The numbers are seductive. Revenue at $117 million, up 283.1% year-over-year. Gross profit climbing 464.8%. Losses narrowing 11%. A classic growth narrative for a Hong Kong-listed AI darling.
But I have audited enough smart contracts to know that the most attractive surface often hides the most brittle structure. And when I ran the math on MINIMAX's filing, one figure stopped me cold: the gross margin.
At roughly 17.8%, this is not a technology company. It's a pass-through entity for GPU costs. Let me show you what the headline missed.
The Growth Is Real. The Economics Are Not.
First, let's establish what MINIMAX has actually achieved. The company, known primarily for its Hailuo series of video generation models, has crossed the chasm from proof-of-concept to production. Revenue of $117 million in the first half of 2026 is not trivial. In a sector where most AI video startups are still trading demo reels for seed rounds, MINIMAX has actual customers.
The 464.8% growth in gross profit relative to revenue growth is a classic unit economics signal. It suggests the engineering team has found ways to optimize inference costs. Better quantization. Smarter architecture. Leaner serving infrastructure. This is the kind of optimization that comes from real-world deployment pressure, not lab experimentation.
But here is where the narrative cracks. With gross profit at $20.8 million on $117 million in revenue, the gross margin sits at roughly 17.8%. For context, a healthy SaaS business runs at 70-80%. Even a capital-heavy cloud provider manages 30-40%. At 17.8%, MINIMAX is spending over 80 cents of every revenue dollar on direct costs, and the largest component of that is compute.
Video generation is not text generation. Generating one minute of high-quality video requires thousands of GPU inference calls. This is a fundamentally heavier compute profile than any LLM chatbot. The architecture forces a cost structure that looks more like a semiconductor foundry than a software company.
The Three-Body Problem of AI Video
Based on my experience auditing ICO-era smart contracts, I have a habit of looking for hidden leverage points. Here, the structural tension is obvious. MINIMAX faces a trilemma that no video generation company has solved yet.
First, model quality. The market expects cinematic output. Any regression in video coherence or aesthetic quality sends users to competitors like Kling, Runway, or OpenAI's Sora. This demands relentless research and development, which demands compute.
Second, price. The consumer market is brutal. Subscription tiers are anchored at $10-30 per month. Enterprise API pricing faces pressure from well-funded rivals. There is limited room to raise prices without losing market share.
Third, gross margin. With a 17.8% margin, there is no room for error. Any increase in GPU prices, any reduction in efficiency, any increase in inference load per user, and the margin evaporates entirely.
This is the core tension. To win in quality, you must spend. To win in price, you must optimize. To optimize, you need scale. And to scale, you need capital. MINIMAX is running a capital-intensive race where the finish line is a moving target.
The $358 Million Question
Let us talk about the loss: $358 million in the first half. That is a burn rate of roughly $7 billion annualized. I have seen this pattern before in DeFi protocols that promise the world but quietly burn through the treasury.
Wait, I have seen the pattern before, and it always ends the same way. The market rewards the revenue number. It ignores the cost structure. Then, one day, the market notices the cost structure, and the stock gets repriced in a single afternoon.
MINIMAX's losses have narrowed by 11%. That is a positive sign. But the narrowing is driven by the denominator effect. Revenue grew faster than costs grew. It is not necessarily because the company has become more efficient. In fact, the gross margin of 17.8% suggests efficiency gains are still far off.
The real question is: How much cash is left? At a $7 billion annualized burn rate, and without knowing the cash position, the company is in a race against time. It must either reach scale to achieve gross margin expansion or return to the capital markets for another round. Both options are risky in the current funding environment.
The Contrarian Angle: What If the Revenue Is the Problem?
Here is the counter-intuitive take. The 283% revenue growth might actually be a negative signal for long-term profitability. When a company grows this fast, it usually indicates one of two things: the market is genuinely hungry, or the pricing is too low.
With a 17.8% gross margin, I suspect the second. MINIMAX may be buying revenue through aggressive pricing. This creates a trap. The customers you acquire at a low price will resist future price increases. The market positioning gets set at a low value proposition. And the unit economics stay permanently depressed.
I have seen this pattern in the crypto world with yield farms that offer unsustainable APRs to attract liquidity. The users come for the yield, and they leave when the yield drops. The user base is not sticky because the attraction was not the product but the price.
The Infrastructure Tax
There is another angle that most analysts miss: the infrastructure dependency. MINIMAX is a Chinese company, operating in a market where access to the most advanced GPUs is constrained by export controls. The company is likely relying on a mix of domestic chips and cloud service providers.
This creates a dual dependency. It is a technology company that must buy its primary input from a supplier that is structurally compromised. Any geopolitical escalation could disrupt the supply chain and cost structure in a way that is outside the control of the management team.
This is the "pick-and-shovel" problem in reverse. Nvidia is selling the shovels. MINIMAX is digging the hole. And the shovels are getting more expensive.
The Signal I am Watching
The next quarter's gross margin. If MINIMAX can push the margin from 17.8% to 25% or higher, it will signal genuine architecture optimization and operational scaling. If the margin stays flat or drops, it will signal that the company is trapped in a pricing war it cannot win.
The revenue growth is exciting. The narrative is compelling. But the fundamental math does not work yet. And in a bull market where capital is easy, it is easy to forget that math eventually always wins.
The Takeaway: This Is Not a Profit Story Yet
MINIMAX is a revenue story, a user acquisition story, a technology story. It is not a profit story. The company is in a strategic consolidation phase, buying time with capital. The question is whether it can translate its technological leadership into a structural cost advantage before the capital runs out.
History does not always rhyme, but it often does. The pattern of high growth, low margin, high burn is the same pattern I saw in the early DeFi protocols, the ICOs, the NFT platforms. Some survived. Most did not. The difference was not the technology. It was the unit economics.
I am not saying MINIMAX will fail. I am saying the current numbers do not support a sustainable business model. The future is a binary outcome. Either the model gets significantly cheaper and the margin expands, or the capital runs out and the story ends. It is a bet on engineering discipline.
That is a bet I have seen lose more often than win. But the ones that do win, they have a structural advantage that is hard to replicate. MINIMAX has not shown that advantage. Not yet.