MiniMax announced a $2 billion financing today. 160 billion Hong Kong dollars. One hundred and sixty-three institutional investors participated. Seven times oversubscribed. The deal was upsized from $1.8 billion to over $2 billion after demand blew past expectations .
The money comes with a twist. 80% is going to AI infrastructure and model R&D. The company has 100,000 business and developer clients now — up 5x from six months ago. Inference demand is exploding .
But the headline isn't the money. It's the founder. MiniMax CEO Yan Junjie sent a company-wide email today, published in full. He announced that until MiniMax achieves AGI, he will not take any salary from the company .
He also pledged 4% of his personal shares — about $400 million at current valuation — to team incentives over the next four years. Another 1% will fund open-source community development. Total founder commitment: roughly $500 million .
He signed the letter with "We will keep going until we get there. Intelligence with Everyone."
Here's the context. MiniMax had its first major lock-up expiration yesterday. 49% of its shares hit the market. The stock dropped 18% in a single day — $2.5 billion in market cap wiped out .
Today, the stock is trading around 270 Hong Kong dollars. Three months ago, it was 1,330. Down roughly 80% .
The investors who said they'd hold? More than 80% of pre-IPO and cornerstone investors publicly pledged to stay. Alibaba. MiHoYo. Aspex. Boyu Capital. But it didn't matter. The stock got hammered anyway .
Why? Because the difference between "holding" and "selling" is a question of which investors you're actually dealing with .
Pre-IPO venture funds have fund cycles. They bought in cheap. They're sitting on massive returns. Even at 80% off the peak, their cost basis is still pennies . When 49% of a company's shares become sellable in one day, you only need 5% of that to hit the market to destabilize the entire stock .
The contrast with Zhipu, MiniMax's closest domestic AI competitor, is stark. Zhipu had its own lock-up expiry this week. Only 5.76% of its shares were eligible. Zhipu's stock rose 13% on its lock-up day . MiniMax dropped 18%. Same industry. Same week. Different math.
Now the financing. $2 billion at a moment when the stock is cratering. The terms: new shares at 268 Hong Kong dollars per share — below the current trading price. Plus a convertible bond with a conversion price of 335 Hong Kong dollars, a 12% premium to yesterday's close .
That structure tells you something. The investors are getting a discount on new shares and a convertible bond with upside. They're not buying the stock at market. They're buying a bet on the company's survival and eventual recovery.
The problem for MiniMax is that its business model is still finding its footing. C-side products — Talkie, Hailuo, Audio — account for 67% of revenue. But C-side gross margin is only 4.7%. B-side gross margin is 69%, but it's only one-third of revenue .
Revenue grew 158% last year to about $79 million. Net loss: $250 million . The company's ARR is roughly $150 million, according to management. They're aiming for $1 billion by end of 2026. That's a 6x jump in six months .
The founder is betting his personal compensation on AGI. That's a strong signal. But AGI doesn't have a date . It could be three years. It could be thirty. It could be never.
Meanwhile, the company has two other problems. Zhipu just raised $4 billion this week — twice MiniMax's round . And DeepSeek, with its aggressively priced models, is eating the "good enough at low cost" market that MiniMax once owned .
P.S. Yan Junjie just pledged $500 million of his own shares to his employees and the open-source community. That's $500 million he won't see if the company fails. But if the company succeeds, that $500 million stake could be worth $5 billion. The trick is getting there before the compute bill catches up.
