 chips, playing a crucial role in the development of DeepSeek’s latest AI model.
Regulatory Challenges and Government Support
DeepSeek’s significant chip acquisition initially drew scrutiny from Chinese regulators. Concerns arose over the company’s stockpiling of advanced semiconductor components. However, the authorities ultimately refrained from intervention—a decision that proved critical when the U.S. imposed an export ban on Nvidia’s A100 chips to China in 2022.
With geopolitical tensions escalating, the Chinese government now fully supports DeepSeek. Officials have reportedly advised the company to maintain a low media profile to avoid excessive Western scrutiny.
Revolutionary AI Architecture and Cost Efficiency
A key factor behind DeepSeek’s success lies in its innovative AI architecture. The company employs Mixture-of-Experts (MoE) and Multihead Latent Attention (MLA) techniques, optimizing data processing while reducing computational demands. Analysts at Bernstein estimate that DeepSeek’s operational costs are 20 to 40 times lower than those of OpenAI.
Global Controversies and Western Reactions
DeepSeek’s rapid rise has sparked concern in Western countries. Governments in South Korea and Italy have removed DeepSeek applications from national app stores over privacy concerns. Additionally, Western AI entrepreneurs, including Scale AI CEO Alexandr Wang, have alleged that DeepSeek possesses up to 50,000 high-end Nvidia chips—despite U.S. export restrictions—though no concrete evidence has surfaced.
DeepSeek’s Expanding Influence
The AI model has already been adopted by various Chinese government agencies and state-owned enterprises, including 13 municipal governments and 10 energy corporations. Major tech players like Lenovo, Baidu, and Tencent have also integrated DeepSeek models into their platforms.
The Chinese government views DeepSeek as a testament to the nation’s ability to challenge global AI dominance, particularly amid ongoing trade restrictions and technological embargoes imposed by the U.S.
Impact on Global AI Markets
In response to DeepSeek’s disruption, Western AI companies are adjusting their strategies. OpenAI has reduced service costs, while Google Gemini has introduced more affordable subscription plans. The shift reflects the growing impact of DeepSeek’s cost-effective approach on the global AI industry.
With the imminent launch of R2, DeepSeek is poised to further reshape the competitive landscape. As the world anticipates its next move, the company’s continued innovation could redefine AI’s future trajectory.
For more updates on global AI developments, visit News Xpress Online.
