Gross profit grew 464.8%. Revenue grew 283.1%. Losses narrowed 11%. The market will read this as a growth story. I read it as a structural signal about the cost of AI video inference and the brutal unit economics of the current AI application layer.
MiniMax, the Chinese AI startup backed by Tencent and Alibaba, released its H1 2026 financials today. Revenue hit $117 million. Gross profit came in at $20.8 million. That math gives you a gross margin of roughly 17.8%. For context, a healthy SaaS business runs at 70-80%. OpenAI operates in the 50-60% range. MiniMax is burning 82 cents of every dollar on direct costs, mostly compute. This is not a software company. This is a compute reseller with a very thin spread.
Let me break down what the numbers actually say, beyond the headline growth.
The Margin Expansion Signal
The 464.8% gross profit growth against 283.1% revenue growth is the only genuinely positive signal in this report. It means the cost per unit of output is falling faster than the price per unit of output. This is classic scale economics kicking in. Model quantization, distillation, better inference scheduling, or simply negotiating harder with cloud providers — something is working. The unit economics are improving, but they are improving from a dangerously low base.
I have seen this pattern before. In 2020, I modeled the yield curves on DeFi protocols and found that 90% of the APYs were fueled by inflationary token emissions, not real revenue. The same structural skepticism applies here. When a company grows gross profit at 464.8% but still sits at 17.8% gross margin, you have to ask: is this a business, or is this a subsidized acquisition engine? The growth is real. The question is whether the economics can ever catch up to the scale.
The AI Capex Trap
MiniMax is an AI video generation company. The Hailuo series is their flagship. Video inference is computationally brutal. Generating a minute of high-resolution video requires thousands of GPU-hours. This is not text generation where you can optimize your way to a 60% margin. The cost curve for video is steep, and the competitive pressure to offer better quality at lower prices is relentless. ByteDance's Jimeng, Kuaishou's Kling, and OpenAI's Sora are all fighting for the same developers and the same enterprise budgets.
This is where the liquidity-first analysis kicks in. In crypto, we watch the pipes — the movement of stablecoins, the velocity of tokens, the flows through DEXs. Here, the pipes are GPU clusters and cloud contracts. MiniMax is likely spending around $96 million per half on compute. That is not a sustainable path to profitability unless the cost per video generation drops by an order of magnitude. The infrastructure burden is the story. The narrative around AI disruption is secondary.
The Growth at All Costs Model
A 283.1% revenue growth rate is impressive by any standard. But the absolute scale is still small. Annualized revenue is around $230 million. OpenAI is doing over $10 billion. This is a $2.3 billion company playing in a $100 billion arena. The growth rate will inevitably decelerate as the base gets larger. The question is whether the margin expansion can outpace the growth deceleration. Based on my experience auditing token utility metrics and identifying liquidity traps in 2017, I can tell you that when growth is the only metric that matters, the structural flaws get hidden. The floor eventually breaks.
The Contrarian Read: This is Not a Warning, It's an Opportunity
Here is where I diverge from the bearish consensus. The market will look at the 17.8% gross margin and say this company is uninvestable. I see something different. The 464.8% gross profit growth rate tells me the engineering team has figured out how to radically reduce inference costs. The 11% loss narrowing tells me they are not just throwing money at the problem. They are getting more efficient. The margin is low because they are choosing to reinvest aggressively into market share. This is a strategic choice, not a failure.
Think of it like an infrastructure play in crypto. When a Layer 1 protocol launches with low fees and high subsidies, the initial metrics look terrible. But if the user adoption curve is steep enough, the network effects kick in and the unit economics eventually normalize. MiniMax is playing the same game. They are buying market share in the AI video generation market before the giants fully wake up. The risk is that the giants have deeper pockets and can subsidize for longer. The opportunity is that MiniMax has a technical edge in video generation that the incumbents are struggling to match.
The Liquidity Question
MiniMax lost $358 million in the first half of 2026. Annualized, that is over $700 million in cash burn. The company will need to raise more capital or go public. A Hong Kong IPO is the likely path. The Hong Kong exchange has a special chapter for pre-profit tech companies. The valuation math is speculative, but if the market applies a 30-50x price-to-sales multiple on annualized revenue of $230 million, you get a valuation range of $7-11.5 billion. That is a meaningful outcome for early investors.
The bigger question is the cash runway. At this burn rate, MiniMax has maybe 12-18 months of capital left, assuming they have a few hundred million in the bank. The IPO window is the critical variable. If the market turns risk-off, the IPO gets delayed, and the burn rate becomes existential. Macro moves before you blink. Adjust.
The Takeaway: Watch the Gross Margin, Not the Headlines
The next 12 months will tell us whether MiniMax is a real business or a subsidized experiment. The signal to watch is the gross margin trajectory. If it climbs from 17.8% to 25% or 30% within the next two quarters, the unit economics are improving at a rate that justifies the growth. If it stalls or declines, the cost structure is fundamentally broken. The revenue growth is a lagging indicator. The gross margin is the leading indicator. Liquidity leaves first. Watch the pipes.
MiniMax has a real product in a high-growth market. The technical execution is solid. But the economics are still in the danger zone. The company is spending heavily to establish a position in AI video generation, and the bet is that the cost curve will eventually bend in their favor. It might. But in a market where compute is the ultimate scarce resource, the company with the best cost structure wins. MiniMax is not there yet. The race is still open.
