MiniMax Group, the Chinese AI company listed in Hong Kong as HKEX: 00100, reported Wednesday that its revenue nearly quadrupled in the first half of 2026, as surging demand for enterprise AI services transformed the startup's top line.
Total revenue for the six months ended June 30 reached US$116.6 million, up 283.1 percent from US$30.4 million in the same period last year, according to the company's unaudited results. That figure already exceeds MiniMax's revenue for all of 2025, which came in at US$79.0 million. For more context on this story, see our ongoing AI news.
Enterprise API revenue is the engine
The standout number in the release is the Open Platform, MiniMax's enterprise and developer business. Revenue there grew 703.1 percent year over year, from US$9.2 million to US$73.9 million, and now represents 63.4 percent of total revenue — roughly double the 30.3 percent share it held a year earlier.
The company attributed the explosion to growing numbers of paying individual and enterprise users, rising API call volumes, and rapid adoption of its Token Plan subscription offering. Revenue from AI-native consumer products, including the Hailuo AI video and content platform, grew 100.9 percent to US$42.6 million, driven by higher engagement and increased willingness to pay.
From consumer products to infrastructure
Looked at as a strategic story rather than a growth headline, the results mark a striking rebalancing of MiniMax's business. A year ago, consumer AI products like Hailuo carried the company, accounting for roughly 70 percent of revenue while the Open Platform contributed just 30.3 percent. Half a year later, the proportions have effectively flipped: enterprise services now deliver nearly two-thirds of every dollar MiniMax earns.
That inversion mirrors a wider pattern across China's AI sector, where the most durable monetization has come not from viral consumer apps but from developers and businesses wiring model APIs into their own products — a market where token-based pricing compounds quietly and churn is low. MiniMax's Token Plan, which lets developers pre-purchase inference capacity, appears to be the sharpest expression of that shift in the company's numbers.
Reuters, which first reported the results, framed the surge as evidence of blistering AI demand across China's technology sector, while Bloomberg cast it as the latest move in a crowded race for the country's AI lead — a contest that also features DeepSeek, Alibaba's Qwen family, and Moonshot AI. Reuters, which first reported the results, framed the surge as evidence of blistering AI demand across China's technology sector, while Bloomberg cast it as the latest move in a crowded race for the country's AI lead — a contest that also features DeepSeek, Alibaba's Qwen family, and Moonshot AI.
Profitability: better margins, bigger losses
Gross profit improved 464.8 percent to US$20.8 million, with gross margin rising from 12.1 percent to 17.9 percent — a jump the company credits to improving infrastructure efficiency.
The bottom line is more complicated. MiniMax's adjusted net loss widened to US$293.0 million for the half, from US$138.7 million a year earlier, as the company poured money into model development. Research and development expenses rose 138.8 percent to US$296.9 million, driven by cloud services costs for training as MiniMax iterates on its foundation models and multimodal capabilities.
The company emphasized that R&D grew significantly slower than revenue, and pointed to other efficiency gains: selling and distribution expenses fell 17.9 percent to US$27.0 million as it pursued organic user growth over paid promotion, and administrative expenses, though up 103.7 percent to US$30.2 million, fell from 48.8 percent of revenue to 25.9 percent.
MiniMax is well capitalized to keep burning. Its cash balance stood at US$1,322.8 million as of June 30, up from US$1,050.3 million at the end of 2025.
'Intelligence can scale almost without limit'
Co-founder and CEO Dr. Yan Junjie framed the results around the economics of intelligence rather than raw growth rates.
"Intelligence can scale almost without limit; energy and compute cannot. By July 2026, token consumption on MiniMax had grown to 20 times its January level," Yan said in the release. "That reinforces a belief we've held since day one: the long-term competition in AI is not just about building more powerful models, but about delivering higher levels of intelligence to more people at lower cost."
The company sums up the strategy as "Minimize the Cost, Maximize the Intelligence" — the mechanism, it says, for making "Intelligence with Everyone" possible.
The catch: pace, not direction
For all the momentum, the growth story has a caveat. The South China Morning Post noted that MiniMax's first-half performance still leaves the company behind the pace needed to meet its full-year forecast — a reminder that in China's AI sector, triple-digit growth is now the price of entry rather than a differentiator.
The competitive backdrop is unforgiving. Open-weight rivals keep compressing API prices, DeepSeek's models dominate developer mindshare, and Alibaba is pushing its Qwen family aggressively into the enterprise market. MiniMax's 703 percent enterprise growth suggests its Open Platform strategy is gaining traction — but with an adjusted loss of nearly US$300 million against US$116.6 million of revenue, the company is still spending roughly $2.50 on R&D for every dollar it earns.
There is also the question of what the widening adjusted loss buys. MiniMax is spending its capital on exactly one thing: model capability — foundation models and multimodal systems, trained at escalating cloud cost. If enterprise token consumption keeps growing the way it did in the first half, that spending is an investment in market share. If it flattens, it becomes the kind of burn rate that has felled plenty of well-funded challengers before it.
Whether that math converges depends on the bet at the heart of Yan's quote: that token demand keeps compounding faster than the cost of serving it. The first-half numbers are the strongest evidence yet that, at least for MiniMax, the demand side of that equation is real. The second half will show whether margin gains can outrun a research budget growing this fast.
---
Stay Ahead of AIGet the latest AI news, analysis, and breakthroughs — all in one place.
Read more AI news →
