The numbers hit my screen at 8:47 AM Mexico City time. MINIMAX, the Shanghai-based AI powerhouse, just dropped its H1 2026 financials. Revenue: $117 million. Up 283.1% year-over-year. Gross profit: $20.8 million. Up 464.8%. Net loss: $358 million. Narrowed by 11%.
I've seen this pattern before. Chasing the white whale in the 2017 ether rush taught me that when growth outpaces losses, something structural is shifting. But when I ran the gross margin calculation, I stopped cold. 17.8%. That's not a typo. That's a company burning cash at a rate that would make a DeFi summer yield farmer blush.
The market will spin this as 'AI unicorn accelerates.' I see something different. This is a story about the brutal economics of AI video generation, the hidden cost of chasing the Sora dream, and a ticking clock that starts with a burn rate of roughly $60 million per month.
Let me break down what this actually means.
The Context: MINIMAX's Position in the AI Arms Race
MINIMAX isn't a household name like OpenAI or Anthropic. But in the AI video generation space, they're a serious contender. Their Hailuo series of video generation models has been turning heads since 2025. While everyone was obsessing over text-to-image, MINIMAX went hard after text-to-video, a far more computationally intensive problem.
The company's trajectory mirrors the broader market shift. We're seeing the AI application layer move from proof-of-concept to production. But here's the uncomfortable truth nobody wants to admit: video generation is a computationally brutal business. Generating one minute of high-quality video requires thousands of GPU hours. That's not a software margin business. That's a manufacturing business with silicon as the raw material.
MINIMAX reported a 283% revenue jump to $117 million in H1 2026. Sounds impressive. But let's put that in perspective. OpenAI is on pace for $10 billion in annual revenue. Anthropic is at $3 billion. MINIMAX's annualized run rate of roughly $230 million puts them at about 2% of OpenAI's scale. They're growing fast, but they're playing a different game.
The Core Analysis: Where the Money Actually Goes
The gross margin math is the story here. At 17.8%, MINIMAX is keeping only 18 cents of every dollar of revenue after direct costs. The other 82 cents? Gone to compute, data, and infrastructure. That's not a software company's profile. That's a capital-intensive industrial operation.
The growth in gross profit (up 464.8%) versus revenue growth (up 283.1%) tells me they're achieving real efficiency gains. The gap between these two numbers is the key signal. It suggests inference costs are dropping faster than pricing pressure. That could be from model quantization, better distillation techniques, or improved GPU utilization.
But here's what's missing from the headline: this company is still losing money at an alarming rate. $358 million in losses over six months. That's a burn rate of about $2 million per day. Even with the 11% loss narrowing, they're consuming cash like a black hole.
I audited yield aggregators during DeFi Summer 2020 that had better unit economics than this. At least those protocols knew their cost per transaction. MINIMAX's cost per video generation is a closely guarded secret, but the gross margin tells me it's still painfully high.
The revenue mix is the critical unknown. If MINIMAX is primarily selling API access to developers, the pricing pressure from OpenAI and ByteDance's Jimeng is brutal. If they're doing custom enterprise deployments, those contracts are high-cost, low-margin affairs. Either way, the path to profitability is not obvious.
The Contrarian Angle: The Efficiency Trap
Here's what the growth narrative misses. The 464.8% gross profit growth sounds like a triumph. But it's happening from a very low base. When I calculated the absolute numbers, the gross profit for H1 2026 is just $20.8 million. The gross profit for H1 2025 would have been around $3.7 million. This is a story of a company climbing out of a very deep hole, not one that has reached solid ground.
The real contrarian play is watching the competitive landscape. The AI video generation space is getting crowded fast. ByteDance's Jimeng, Kuaishou's Kling, OpenAI's Sora — all are spending billions to capture this market. MINIMAX is trying to fight a war on multiple fronts with a treasure chest that's running low.
I've been hunting spreads while the market sleeps long enough to recognize a market that's about to get crushed. The AI video generation space is heading for a consolidation phase. The players with the deepest pockets and the most efficient compute will survive. Everyone else will either get acquired or fade into irrelevance.
MINIMAX's differentiator was supposed to be their Hailuo video generation model. But video generation quality is becoming commoditized. The gap between Sora, Kling, and Hailuo is narrowing by the quarter. If MINIMAX can't maintain a technical edge, their only weapon is price. And price competition in a high-compute business is a race to the bottom.
The other overlooked issue is the regulatory environment. As a Chinese company listed on the Hong Kong exchange, MINIMAX faces dual regulatory oversight from mainland China and Hong Kong. China's rules on AI-generated content labeling are strict. Deepfake prevention is a growing concern globally. Any regulatory action could disrupt their growth trajectory overnight.
And let's talk about the deepfake problem. AI video generation is the primary tool for creating convincing fake videos. As MINIMAX's user base grows, so does their exposure to misuse. They'll need to invest heavily in content moderation and provenance tracking. That's another cost center dragging on margins.
The Takeaway: Watch the Margin, Not the Headline
MINIMAX's financial results are a classic 'good news, bad news' story. The growth is real. The technical progress is real. But the unit economics are still broken. The company is trading money for market share, hoping that scale will eventually solve the margin problem.
I've seen this movie before. It's the same pattern as the NFT minting frenzy of 2021 — massive revenue growth, terrible fundamentals, and a clock ticking toward reality. The question is whether MINIMAX can improve gross margins from 17.8% to 30%+ before the cash runs out.
At their current burn rate, MINIMAX needs to raise more capital or reach an IPO within the next 12 to 18 months. The Hong Kong exchange has a special chapter (18C) for pre-revenue tech companies, which could be their lifeline. But an IPO in a volatile market for AI stocks is not a guarantee.
The numbers say one thing: this is a company in a footrace against its own cost structure. Speed kills slower than greed. The question is whether MINIMAX can outrun its burn rate before the market loses patience with AI stories that don't make money.
I'll be watching their next quarterly report for two things: gross margin trajectory and any announcement about compute cost reductions. If the margin starts approaching 30%, this company becomes a serious contender. If it stalls, we're watching the beginning of the end.
Volatility is just noise until it becomes signal. This earnings report is signal. The signal says: AI video generation is a brutal business, and the companies that survive will be the ones that solve the compute cost puzzle first.
We don't need more AI demos. We need better unit economics. MINIMAX just showed us both sides of that coin.
