Hong Kong-listed AI computing power stocks experienced a strong upward surge. As of the time of writing, shares of MINIMAX-W (00100) rose 7.32% to HKD 352, while Z.AI (02513) gained 6.04% to HKD 1,264. GBA AI COMP (01396) climbed 6.03% to HKD 10.9, and GDS-SW (09698) advanced 5.06% to HKD 32.

The rally follows news that DeepSeek plans to raise pricing for its API services overall. Analysts at GF Securities noted that unlike previous peak-hour adjustments, this move targets the entire pricing structure, driven by surging demand from AI application growth and resulting tight supply of computing power. The price increase signals a shift in the large model industry's approach to price wars, with companies increasingly focusing on cost control and sustainable profitability. On the computing side, the rise of domestic large models directly fuels demand for local computing power, benefiting domestic AI chips and computing rental services.

Notably, GBA AI COMP recently announced that its subsidiary Shenzhen Hongrui had signed two financial leasing agreements with Puyin Financial Leasing, totaling approximately RMB 1.185 billion. The latest lease is valued at around RMB 790 million, with a term of 62 months. In the first half of the year, GBA AI COMP signed new contracts worth RMB 15 billion, bringing cumulative orders to over RMB 30 billion, with more than RMB 2 billion already delivered. An additional RMB 2 billion in equipment has been purchased and is awaiting delivery. With a strong order backlog and accelerating delivery pace, the company is well-positioned amid tight global AI computing supply and surging demand. Those who can quickly convert orders into operational computing assets will gain a competitive edge.

Analysts at CITIC Securities suggest that the current model competition is shifting from single-parameter expansion to collaborative evolution across pre-training, reinforcement learning, inference-time computing, RSI, and long-term Agent capabilities. Computing demand is expected to expand from training to inference and Agent execution. The firm remains bullish on ecosystem capital expenditure from major tech companies, domestic chips and supernodes, computing services, pre-AI infrastructure, B-end AI applications, and local inference capabilities.