Merge the "Six Little Tigers" to build China AGI.

"Don't Evaluate AI Companies Like You're Solving a Math Problem"

Yesterday I shook hands with Brother Dao at the Shenyang Street ruins. Great time. Today I'm here to roast Zhipu AI and MiniMax's IPOs.

Reading Zhipu AI and MiniMax's prospectuses is pure comedy gold.

I was ready to absolutely demolish these two — I call them the Dip Brothers. Putting out such terrible financials this fast? Aren't they just tanking the market for AI model companies?

To be precise, they're tanking the market for Kimi.

The reason is obvious. Of the "Six Little Tigers," two others have already bowed out of this race. Only StepFun still occasionally drops a product update or pulls a stunt, stubbornly doing pre-training like Brother Tiger from the Northeast doggedly showing up at anime cons.

Of course, my favorite part is Yutong's counterattack.

The moment these two filed their prospectuses, I bet a lot of you were maliciously speculating about Kimi's mindset like I was. On one hand, your competitors just self-destructed by publishing their financials — hard not to secretly enjoy that. On the other hand, such garbage fundamentals drag down the entire large model layer's valuation. Aren't they just straight-up murdering Kimi's valuation?

So evil, fam 😭

But Kimi still has some fight in them, and Yutong's counter came swift and direct. So we got this round of Kimi PR disguised as an internal memo. Emphasizing they just raised $500 million in a new round, with over 10 billion RMB in cash on hand.

Saint Yang even straight-up said: "Our Series B/C funding already exceeded the vast majority of IPO proceeds and listed companies' private placements. So we're not rushing to go public in the short term."

Exactly which two companies are targeting ~$500 million in expected proceeds, I'll pretend I don't know 🤡

The funniest part? Kimi's funding news and Saint Yang's all-hands letter dropped on the last day of 2025. Not only did they pull a major stunt right before the holiday, they timed it to when their competitors just started book-building but hadn't officially listed yet — deliberately hitting where it hurts. Gotta admit, Kimi does have some showmanship.

Back to Zhipu AI and MiniMax.

Let's review the Dip Brothers' financials:

Zhipu AI: 2023 revenue of 125 million RMB, adjusted net loss of 621 million RMB; 2024 revenue of 312 million RMB, adjusted net loss of 2.466 billion RMB; H1 2025 revenue of 191 million RMB, adjusted net loss of 1.752 billion RMB.

MiniMax: 2023 revenue of $3.46 million, adjusted net loss of $89 million; 2024 revenue of $30.52 million, adjusted net loss of $244 million; first 9 months of 2025 revenue of $53.44 million, adjusted net loss of $186 million.

Combined revenue doesn't even match Manus. Here I need to call out some media folks — didn't you say MiniMax's 2024 revenue was over $70 million and that number was conservative?

Zhipu AI only has 2.55 billion RMB in cash left (as of June 30, 2025). Yeah, they really need to IPO soon or where's the money gonna come from? But MiniMax still has $1 billion on hand. Someone should ask Chen Mian at Lovart if he's jealous of Manus and wants to chat about an acquisition?

That way they could boost revenue and gross margin for my Pangmao AI 😭

But over the past couple days, I realized something: So what if Zhipu AI and MiniMax's financials are absolutely terrible?

People buying into AI are buying faith — belief that AGI will happen. That has nothing to do with currently awful financials.

The logic is: you can't evaluate AI model companies with a test-taker's mindset. If you want pretty financials, go buy bank or consumer stocks. If you're already here caring about faith-driven AI, why are you even looking at fundamentals?

AI is faith investing — completely different from the value investing that middle-aged dudes love. When I explained this to my former fashion editor friend Luozi Ma, his first reaction was: isn't this just Bitcoin logic — valuable because people believe?

Yeah, that's basically it 🥵

So I think Zhipu AI and MiniMax's IPO marketing strategy has major problems. These two keep trying to emphasize how good their financials look, how healthy their cash flow is, how solid their fundamentals are. Aren't they just clowning us 😭

For example, Zhipu AI emphasizes having 8.9 billion RMB in available funds (as of October 31, 2025), but actually counts 6.1 billion in bank credit lines. A 59% gross margin on on-premise deployment is nice, but why do your top five customers change completely every year? Zero loyalty in B2B.

