Yan Junjie is included in the 'Perspective' section because his entrepreneurial path demonstrates a rare industrial rationality. While the AI industry generally chased parameter races and concept packaging, he chose a harder but more essential technical route and persisted for four years. His story is not about revelry on the trend, but about how to maintain judgment amid noise. From a county town to academic summit, from big-tech executive to entrepreneur, each step was based on a calm deduction of technical essence and business laws. This steadfastness is the archival value that the Perspective section aims to preserve.

Yan Junjie's entrepreneurial journey began in 2021 when he resigned as Vice President of SenseTime to found MiniMax. In the early days, he chose the MoE (Mixture of Experts) architecture as the technical route, which was not mainstream at the time, but Yan believed it had advantages in inference efficiency and long-term cost. In 2022, MiniMax was officially established with a team of fewer than 20 people. The company initially kept a low profile, focusing on underlying architecture and engineering stability. In 2022, the product Glow was launched, and in August 2024, Hailuo AI was introduced, with user numbers growing steadily. As of September 2025, Hailuo AI had 42.348 million global users. In January 2026, MiniMax listed on the Hong Kong Stock Exchange, with market value briefly exceeding HK$300 billion. ^{[1] [2] [3]}

Yan Junjie's business philosophy emphasizes long-termism and returning to fundamentals. He insisted on the MoE technical route despite its high R&D difficulty and long convergence cycle, believing it was key to achieving scaled commercialization. He advocates that AI must solve the three core problems of 'good to use, stable, and cheap' to promote technological inclusivity. In commercialization, MiniMax focuses on healthy growth, with revenue growing 170% year-over-year in the first three quarters of 2025, and overseas revenue accounting for over 70%. The company has raised nearly US$1.5 billion in cumulative funding, with valuation rising from RMB 1.2 billion at angel round to US$4.2 billion. Key battles include the successful launches of Glow and Hailuo AI, and the HKEX listing, which saw 1,837 times oversubscription and participation from over 460 institutions. ^{[1] [2] [3]}

Yan Junjie and MiniMax's practice provide a sample for China's AI industry: in the large model track surrounded by giants, startups can break through through technical routes and product experience. MiniMax's success proves the value of low-key pragmatism and long-termism. Yan himself maintains a low-profile, high-focus style, rarely giving public speeches, concentrating on technology, products, and organizational building. His experience is seen as a typical sample of China's first generation of AI technology entrepreneurs, and his insistence on MoE architecture and global commercialization strategy has had a profound impact on the industry. The market's high valuation of MiniMax is not only optimism about short-term growth but also a bet on its long-term potential in the AGI era. ^{[1] [2] [3]}

The business proposition Yan Junjie leaves is: in technology cycles, how to identify true long-term value and persist to the end. MiniMax's success proves that startups do not need to compete head-on with giants in resources but can find structural opportunities in technical routes and global layout. His experience also reminds future entrepreneurs that true counterattack is not an overnight explosion but continuous accumulation in the right direction. In the process of AI moving from frenzy to maturity, the practice of Yan Junjie and MiniMax provides a replicable rational sample for the industry.