MiniMax increases registered capital to 4 billion yuan, as Yan Junjie builds a computing power moat with capital depth to navigate the brutal shift from algorithm competition to efficiency harvesting in the large model race.
On the battlefield of China's AI sector, which has already turned bloody red, money is no longer everything, but without money, nothing can be done. Recently, a business registration change involving MiniMax's affiliated company, Shanghai Xiyu Jizhi Technology Co., Ltd., wrote the most expensive footnote to this brutal consensus. Tianyancha data shows that the company's registered capital surged from 1 billion RMB to 4 billion RMB — a threefold increase that underscores this large-model unicorn's full-speed transition from the "technology frontier era" into the "capital showdown era."
This capital increase is far from a simple bookkeeping exercise — it represents the most fundamental defensive fortification Yan Junjie laid for MiniMax in early 2026.
Looking at the deeper cause, the core driver behind this capital expansion is the large-model sector's shift from "algorithmic competition" to "efficiency harvesting." In previous years, the industry debated whose parameters were more impressive and whose logic edges closer to Altman. But by 2026, the logic of large models has become extremely simple and ruthless: whoever can drive inference costs to the lowest point in large-scale commercial deployment is the one qualified to stay at the table. The 4 billion RMB in registered capital is, in effect, a massive credit check MiniMax has handed to the market, to competitors, and to upstream and downstream supply chains. This means that whether it is advance payments for compute leasing or deposits for chip procurement, MiniMax now holds financial parity with top-tier giants.
The equity structure disclosed by Tianyancha is equally noteworthy. Shanghai Xiyu Jizhi is wholly owned by Hong Kong Xiyu Jizhi Co., Ltd. This typical cross-border structure not only facilitates absorbing the mixed capital of top-tier dollar or RMB funds like Alibaba and Sequoia, but also preserves flexible capital exits for future global expansion. In the context of 2026, a purely domestic AI company can hardly survive the dual pressures of compute restrictions and technology going global. MiniMax's strategy of strengthening its corporate entity's creditworthiness through a capital increase while maintaining an offshore holding structure is, in essence, a highly sophisticated form of "identity hedging."
Why now? Because the "hundred-model war" in the large-model market has entered a substantive phase of reckoning. Alibaba's Tongyi and Tencent's Hunyuan are closing in from behind, while fellow "AI tigers" like Zhipu and Moonshot AI are also pushing heavy capital penetration through different paths. By choosing this moment to raise its registered capital to 4 billion, MiniMax is sending an extremely powerful signal to potential price-war entrants: the entry fee has risen to astronomical levels, and without a capital base of comparable scale, any small-scale technological tinkering will be unable to shake the industrial ecosystem it has already built.
From the business scope shown on Tianyancha — computer system services, information system integration, and AI application software development — these seemingly mundane terms conceal MiniMax's ongoing horizontal expansion. The 4 billion in capital will most likely be directed toward building or deeply customizing more asset-heavy compute centers. When general-purpose compute resources become expensive and scarce, only companies like MiniMax that dare to directly invest in underlying assets can escape absolute path dependency on cloud service giants.
The contrast generated by this move lies in the fact that outsiders assumed the core assets of an AI company were its code-writing engineers. But Yan Junjie used this 4 billion to tell the market that in the hard-tech jungle, the depth of capital is the fault tolerance for technological breakthroughs. The iteration of large models leaves no room for hesitation — a single failed training run could mean losses in the tens of millions or even hundreds of millions. The 4 billion platform gives MiniMax the strategic resolve to tear things down and start over multiple times when faced with unknown technical paths.
The ultimate endgame of business is always the struggle for control. Through this capital increase visible on Tianyancha, MiniMax is essentially accelerating its metamorphosis from "unicorn" to "industry platform." When a platform's registered capital reaches the 4 billion scale, its gravitational pull on talent, its credibility endorsement to customers, and its bargaining power with government industrial policies all undergo a qualitative leap.
This game over AI sovereignty is increasingly resembling an extremely capital-intensive heavy industry race. Yan Junjie's move in 2026 signals that the first tier of China's large-model players has completed its initial capital accumulation. What comes next will be a bloody harvest of application depth and commercial boundaries, built on this 4 billion capital foundation. Any players still clinging to PowerPoint storytelling or trying to extract outsized profits through asset-light models will feel a physical sense of despair against such a formidable wall of capital. This is not just a victory for MiniMax — it is a milestone in China's AI industry's complete transformation toward an asset-heavy, high-barrier, long-cycle logic.
