MINIMAX's Gross Margin Is a Red Flag Disguised as Growth

The numbers hit the wire at 9:00 AM Abu Dhabi time. MINIMAX, the AI unicorn with a video-generation obsession, reported H1 2026 revenue of $117 million, up 283.1% year-over-year. Gross profit climbed 464.8%. Net loss narrowed 11% to $358 million. The market will read this as a growth story. I read it as a cost structure confession. A 17.8% gross margin tells me more about this company than any revenue headline ever will. That is not a software business. That is a commodity processing operation with a model wrapper. And I've seen this pattern before — in 2017, when I audited 14 ICO whitepapers and found that 94% of them had emission schedules designed for founders to dump on retail. The mechanics differ, but the shape is the same: impressive topline, broken unit economics, and a clock ticking toward a liquidity event. The context matters. MINIMAX is a Chinese AI company with serious backing — Tencent and Alibaba sit on its cap table. Its Hailuo video generation models have made noise in the industry. The company filed a financial announcement with the Hong Kong Exchange, which smells like preparation for an 18C listing — the HKEX's special channel for pre-revenue tech companies. But the filing contains exactly four data points: revenue, gross profit, net loss, and the growth rates. No customer metrics. No unit economics. No segment breakdown. Let me do the math that the press release leaves out. Gross profit for H1 2026 was $20.8 million. Revenue was $117 million. That gives you a 17.8% gross margin. Now, for context: OpenAI reportedly operates at roughly 50-60% gross margin. Pure SaaS companies in the US run 70-80%. A company at 17.8% is not selling software. It is selling compute at a loss, with a model attached. The implication is that direct costs — almost entirely GPU compute — consumed 82.2% of every dollar of revenue. In absolute terms, that is roughly $96 million spent on infrastructure in six months. That is the cost of doing business in AI video generation. It is a compute-hungry game. Generating one minute of high-quality video costs an order of magnitude more than generating 10,000 tokens of text. But there's a second-order signal here worth noting: gross profit grew 464.8% while revenue grew 283.1%. That divergence suggests the company made real progress in inference optimization during the period. Quantization, model distillation, better scheduling — something worked. Gross margin improved from roughly 10% in H1 2025 to 17.8% in H1 2026. The unit economics are improving, but they're improving from a base that's still deeply unattractive. Now let's talk about what this means for the AI video generation narrative. If MINIMAX's revenue growth is primarily driven by video generation — and the evidence points that way — then the market is starting to pay real money for generated video. That's a thesis I've been tracking since the DeFi Summer of 2020, when I built a Python-based stress test that predicted cascading liquidations on Compound and Aave three weeks before they happened. The lesson from that exercise: when a sector grows fast but the underlying infrastructure is fragile, you prepare for a correction. In this case, the fragility is in the cost structure, not the code. The contrarian angle is this: everyone is chasing the revenue growth rate. But if you look at the gross margin trajectory, what you're seeing is a company that may have already peaked in its growth story. The problem is not the absolute level of the margin — that will improve as compute costs decline. The problem is the competitive dynamics of the market. In text generation, China's giants — ByteDance's Doubao, Zhipu, Baidu — are fighting a price war that's already collapsed margins to near-zero. Video generation is the next battleground, and ByteDance's Jimeng and Kuaishou's Kling are spending heavily. OpenAI has Sora. Everyone is chasing the same compute. And if the gross margin is 17.8% today, with leading-edge competitors subsidizing their products, what happens when the price war starts in video? Liquidity is a mirage in high heat. That's the phrase I keep returning to. The market is pricing MINIMAX like a tech company with network effects. But the balance sheet reveals a different story. The revenue growth is real — 283% is not a rounding error. But the burn rate is staggering: $358 million in half a year, or roughly three times revenue. At that pace, the company needs either a successful IPO or a continued stream of private capital to survive. The Hong Kong listing route is plausible — the 18C chapter was designed precisely for pre-profit tech companies — and the timing, right after a strong growth report, is consistent with a pre-IPO narrative. But here's the uncomfortable math: if MINIMAX needs to raise $500 million to sustain operations for another year, and the market sees a 17.8% gross margin, the valuation negotiation gets tricky. At 10-20x revenue, the company might command a $23-46 billion valuation. But those multiples are reserved for companies with clear paths to profitability. A company with a 17.8% gross margin and a 300% burn rate is not priced like a software company. It's priced like a semiconductor fab — capital intensive, high risk, long timeline. What's the actual investment case here? The short-term view is straightforward: MINIMAX is a high-growth, high-burn, low-margin AI company that will need to either drastically improve unit economics or secure a massive capital infusion. The long-term view is more complex. If the AI video market is actually going to be a multi-billion dollar market — and the evidence suggests it will — then MINIMAX's positioning as a specialized video generation provider could be defensible. The key is whether the gross margin can cross 30% within the next 12-24 months. That's the number I'm watching. Not the revenue growth. Not the gross profit growth. The gross margin. Because in this market, growth is cheap — capital is abundant and every AI company is growing. What matters is the margin trajectory. Here's my prediction. If MINIMAX's H2 2026 report shows gross margin above 25%, the IPO story is credible. If it's flat or declining, the company will face a brutal down-round and possibly a strategic sale. The video generation market is not a winner-take-all game — the technology is still improving, and the barriers are not insurmountable. The real question is not whether MINIMAX can generate revenue growth, but whether it can translate that growth into sustainable gross margin. The signal from this report is mixed. On one hand, the revenue growth and gross margin improvement are real. On the other, the absolute level of the gross margin is a red flag that masks the growth narrative. The market will see this as a success story. But I see the 17.8% as a warning. Consensus is fragile. And so is a business that spends 82 cents of every dollar it earns on the cost of goods sold. Watch the margin, not the headline. Because if the margin doesn't improve, the next headline you read about MINIMAX won't be about growth. It'll be about layoffs, down rounds, and dilution. That's the cycle. I've seen it play out four times now. Every time, the pattern is the same: revenue growth first, margin compression second, and the reckoning comes third. The only question is timing.