MiniMax Revenue Surges 283%, Missing Full-Year Forecasts

Chinese artificial intelligence firm MiniMax reported a 283 per cent surge in first-half revenue to US$116.6 million, powered by a 700 per cent jump in its enterprise business. This growth, however, falls short of full-year analyst forecasts. Performance indicates strong competition from American and Chinese AI laboratories.

Credit: UNSPLASH

Revenue for the six months ending the first half amounted to approximately 32 per cent of the US$363.77 million analysts anticipated for the current financial year, according to Bloomberg. The organisation had reported US$79 million in revenue for the prior financial year.

So, MiniMax's top-line expansion stemmed primarily from its enterprise-facing operations. Income from its Open Platform and other AI enterprise services climbed over 703 per cent year on year, rising to US$73.9 million from US$9.2 million. This segment made up 63.4 per cent of total revenue, up from 30.3 per cent a year earlier, reflecting growth in paying users and corporate clients. Revenue from other AI-native products also doubled.

The firm's total loss for the period reduced by 11 per cent to US$358 million. Adjusted net loss, however, grew by 111.2 per cent, reaching US$293 million from US$139 million reported a year ago. Gross profit increased five-fold to US$20.8 million from US$3.7 million. Gross profit margin improved to 17.9 per cent from 12.1 per cent a year earlier. Hong Kong-listed MiniMax shares closed up 1.13 per cent at SGD 52.72 before the earnings announcement this week.

Shanghai-based MiniMax, established in 2021, is known for its M-series large language models, its recent H3 video-generation model, and popular consumer applications such as Hailuo AI. The company stated earlier in the year that its annual recurring revenue, a metric projecting monthly subscription income, reached approximately US$300 million, doubling since early in the year.

But the firm's annual recurring revenue increased to US$800 million by late this year, driven by enhancements in model capabilities, according to Yan Junjie, MiniMax's founder and Chief Executive Officer. Corporate customers contributed 80 per cent of MiniMax's annual recurring revenue. The company's total token consumption recently increased 20-fold from the beginning of the year.

MiniMax and Beijing-based Zhipu AI, China's first two publicly traded large language model developers, saw valuations increase following initial public offerings earlier in the year. Both now trade at a fraction of their peak share prices, with MiniMax's shares falling about 77 per cent from its high earlier in the year. Zhipu AI also saw its share price significantly reduced.

Both organisations face substantial competition from Chinese AI model developers, including Moonshot AI and DeepSeek. They also strive to match US frontier laboratories such as Anthropic and OpenAI. JP Morgan analysts noted MiniMax's latest M3 large language model, introduced recently, "yet to establish a clear advantage on either capability or cost-performance," leaving it "vulnerable to stronger models at one end and DeepSeek's historically aggressive economics at the other."

And the upcoming M3.1 model will serve as a significant test. JP Morgan analysts Olivia Xu, Alex Yao, and Daniel Chen stated it will determine if its large language model can "regain relevance on the capability-cost frontier." Yan also mentioned MiniMax's next models, M3.1, M3 Pro, and H3.1, are nearly completed.

MiniMax possesses adequate computing power for its M3 Pro and video models' training and inference processes. The company is constructing an "independent and controllable" computing cluster. Yan added the organisation has begun utilising domestically produced AI chips, expecting these to account for a greater proportion of its computing power from the final quarter of this year.

• MiniMax reported a 283 per cent first-half revenue surge to US$116.6 million, largely from a 700 per cent jump in enterprise business.

• The company's top-line growth remains below full-year analyst forecasts amidst intense AI sector competition.

• MiniMax and Zhipu AI, China's leading large language model developers, currently trade at fractions of their post-IPO peak share prices.

• JP Morgan analysts suggest MiniMax's M3 model has not established a clear advantage, making upcoming models critical for its market position.

• The organisation is building an independent computing cluster and plans to increase its use of domestically produced AI chips.

Source: SCMP