MINIMAX's 17.8% Gross Margin Exposes the Ugly Math Behind 283% Revenue Growth
Revenue up 283.1%. Gross profit up 464.8%. Net loss narrowed 11%. Three numbers from MINIMAX's H1 2026 report, and the market will read them as a growth story. I read them as a warning.
Here is the math nobody is doing: gross margin is 17.8%. That means for every dollar of revenue, 82 cents goes to direct costs—mostly compute. This is not a software company. This is a hardware company wearing an AI costume.
Context: MINIMAX is a Chinese AI startup betting heavily on video generation—the Hailuo series. They operate in the most compute-intensive corner of the AI market. Video generation inference costs are an order of magnitude higher than text models. The 283% revenue growth is real, but it's built on a foundation that burns capital at a rate that would make most venture funds flinch.
Let's break down the core data. Half-year revenue: $117 million. Gross profit: $20.8 million. Loss: $358 million. That's a loss-to-revenue ratio of 3:1. They are spending three dollars for every dollar they bring in. The loss narrowing is technically true—11%—but when your revenue triples and your loss barely moves, that's not improvement. That's scale without efficiency.
I've seen this pattern before. In my DeFi arbitrage days, we had a rule: if your gross margin is below 30%, you are not building a business, you are renting market share. MINIMAX is renting. The question is whether they can convert that rental into ownership before the capital runs dry.
The contrarian angle here is uncomfortable: the headline growth is actually a liability. When you grow at 283% with 17.8% gross margins, you are locking in a customer base that expects a certain price point. Raising prices later is not a strategy—it's a churn event waiting to happen. The smart play would have been slower growth with better unit economics. But in this market, slower growth means you disappear. So they're trapped.
Compare this to the competition. OpenAI runs gross margins around 50-60%. Anthropic is in a similar range. MINIMAX's 17.8% puts them in a fundamentally different category—not a premium AI company, but a commodity infrastructure provider. Their differentiation is supposed to be video generation, but that's the most expensive inference workload in the industry. The worse the technology is at cost containment, the more revenue it needs to generate to survive.
Based on my audit experience—I spent three months dissecting 0x protocol's smart contracts in 2017 looking for exactly this kind of structural inefficiency—the problem here is not technical. It's architectural. MINIMAX has built a business model where success (more users) directly increases their burn rate. That's not a growth company. That's a charity for GPU manufacturers.
Here's what nobody tells you: the AI video generation market is a trap. It's expanding, yes, but the cost curves are brutal. Every video generation query requires hundreds of times more compute than a text prompt. The industry is competing on price while the underlying input costs remain tied to chip supply constraints. MINIMAX, as a Chinese company, is also navigating export controls that limit access to the best hardware. They are fighting with one arm tied behind their back.
The data shows a company with 3.58 billion in half-year losses. The entire market cap potential, based on PS multiples, sits between 23 and 115 billion—but that range assumes they can fix the margin problem. If they can't, the equity is worth a fraction of that.
Efficiency eats sentiment for breakfast. The sentiment is bullish. The efficiency metrics are not. The gross margin trend is the only number that matters for the next 12 months. If it stays below 25%, this is a subsidized acquisition funnel, not a business. If it crosses 30%, they might have something.
Spread the truth, not the panic. But understand this: MINIMAX is a bet on one thing—that video generation inference costs drop by 50% in the next 18 months. If that happens, their unit economics transform. If it doesn't, they're holding a burning GPU bill with no way out.
Data doesn't lie; emotions do. The emotional narrative is AI growth. The data narrative is margin compression. I'll trade on the data.
The takeaway: watch the gross margin line. Not the revenue growth, not the loss narrowing, not the market narrative. Code is law; liquidity is life. Right now, MINIMAX's liquidity is being drained by their own success. Whether they survive depends on whether they can flip 82 cents of cost per dollar of revenue into something closer to 60 cents. That's the entire ballgame.