MiniMax emphasizes that over 70% of revenue comes from overseas, but your CAC on consumer products is nearly 1:1 with revenue, gross margin only 4.7%. And I won't even mention why they highlight their board's average age being post-'95 😴

None of this data is pretty, but it doesn't matter.

Just admit your fundamentals are terrible and your financials suck right now. No problem, because it doesn't matter.

What people are chasing is AGI, and AGI achieved by a Chinese company. That's an incredibly valuable faith.

And this has no linear relationship with fundamentals. NVIDIA at $1 trillion market cap might have fundamentals. NVIDIA at $5 trillion clearly has no fundamentals — it's purely a faith stock. AI is currently a faith industry.

What I realized is: you can't approach the secondary market with a test-taker's mindset, and you absolutely can't approach AI with one.

The only thing that matters for these two companies is whether they can build impressive AI.

Well, obviously their AI isn't that impressive. Zhipu AI's GLM model — plenty of people around me use it, but it's just one of several Claude alternatives. MiniMax has special meaning to me because they created Pangmao, but whether it's video model Hailuo AI, text model M2, or STARFIELD, none are top-tier.

You can see this from OpenRouter call volume rankings too. For full year 2025, DeepSeek ranked 5th, Qwen 6th, with Zhipu AI and MiniMax at 8th and 9th.

So we can reasonably conclude: Zhipu AI and MiniMax's models are usable, with some distinctive features and highlights, but none are truly top-tier — kind of stuck in the middle.

Whether you believe in Zhipu AI and MiniMax then depends on how much you believe in Chinese AI and AGI — two great narratives of nationalism and progressivism.

I obviously believe strongly in Chinese AI, so I subscribed to MiniMax, trying to support Pangmao with real money. Muqiu subscribed to both MiniMax and Zhipu AI — that's just overconfident.

Of course, our decisions are purely for fun, paying for faith (Muqiu says he's paying for the lolz), not investment advice. After all, how many people in this world are still willing to pay for Pangmao 😭

Anyway, what I'm saying is: doing the math on large model companies is meaningless. Because at this stage, the only decisive factor for large model companies is technical strength.

Reading the financials, I also discovered Alibaba is the one guaranteed winner.

Alibaba owns 13.66% of MiniMax. Even at the last round's $4 billion valuation, their investment was under $400 million. But MiniMax spent $58.3 million on Alibaba Cloud in the first 9 months of 2025 alone, and the prospectus forecasts annual Alibaba Cloud spending caps of $115 million (2026), $125 million (2027), and $135 million (2028).

So even ignoring IPO returns (which may or may not happen), Alibaba can earn back its investment purely through cloud services. And a significant portion of Alibaba's investment was paid in compute credits — virtue coupons.

I'm very much looking forward to Kimi's IPO to see how much Alibaba can make from them. You can't help but notice that in the VC market, Alibaba ends up being the most generous big brother. So great, Alibaba 👋😭👋

That said, Zhipu AI and MiniMax's middling performance is genuinely worrying, so I consulted a securities veteran on how to break through. My guy roasted them for lacking vision, then gave some very practical advice:

Merge the Six Little Tigers into one company.

The new company could have Jie Tang as Chairman, Zhilin Yang as Chief Scientist, Junjie Yan as CEO, Yeyi Yuan as CMO, and Yutong Zhang as CFO. If they don't mind Chuan Wang and Kai-Fu Lee's age, they could bring them in as Senior VPs handling BD. And nobody needs to go through the IPO process — just backdoor into Qianli Technology, giving StepFun's people a warm home.

For this company carrying half the world's AGI dreams, I propose naming it Panghu AI — combining Pangmao and the Six Little Tigers.

Truly win-win, win to the max. Directly welding scarcity shut .

After all, if Mistral AI — which nobody but European bros cares about — can be valued at €12 billion, then our East's most diao, most promising AI large model at 500 billion RMB is completely reasonable 👊🇨🇳🔥

(Images in this article generated by ChatGPT; text purely human-written. Final disclaimer: this article does not constitute investment advice, purely for entertainment. Nobody's actually buying stocks to support Pangmao, right? Right?)