Overnight, the US compute supply chain came under broad pressure: the Nasdaq fell 1.25%, the Philadelphia Semiconductor Index dropped 3.39%, with Micron down 4.79%, Nvidia down 2.94%, and Oracle down 5.5%. The trigger was OpenAI disclosing to investors that its annualized revenue reached nearly $50 billion by the end of September, below the $70 billion widely reported by media previously. The market interpreted this as an expectation gap in the AI demand narrative, compounded by oil prices surging more than 4% and the 10-year US Treasury yield remaining elevated, putting concentrated pressure on high-valuation assets.

However, this expectation gap stems from differences in statistical methodology and short-term sentiment disruption, rather than a weakening of AI industry fundamentals. During periods of market volatility, portfolio tools are better suited to smoothing fluctuations. E Fund AI (03489) and E Fund Asia Semiconductor (03486), with their precise sector positioning and full-chain coverage capabilities, see their allocation value further highlighted at present.

According to disclosures, the $70 billion figure originated from investors attempting to "gross up" OpenAI's annualized revenue for direct comparison with Anthropic. The two companies use different accounting methods: Anthropic includes revenue from cloud partners such as AWS and Google Cloud, while OpenAI does not. The new presentation showed annualized revenue approaching $30 billion in July and nearly $50 billion in September, still representing over 60% growth in two months. The lack of unified standards in statistical methodology is the direct source of this expectation gap.

It is worth noting that industry data disclosed on the same day does not support weakening demand, forming a clear divergence from short-term market pessimism, with upward industry prosperity verification still trending higher: TSMC reported September revenue of NT$511.86 billion, up 54.6% year-on-year, with Q3 2026 revenue of approximately NT$1.49 trillion, setting a quarterly record high and exceeding expectations; Samsung reported Q3 revenue of 195 trillion Korean won (+126.6%) and operating profit of 107.4 trillion Korean won (+782.5%). On the memory side, under TrendForce's estimates, general-purpose DRAM prices are expected to rise another 10%-15% in Q4, HBM pricing may more than double by 2027, and 75% of Micron's 2027 capacity has already been locked up through long-term agreements.

On the model side, Anthropic released Claude Haiku 5.5, with operating costs reduced by approximately 75% compared to the previous generation and context expanded to 1 million tokens; OpenAI launched a GPT-6 interactive interface on the same day, covering 1.2 billion weekly active users. Capability improvements and cost reductions are advancing simultaneously, and the direction of inference compute demand remains unchanged.

The current situation more closely resembles a narrative downcycle vacuum, where sentiment sensitivity exceeds fundamental changes. A near-term catalyst worth tracking is Anthropic's Investor Day on October 14. As a pre-IPO investor communication event, if disclosed revenue and order data are positive, it could stage a recovery in market expectations for AI commercialization pace.

However, two risks warrant attention: first, elevated long-end US Treasury yields continue to pressure high-valuation growth stocks, and the race between valuation and earnings still leans toward the former facing pressure; second, OpenAI itself expects cumulative consumption of nearly $280 billion by 2030, and changes in its private financing valuation and IPO timeline could still repeatedly disrupt market sentiment.

During a narrative vacuum, single events have a stronger amplifying effect on sentiment than changes in fundamentals themselves, making portfolio tools preferable to individual stock bets. E Fund AI (03489) covers US compute leaders and Hong Kong hard-tech companies, with Nvidia, TSMC, Micron, Alibaba, and Tencent all among its constituents, suitable for capturing the entire AI chain through index exposure during periods of sentiment disruption, avoiding amplified volatility from single-narrative targets while steadily seizing long-term growth opportunities in the AI industry.

E Fund Asia Semiconductor (03486) focuses on high-prosperity segments including memory, foundry, advanced packaging, and equipment, with SK Hynix and TSMC as core holdings, serving as a pure hardware tool for capturing memory and wafer manufacturing prosperity. The two funds respectively cover the entire AI industry chain and core semiconductor hardware sectors, forming differentiated allocation advantages, providing investors with steady and efficient allocation tools amid short-term sentiment volatility and long-term industry uptrend.