MiniMax: A Young Man from a Henan County and His 300 Billion

Author: Lin Wanwan

In 2014, Baidu Research Institute had an intern, a PhD from the Institute of Automation, Chinese Academy of Sciences, from a small town in Henan. He calculated his own prospects: after graduation, his ideal destination was IBM, coding in Java, with an annual salary of 280,000 yuan.

During the Spring Festival of 2026, a tool called OpenClaw Agent exploded globally. Developers needed a large underlying model to support the lobster framework. There was a model that was fast and cheap, consuming 1.44 trillion tokens in a week on OpenRouter, topping all platforms.

This model was called M2.5, produced by MiniMax.

Two months after going public, its stock price soared from HKD 165 to HKD 1,300, with a market value surpassing HKD 300 billion, despite the company’s annual revenue being less than $80 million.

The person behind MiniMax is the same intern from twelve years ago, Yan Junjie.

A Bet Made Over a Year in Advance

During the Spring Festival of 2021, Yan Junjie returned to his hometown in Henan to visit his grandfather.

His grandfather told him he wanted to write a memoir, to record his 80 years of life. But he couldn’t type, nor organize his stories well. After mentioning it several times, he gave up.

Yan Junjie had been in the AI industry for over a decade. At that moment, he suddenly realized that although what he had built had been implemented in the industry and helped many companies, it was of no use to an elderly person wanting to write a memoir.

This detail was later repeatedly cited, giving it a somewhat inspirational story feel. But it truly explained one thing: his motivation for doing AI was simple—to make it accessible for ordinary people. This obsession later drove a series of counterintuitive decisions.

At the end of 2021, he left SenseTime.

The timing was crucial. SenseTime was preparing for a Hong Kong IPO. He was a vice president, deputy director of the research institute, CTO of the Smart City Business Group. He left at a time when the company was most valuable. He didn’t wait for the IPO or the wealth realization; he left before that.

ChatGPT was released only in November 2022.

MiniMax was founded in December 2021.

This time gap laid the foundation for everything that followed. Yan Junjie later said that if he hadn’t started early, in the later environment where “star researchers and big tech AI backgrounds” were more favored in fundraising, MiniMax wouldn’t have been able to compete.

His parents were ordinary people. He studied high school in a county town, then got into Southeast University’s Mathematics Department, later earned a PhD at the Institute of Automation, Chinese Academy of Sciences, did a postdoc at Tsinghua, and then joined SenseTime. He rose step by step, with no overseas background or prominent connections.

During his internship at Baidu, he interacted with Yu Kai from Horizon Robotics. Yu Kai later said that academic ability can be trained, but those who can engineer AI technology into real products are rare. Yan Junjie was one of them.

After joining SenseTime, he spent seven years from intern to vice president. In 2018, with limited staff, he led the team to develop a set of “All for One” model algorithms, surpassing Megvii and Yitu in bidding competitions, achieving industry first. Someone commented that he “reads papers very quickly, focusing only on the essence, ignoring clichés.” This efficiency later became the company culture at MiniMax.

He named the company MiniMax, inspired by von Neumann’s minimax algorithm in game theory.

He explained that decision-making should first guard against the worst risks, then choose the relatively optimal solution.

An Unusual Shareholder Chart

In December 2021, MiniMax completed its angel round, raising $31 million, with a pre-money valuation of $170 million. Investors included MiHoYo, IDG, Hillhouse, and Yunqi.

The money from MiHoYo was somewhat special. Yan Junjie had a good relationship with Liu Wei, chairman of MiHoYo. They invested during the angel round, and Liu Wei remains a non-executive director on MiniMax’s board.

MiHoYo is also a client of MiniMax, using their models for NPC dialogue and story generation in games.

After the angel round, a small incident occurred.

In March 2023, Silicon Valley Bank declared bankruptcy. At that time, all of MiniMax’s funds were in that bank. It was the riskiest moment early in the startup’s life—money was lost, and the fundraising environment was chaotic. But they survived. Two months later, they secured Series A funding of $257 million, with a valuation of $1.157 billion.

The subsequent investors became increasingly prominent. Alibaba invested, Tencent invested, Sequoia Capital followed. Before going public, they completed seven rounds of financing, raising nearly $1.5 billion in total, with a valuation of $4.2 billion. After the IPO, Alibaba held 12.52%, making it the largest external shareholder.

Yan Junjie had a habit in early fundraising: only meet with top-tier investors. He met with Sequoia’s Shen Nanpeng and Hillhouse’s Zhang Lei.

But there’s another person on the shareholder chart worth mentioning: Yuan Yeyi.

Born in 1994, she graduated with a bachelor’s in electronic engineering from Johns Hopkins University, minoring in economics and mathematics. She joined SenseTime right after graduation in 2017, working on financing and strategic investments. A year later, she was promoted to executive assistant to CEO Xu Li and director of the strategic department. She was deeply involved in SenseTime’s entire journey from early days to Hong Kong listing.

In 2021, she left to start her own venture with Yan Junjie.

An investor commented that she is “capable, commanding, highly driven, with a maturity beyond her age.” Her role with Yan Junjie was clear: one defined the technical vision, the other turned that vision into money and resources. Yan Junjie could dive into technology, even with shaved hair, but the market, capital, and globalization were Yuan Yeyi’s battleground.

On the day of the listing, both stood on the same stage. Yuan Yeyi, 31, valued at over HKD 40 billion.

385 People and 1% of the Money

At the time of MiniMax’s IPO, the company had 385 employees, with an average age of 29.

