Chinese artificial intelligence company MiniMax has significantly increased its cloud computing spending commitments with Alibaba Group Holding, reflecting growing demand for computing power as the firm expands its AI capabilities. The Shanghai-based company raised its three-year purchase ceiling for Alibaba Cloud services by 220 percent to US$1.2 billion, according to filings submitted to the Hong Kong stock exchange on Wednesday.

MiniMax had initially planned to allocate up to US$115 million for Alibaba Cloud usage this year but revised that figure to nearly US$300 million after consuming about two-thirds of its original budget by the end of June. Under the new arrangement, the company's annual spending limits are set to rise from US$125 million to US$400 million in 2027 and from US$135 million to US$500 million in 2028.

In addition to cloud services, MiniMax also expanded its budget for application programming interface (API) services with Alibaba, increasing its 2026 cap from US$650,000 to US$7.5 million. This adjustment raises the total three-year API spending ceiling to US$62.5 million, marking a nearly 20-fold increase from previous levels.

These expansions are driven by MiniMax’s heightened needs for computational resources to support both AI model training and live inference tasks. The company, recognized for its M-series large language models, the H3 video generation model, and the consumer-focused Hailuo AI app, reported a 283 percent surge in first-half revenue to US$116.6 million this year, propelled by a 700 percent increase in enterprise sales.

Founder and CEO Yan Junjie stated that upcoming models—including M3.1, M3 Pro, and H3.1—are nearing completion. To sustain its growth, MiniMax raised HK$16 billion (approximately US$2 million) in July through a share placement and convertible bond issuance, allocating around 80 percent of the net proceeds toward AI infrastructure and model development planned for 2027.

The company's revised agreements with Alibaba reflect a broader Chinese industry trend shifting focus from foundational AI model training to the scaling of compute-intensive inference workloads driven by agentic AI applications. Market research firm IDC projects China’s model-as-a-service sector to handle 40,000 trillion token calls this year.

Alibaba has reported a 45 percent increase in revenue from its cloud and AI operations for the quarter ending in June, marking its fastest growth in nearly six years. The company also commenced commercial-scale deployment of its Zhenwu M980 AI processor earlier this month.

MiniMax disclosed that purchases of Alibaba cloud services accounted for approximately 30 percent of its US$253 million research and development expenses last year, a ratio it expects to maintain as it expands infrastructure to support future operations. The company emphasized that Alibaba’s cloud products provide secure, stable, scalable, and flexible infrastructure crucial for its core business and foundational model development.

Beyond its deepening partnership with Alibaba, MiniMax is also building its own independent computing infrastructure. Yan revealed plans to establish a self-controlled computing cluster equipped with high-bandwidth networks, dedicated lines, servers, and storage to become a central technology asset. The company has started integrating domestically produced AI chips, which will comprise an increasing portion of its computing power starting in the fourth quarter of this year.