Yan Junjie

ACF Cover Figure | Yan Junjie: Four Years to Turn a Dismissed Company into a 300-Billion Enterprise

Asia "Cover Figure" · Perspective

Issue No. 044 — ACF-GJ-2026

From a county town, through academic foundation, big-tech experience, and resignation at an industry high point to start a company — Yan Junjie's career trajectory is a typical sample of China's first generation of AI technology entrepreneurs. Following a technical route contrary to the mainstream, advancing quietly, he pushed a startup to 300 billion HKD market cap on HKEX in four years, becoming one of the most representative companies in the global large model track. His growth and choices reflect how a generation of technology managers maintains essence amid noise and chooses long-term amid cycles.

I. From County Town to Academic Summit: Technical Foundation Determines Long-Term Judgment

Yan Junjie's early career centers on continuous academic climbing. From a Henan county town, he advanced through systematic academic training to earn a PhD from the Chinese Academy of Sciences and complete postdoctoral research at Tsinghua University. This rigorous engineering and AI academic training shaped his logic-first, fundamentals-first, deployment-first thinking — the foundation for critical judgments in later entrepreneurship. Before deep learning and large models became mainstream, he had already accumulated solid capability in underlying algorithms, system architecture, and engineering implementation.

With this strong technical background, Yan Junjie joined SenseTime, rising from core R&D to Vice President, deeply participating in technology architecture and business layout, witnessing computer vision's full journey from lab to industrialization. At SenseTime's critical pre-IPO stage, he held a stable, well-resourced senior position with a clear career path — mature team, industry resources, and defined future expectations.

It was precisely at this stage that he developed a cooler industry observation: massive AI resources were flowing into parameter races, concept packaging, and launch marketing — yet few models could truly achieve scaled, stable, low-cost deployment. Many models remained at demo stage, unable to support high-concurrency, long-cycle, low-cost commercial scenarios. He judged the industry's next core wouldn't be about volume, but technical efficiency and genuine commercialization. This judgment ultimately drove him to abandon the established platform and choose independent entrepreneurship.

II. Turning at the Summit: Abandoning the Big-Tech Path for a Harder Technical Route

In 2021, before industry expectations for large models had fully exploded, Yan Junjie formally resigned from SenseTime and launched his venture. This decision wasn't widely understood: giving up a mature platform, equity returns, and a stable team to enter a highly uncertain track meant reinvesting all accumulated assets into risk. Many saw it as "an unnecessary gamble" — but for Yan Junjie, it was an inevitable choice based on long-term judgment.

In early 2022, MiniMax was founded with fewer than twenty people. On the technical route determining the company's long-term fate, Yan Junjie did not choose the dense architecture that was easier to demo and attract capital attention. Instead, he committed to MoE (Mixture of Experts) architecture as the core direction — higher R&D difficulty, longer convergence cycles, harder to produce early demos — but with structural advantages in inference efficiency, concurrency, and long-term costs, more suitable for scaled commercialization. He insisted: AI must ultimately become universal, requiring solutions to three core problems — "good, stable, cheap."

During the early stage, MiniMax maintained extremely low market exposure. The team didn't chase hotspots, hold hype events, or expand aggressively. Instead, they prioritized capital and manpower toward underlying architecture, engineering stability, and real scenario refinement. While the industry pursued speed and volume, MiniMax chose slowness against speed — first consolidate the technology foundation.

III. Product & Commercialization: From Quiet Refinement to Global Validation

Long-term technology investment gradually translated into genuine product and market competitiveness. In 2022, MiniMax's Glow launched, quickly building reputation among target users. In August 2024, Hailuo AI launched, achieving sustained user growth through stability and practicality. Neither product relied on massive traffic spending — growth came primarily through user word-of-mouth, unusual in the marketing-dependent internet industry.

As of September 2025, Hailuo AI reached 42.348 million global users with over 590 million pieces of generated content. Social products globally surpassed 147 million cumulative users with over 11 million monthly active users. On the commercialization front, MiniMax showed high-quality growth: 2025 first three quarters revenue grew 170% YoY, overseas revenue exceeded 70%, proving a globalized business model. In four-plus years, cumulative financing approached 1.5 billion USD, valuation rising from 1.2 billion (angel round) to 4.2 billion USD — a global AI track head unicorn.

IV. HKEX Listing: Market Recognition Behind 300 Billion Market Cap

On January 9, 2026, MiniMax listed on HKEX at 165 HKD per share, rising 42.67% on the first day with market cap exceeding 70 billion HKD. Within two months, stock price peaked at approximately 1,300 HKD, market cap reaching 300 billion HKD. The IPO public offering was oversubscribed 1,837 times, with 460+ institutional investors participating — among the top in HKEX tech IPOs. Capital market recognition fundamentally confirmed the company's technical route, product capability, global commercialization, and team execution.

MiniMax's growth path provides an industry sample: in the large model track surrounded by giants with capital intensity, a startup with humble origins and a low-profile start can still break through through technology and product. It proves startups need not compete head-on with giants on resources — they can find structural opportunities in technical routes, product experience, and global deployment.

V. Long-Termism: Maintaining Restraint and Focus Amid Noise

Even after listing and achieving high market cap, Yan Junjie maintains low exposure, high focus. He rarely gives public speeches, doesn't build personal IP, concentrating energy on technical direction, product iteration, and organizational building. His core logic remains clear: AI's ultimate competition is not about short-term buzz, but who solves problems most stably, creates value most sustainably, and builds the healthiest commercial closed loop.

For Yan Junjie, entrepreneurship was never about rapid IPO and cashing out — it was about building a genuinely long-term enterprise with technical substance and social value. He hopes MiniMax can contribute at the AGI infrastructure level, bringing AI technology to more scenarios at lower cost and higher efficiency.

VI. Lessons: What's Scarce Is Not Trends — It's Long-Term Conviction

From county-town student to academic researcher, from big-tech executive to listed company leader — Yan Junjie's path involved no dramatic reversals or concept packaging. Instead, it relied on continuous judgment, execution, and conviction, forging a steady technology entrepreneurship route. His experience shows that true breakthroughs are not overnight explosions, but continuous accumulation in the right direction, restraint before temptation, and holding fast to its original mission amid noise.

Every cover publication is registered and kept in the ACF Cover Archive. The code is permanent.

Further ReadingRelated

Wang Cen

ACF Code

ACF-01-00198

Wang Cen

Chairman & Founding Partner

Zhang Luokang

ACF Code

ACF-01-00199

Zhang Luokang

Founder & Chairman

Frank Wang

ACF Code

ACF-01-00200

Frank Wang

Founder

Cao Hui

ACF Code

ACF-01-00102

Cao Hui

Chairman

Not Everyone Makes the Cover

Asia Cover Figure · Influencing Those Who Influence