Domestic Compute Power Rises As Five AI Chip Firms Forecast Revenue Doubling; Hong Kong Leaders Rally Over 16%

On April 16, Hong Kong shares tied to compute power advanced broadly, reflecting investor optimism about the AI hardware segment amid accelerated model iteration and commercialization across the industry. As of the time of publication, Biren Technology（06082.HK）was up more than 26%, TianShu ZhiXin（09903.HK）rose over 16%, and Cambridge Technology（06166.HK）along with several peers climbed more than 5%.

The market move coincided with a fresh wave of price increases for compute services. On April 15, Alibaba Cloud issued a notice raising prices for certain model unit services on its “Bailian” platform by 2%–7%. Media reports indicate that other major providers, including Anthropic as well as Tencent and Baidu, have also implemented or signaled increases in compute‑service fees.

Demand for compute has accelerated sharply since the start of 2026, producing a steep growth trajectory. Data show that weekly token consumption on OpenRouter, the world’s largest API aggregation platform, has risen roughly seven to eight times compared with the same period a year earlier. This new iteration cycle for large models, together with faster commercialization, has created immediate supply gaps and upward pricing pressure that support industry fundamentals.

Concurrently, the rise of domestic compute solutions is becoming more pronounced. IDC reports that China shipped about four million AI accelerator cards in 2025, of which domestic vendors accounted for approximately 1.65 million units, representing a 41% market share. Looking to 2026, Wind consensus forecasts indicate that five listed AI chip companies, including Cambricon, Biren Technology and TianShu ZhiXin, are expected to see revenue increase by about 120%, reaching roughly RMB 25.7 billion.

Market indicators in the compute rental sector corroborate these trends. SemiAnalysis data show that one‑year rental rates for H100 GPUs climbed from about USD 1.70 per hour per GPU in October 2025 to USD 2.35 per hour per GPU in March 2026, an increase approaching 40%, underscoring persistent supply‑demand tightness. At the same time, product innovation continues: Alibaba’s ATH business group has opened early access to “Happy Oyster,” an open‑world multimodal model that supports audio‑video generation and enables continuous user interaction during content creation, delivering real‑time responses and dynamic rendering.

Industry research suggests that as AI applications are deployed at scale, inference demand will expand rapidly and further strengthen the case for domestic compute capacity. Huatai Securities notes that constraints on overseas compute providers entering the Chinese market have reinforced the shift toward local solutions, supporting a structural acceleration in demand for domestically produced AI hardware.