From founding until September 2025, the company spent about $500 million. OpenAI spent between $40 billion and $55 billion during the same period.

This comparison is somewhat absurd. Using less than 1% of the opponent’s money, they built a globally leading multimodal company. Cost-saving was just the result. The real reason was their relentless pursuit of AI excellence. 80% of their code was generated by AI, internally called “interns.” These interns had permissions high enough to access code repositories, modify online environments, chat in Feishu, review, and deploy directly.

This efficiency made MiniMax’s per capita output abnormally high.

Product-wise, they adopted a full multimodal approach from the start: language, video, speech, music—pushing all four simultaneously. While others were learning ChatGPT for dialogue, Yan Junjie focused on multimodal fusion. He believed that multimodality was the fundamental premise for continuous intelligence improvement. Without it, the next-generation models would have no chance.

In summer 2023, he made a more radical decision.

To allocate 80% of computing power and R&D resources to MoE (Mixture of Experts).

At that time, domestic mainstream models were still iterating dense models. MoE was considered “cutting-edge but immature.” Yan Junjie’s logic was simple: to serve tens of millions or hundreds of millions of users, the cost and latency of token generation with dense models were unsustainable. Without MoE, scaling was impossible; everything else was pointless.

In early 2024, MiniMax released China’s first MoE large model.

In products, they didn’t compete domestically. They launched Xingye and Talkie for C-end, one in China, one overseas, focusing on AI companionship; Heluo AI for video generation, which ranked first globally in monthly active users for video generation applications for six consecutive months in late 2024.

Current figures: 236 million users, covering 200 countries and regions, with 73% overseas revenue. 214,000 enterprise clients and developers, with deployments in Google Vertex AI, Microsoft Azure, AWS. Notion’s first open-source model choice was also MiniMax.

In February, ARR surpassed $150 million. The M2 series’ daily token consumption was six times that of December last year, with programming-related growth exceeding tenfold.

This is why the market is willing to assign a 200x sales multiple.

But some numbers need to be viewed carefully.

In the annual report, C-end gross profit margin was 4.7%, B-end was 69.4%. 67% of the company’s revenue came from C-end, but C-end contributed almost no gross profit. In Q4, C-end gross margin dropped to about 2.1%. Overall gross margin increased from 12.2% to 25.4%, mainly because B-end revenue surged in Q4, pulling up the overall figure.

This is an unresolved question.

The Impossible Mountain Isn’t Unclimbable

In June 2025, MiniMax released the M1 model.

Yan Junjie posted a sentence on WeChat Moments:

“For the first time, I feel that the mountain isn’t unclimbable.”

The reality behind this statement is that the top-tier model capabilities of China and the US might only differ by about 5%, but that 5% allows overseas companies to dominate scenarios worth ten times more, charge ten times higher prices, and ultimately create a nearly hundredfold commercial gap. OpenAI’s latest valuation exceeds $700 billion. MiniMax’s market cap at listing was HKD 800 billion, less than $100 billion.

He made a judgment that in the future, there will be five top-tier AGI companies worldwide, at least two from China, and perhaps one that becomes number one.

After going public on January 9, he appeared at a government-led expert entrepreneur symposium on January 19, becoming the second AI large model founder after Liang Wenfeng of DeepSeek to attend.

Then, on March 2, the first annual report was released, causing a surge in Hong Kong stocks.

During the earnings presentation, Yan Junjie spent a long time discussing one thing: MiniMax needs to transform from a “large model company” into an “AI platform company of the era.”

He proposed a formula for platform value: intelligence density × token throughput. In the internet era, platforms are traffic gateways; in the AI era, platforms are companies that define the boundaries of intelligence and share in the commercial dividends. Google is doing it, OpenAI is doing it, and they want to do it too.

His competitors are many times larger than him.

Listing in Hong Kong only pushed him into another battlefield. Quarterly reports, analysts, market cap pressures—these are completely different from coding. The secondary market doesn’t believe in sentiment; it only looks at numbers. Whether the C-end story can translate into gross profit, whether B-end growth can be sustained, when M3 will come out—these are questions that must be answered each quarter.

But looking at the bigger picture, MiniMax’s story isn’t just about one company.

In recent years, the US has tightened chip access more and more. A100s are restricted, H100s are restricted, H800s are restricted. The logic is straightforward: control computing power, and you control AI’s throat.

China has been forced onto a completely different path.

DeepSeek achieved near-H100 performance using H800. MiniMax spent $500 million to do what OpenAI took hundreds of billions to achieve. Yan Junjie’s gamble on MoE in 2023 was because the limited hardware couldn’t support inference at the scale of hundreds of millions of users. M2.5 running continuously for an hour costs only $1, which is one-twentieth of GPT-5. Innovations like hybrid attention, linear attention, CISPO algorithms were all driven by necessity.

The chip blockade was meant to widen the gap, but in practice, it pushed Chinese AI companies into a low-compute, high-efficiency evolution route.

Less money, fewer chips, fewer people—yet this pushed out extreme engineering capabilities and architectural innovations.

This is similar to Huawei’s chip strategy: block one capability, and I’ll make up for it in other dimensions. During this process, new things you don’t have might emerge.

OpenAI now has over 4,000 employees, spent $8 billion in cash by 2025, and plans to invest $600 billion in computing power by 2030. MiniMax has 385 people, with a total investment of $500 million.

Who will win remains uncertain. But at least now, fewer and fewer people believe MiniMax will fail.

The PhD student from Henan who interned at Baidu in 2014 probably never imagined that twelve years later, he would stand at this position, backed by a whole national-level technological competition.

He chose to keep running.