Hong Kong Stock Concept Tracker｜Oracle (ORCL.US) RPO Surge Ignites AI Computing Power Chain—Domestic Opportunities in Focus

Global technology leaders have recently signaled an acceleration in AI infrastructure deployment. On September 10 (U.S. Eastern Time), Oracle (ORCL.US) reported its fiscal 2026 first-quarter results, revealing that while revenue and earnings per share slightly missed analysts’ forecasts, Remaining Performance Obligations (RPO) soared to USD 455 billion, up 359% year-on-year. This disclosure sent Oracle shares up nearly 36%—their largest single-day gain since 1992—and lifted the company’s market capitalization to approximately USD 933 billion at one point.

Oracle’s cloud infrastructure business, supported by access to NVIDIA GPUs for large-scale AI workloads, has positioned the company as a key beneficiary of the AI boom. Management highlighted that the AI inference market now dwarfs the training segment, and available inference compute capacity is being rapidly consumed. By leveraging its enterprise database leadership, Oracle enables organizations to securely combine private and public data for large language model inference while preserving data privacy.

Looking ahead, Oracle projects its cloud infrastructure revenue for fiscal 2026 to reach USD 18 billion—an increase of 77% over the prior year. Over the subsequent four years, the company expects this figure to grow to USD 32 billion, USD 73 billion, USD 114 billion, and USD 144 billion, respectively. CEO Safra Catz noted that most of these anticipated revenues are already reflected in RPO, underscoring Oracle’s strong entry into the new fiscal year.

Analysts see significant upside for Chinese firms that directly support global AI supply chains. With China’s comprehensive industrial ecosystem, the country holds advantages in large-scale AI implementation, and segments of its domestic AI industry chain may remain undervalued.

NVIDIA has also unveiled new initiatives. The company plans to release Rubin CPX, a product designed to accelerate complex tasks such as video generation and software development. Scheduled for launch by the end of 2026, Rubin CPX will be available in card form for integration into existing server designs or as part of standalone high-performance systems. This addition to the Rubin product family promises up to 6.5 times the performance of the current GB300 NVL72 rack when handling large context windows. CEO Jensen Huang described CPX as the first chip optimized for one-pass inference on models requiring millions of tokens.

Meanwhile, Microsoft and Nebius have agreed to a USD 19.4 billion multi-year AI cloud computing partnership, including a USD 2 billion optional extension through 2031, and Meta’s Mark Zuckerberg announced plans to invest approximately USD 600 billion in AI infrastructure by 2028. These commitments from industry leaders are accelerating the shift from anticipated AI deliveries to actual performance rollouts.

Policy support is also mounting. On September 9, Zhang Yunming, Vice Minister of the Ministry of Industry and Information Technology, stated that the ministry will drive high-quality AI industry development, accelerate intelligent industrialization, and issue an “AI + Manufacturing” action plan outlining tasks for key sectors and publishing guidelines for AI applications in manufacturing enterprises. Concurrently, Hangzhou’s Bureau of Economy and Information Technology released a draft three-year action plan targeting a RMB 300 billion AI terminal industry by 2027, promoting high-performance integrated AI computing systems and edge servers powered by domestic AI chips.

Guojin Securities cites IDC estimates forecasting China’s AI market to reach RMB 815.9 billion by 2028, with a five-year CAGR of 33%. The industry chain follows a gradient of upstream compute capacity expansion, midstream model and MaaS commercialization, and downstream application proliferation driven by enterprise-level AI Agents. Technological breakthroughs, global and domestic demand, and robust capital expenditure are converging to establish a clear upcycle in AI, centered on compute supply, MaaS platforms, and industry-specific AI solutions.

Related Concept Stocks:

The Southern Fund Asia-Pacific Select ETF (159687.HK) blends high-quality dividend assets and leading semiconductor companies across the region, including Toyota Motor, Tencent Holdings, Alibaba, Mitsubishi Group, Taiwan Semiconductor Manufacturing Company, Samsung Electronics, Tokyo Electron, and MediaTek. According to a 2024 Deloitte & GSA report, the Asia-Pacific region accounted for 57.6% of global semiconductor industry revenue in 2022.

Baidu Group-SW (09888.HK) introduced its Wenxin large model X1.1 at the 2025 WAVE SUMMIT Deep Learning Developer Conference. Accessible via the Wenxin Yiyan portal, Wenxiaoyan app, and Baidu Smart Cloud’s Qianfan platform, X1.1 delivers a 34.8% improvement in factual accuracy, 12.5% better instruction adherence, and a 9.6% increase in agent capabilities versus X1. It has outperformed DeepSeek R1-0528 in multiple benchmarks and matches the performance of GPT-5 and Gemini 2.5 Pro on certain tasks.

Alibaba-SW (09988.HK) retains an “Outperform” rating from Guosen Securities, supported by its extensive AI model portfolio and application integration. The Qwen series leads domestic model development, and Alibaba is embedding AI across core services such as Taobao, Quark, DingTalk, and AutoNavi to serve both B2B and B2C use cases.

SenseTime-W (00020.HK) received a “Buy” rating upgrade from Goldman Sachs following the State Council’s “AI +” initiative. SenseTime reported first-half revenue growth of 36% year-on-year and a 73% surge in its generative AI segment, buoyed by productivity and interactive tools. Goldman projects this segment’s contribution to rise from 64% in 2024 to 91% by 2030.

Kingboard Integrated Circuits (01888.HK) saw copper-clad laminate prices rise in the first half of the year amid strong PCB demand. With high-end CCL and material capacity expansion accelerating for 2026 and AI-related options entering a realization phase, valuation fundamentals are expected to strengthen, with significant earnings contributions concentrated in 2027.

Kingboard Holdings (00148.HK) reported mid-2025 revenue of HKD 21.608 billion, up 6% year-on-year, and net profit of HKD 2.582 billion, up 71%. Basic earnings per share stood at HKD 2.329, accompanied by an interim dividend of HKD 0.69 per share. Its subsidiary Kingboard Laminates achieved HKD 9.588 billion in revenue, up 11%, and HKD 933 million in net profit, up 28%, with basic EPS of HKD 0.299 and a proposed interim dividend of HKD 0.15 per share.