AI-themed ETFs advanced in early trading, with E Fund Asian Semiconductor ETF (03486) rising 6.26% to HK$15.96 and E Fund Artificial Intelligence ETF (03489) gaining 2.54% to HK$12.11 as of press time.
On the news front, on October 21, Musk outlined SpaceX's AI compute roadmap, targeting 13 gigawatts of compute capacity online by the end of next year, with an extreme scenario of up to 20 gigawatts. He set an annual recurring revenue goal of US$500 billion to US$1 trillion by the end of 2027, with roughly half of compute for internal use and half for external leasing. The external leasing approach favors spot arrangements rather than long-term contracts.
More critically, the rental benchmark stands out: cloud compute rental rates exceeding US$50 billion per gigawatt imply that the payback period for compute assets is being repriced, shifting neoclouds from a cash-burning narrative to a recoverable-cost narrative. Enthusiasm for AI hardware has been reignited, with Nebius and CoreWeave both strengthening in after-hours trading. At the same time, Grok's bot-based applications continue to consume inference compute, broadening the demand base.
It is worth noting that although the symbolic significance of these figures outweighs near-term verifiability, they do extend the AI hardware demand curve further out, providing a near-term rebound window for semiconductors and the compute supply chain. What ultimately determines the sustainability of the rally is the pace at which cloud providers' capex guidance, memory contract prices and HBM orders materialize.
It is understood that the E Fund Artificial Intelligence ETF tracks the FTSE AI Select Index, covering both U.S. compute and Hong Kong application segments, with heavy weights in compute leaders such as Nvidia, making it a broad-based tool to capture the expansion of the current AI demand narrative. The E Fund Asian Semiconductor ETF focuses on memory, foundry, advanced packaging and equipment, directly corresponding to the hardware increment brought by compute expansion and suitable for capturing the rebound elasticity of the semiconductor chain.
Comments