Open-Source AI Models Advance Further Amid Persistent Computing Power Shortage: Focus on E Fund Global Artificial Intelligence ETF (03489)

The recent simultaneous launch of models like Kimi K3, DeepSeek V4, and Qwen 3.8 marks a significant industry milestone, following last year's 'DeepSeek moment'.

The Kimi K3 model, with its 2.8 trillion parameters and million-token context window, demonstrates capabilities in coding and agentic functions that approach or even surpass top global proprietary models in certain areas, contributing to an upward trend in API pricing.

The proliferation of open-source and cost-effective models is significantly reducing the cost of application calls, accelerating the large-scale adoption across sectors like finance, office productivity, and enterprise services. The primary competitive battleground is expanding from the models themselves to product strength and ecosystem development, benefiting the entire chain from cloud services and applications to hardware.

As model capabilities grow stronger, computing power becomes tighter. Following the launch of Kimi K3, request volumes within 48 hours far exceeded expectations. The developer, Moonshot AI, has publicly stated that computing power is nearing cluster limits, leading to a temporary suspension of new consumer member subscriptions to focus on capacity expansion.

This leap in model performance is shifting the scarcity of resources from the training phase to the inference and application phases. Cloud services and semiconductors remain the most certain fundamental anchors in this landscape.

The industry cycle of 'model breakthrough → surging usage → computing power strain → increased capital expenditure' continues to strengthen. During periods of market adjustment, it is prudent to focus on high-growth potential targets.

The E Fund Global Artificial Intelligence ETF (03489) closely tracks the FTSE Global Artificial Intelligence Index, providing exposure across Hong Kong and US markets and covering the entire industry chain from cloud and software to computing hardware.

It focuses on core Asian semiconductor and computing hardware companies, positioning it to benefit directly from the rising demand for chips, memory, and advanced manufacturing processes driven by the scaling adoption of these powerful AI models.