Gross profit grew 464.8% while revenue grew 283.1%. Losses narrowed by 11% on a base that is still three times the size of revenue. In the DeFi winter, we didn't see numbers like this. But then again, we weren't watching AI application companies. We were watching liquidity pools bleed out.
Today, I'm looking at MINIMAX's H1 2026 financial disclosure. It's a classic high-growth, high-burn narrative. But the underlying math tells a story that the headline numbers don't. Let's break down the order flow of this balance sheet.
The Context: A Video Generation Bet
MINIMAX is not OpenAI. It's not trying to be. This is a Chinese AI company that has staked its claim on multi-modal content generation, specifically AI video generation with the Hailuo series. In 2025, they made noise in the industry with their video models. This financial report is the first hard look at whether that bet is paying off in dollars, not just demos.
The company posted $117 million in revenue for the first half of 2026. That's a 283.1% year-over-year jump. Gross profit came in at $20.813 million, which is a 464.8% increase. Net loss narrowed to $358 million, an 11% improvement. On the surface, this looks like a company hitting an inflection point. Costs are falling faster than prices, and the machine is scaling.
But here's where I start to squint. The gross margin is sitting at roughly 17.8%. Let me put that in perspective. A pure SaaS company runs 70-80% gross margins. OpenAI, even with its massive inference load, is reportedly operating in the 50-60% range. MINIMAX is running at 17.8%. That's not a software company. That's a manufacturing company with a GPU supply chain. Every dollar of revenue costs 82 cents to produce. In this market, that's a red flag, not a green one.
The Core: Reading The Unit Economics
Based on my audit experience with DeFi protocols, I learned that you don't trust the headline APY. You trace the collateral. Here, the collateral is compute. The gross profit growth of 464.8% versus revenue growth of 283.1% tells me that they are getting more efficient at inference. Quantization, distillation, architecture optimization—something is working. The cost per token or per video frame is dropping. That's the good news.
The bad news is the absolute level. The 82.2% cost of revenue is brutal. If MINIMAX's revenue is primarily driven by video generation—which is a compute-intensive nightmare—then they are essentially buying GPU cycles at retail price and reselling them at wholesale. That's not a sustainable model unless they control the hardware.
Let's do the math on the loss structure. A $358 million half-year loss on $117 million in revenue. That's a burn ratio of over 3:1. They are spending three dollars to make one. The loss narrowed by 11%, which is positive, but it's not narrowing fast enough to suggest a clear path to break-even. At this rate, they are burning roughly $60 million a month. If they have $500 million in the bank, that's less than a year of runway. They need either a massive uptick in revenue or a new round of capital.
The Contrarian Angle: The 'Cheap AI' Trap
Everyone is celebrating the growth. I'm looking at the pricing strategy. A 17.8% gross margin suggests they are buying market share with low prices. This is the classic 'subsidize the TVL' play that I saw in DeFi in 2020. Projects offered 1000% APY to attract liquidity, and when the incentives stopped, the users vanished. MINIMAX might be doing the same thing with API pricing. They are likely undercutting OpenAI and domestic competitors like ByteDance and Zhipu to attract developers.
This works in a bull market for AI. But the moment funding dries up, or a bigger player decides to price them out, the revenue will evaporate. The gross margin is the moat. And 17.8% is not a moat. It's a puddle. I didn't lose $110,000 in 2017 by trusting narratives; I lost it by ignoring the unit economics. The narrative here is "AI video is the future." The reality is that MINIMAX is currently a high-volume, low-margin reseller of cloud compute with a very expensive AI wrapper.
Another blind spot is the competitive pressure. ByteDance has Jimeng, Kuaishou has Kling, OpenAI has Sora. These are giants with infinite capital and captive distribution. MINIMAX is trying to out-execute them in a niche. They might win the technology race for a quarter, but the moment Sora or Kling reaches parity on quality, MINIMAX's pricing power evaporates. They have no ecosystem lock-in. They have no switching costs. They have a feature, not a platform.
The Takeaway: Watch The Margin, Not The Headlines
The key metric for MINIMAX over the next 12 months is not revenue growth. It's the gross margin. If they can push from 17.8% to 30% through inference optimization and scale, they have a viable business. If it stays below 25%, they are in a commodity business, and the valuation will compress hard.
The IPO signal is also interesting. Filing financials in Hong Kong suggests they are prepping for a listing. If they can tap the public markets, they extend the runway. But the market is bearish on unprofitable tech. They need to show a credible path to profitability, and a 17.8% gross margin is not that path.
Every crash is just a story that hasn't been told yet. The story here is whether MINIMAX can transition from a high-growth compute reseller to a high-margin software company. The next earnings report will tell us. If gross margin stays flat, it's a short. If it jumps to 25%+, it's a long. The data will speak. It always does. t saying.
