Stronger-than-expected U.S. nonfarm payroll data has prompted an upward revision in pricing for this month's Federal Reserve rate hikes. Market attention remains fixed on investment opportunities within the artificial intelligence supply chain, where memory demand and shortages could persist for some time. However, the key takeaway is that the entire AI industry continues to invest across chips, packaging, computing power, and Hong Kong-listed technology firms, with the pace of benefits varying across segments.

Since July, market views on the memory cycle have diverged into two camps. The optimistic side argues that HBM has transformed the economics of capacity: supply cannot be quickly replenished by merely adding production lines, as capacity concentration, cleanroom constraints, and advanced packaging bottlenecks remain; leading players prefer process migration over significant wafer expansion; and long-term contracts now account for a higher share, with demand extending from training to inference and agentic workloads, offering better visibility than smartphones or personal computers.

The conservative side counters that gross margins approaching 80% are themselves a signal for expansion. Long-term contracts are not infrequently renegotiated during sharp price declines, markets typically price in oversupply well ahead of new capacity arriving in 2028, and elevated prices are already curbing consumption demand.

Closer to reality, AI demand can push shortages higher and extend their duration, yet it remains insufficient to equate memory with a non-cyclical manufacturing asset. A more likely scenario is that supply remains tight through 2026 to 2027, price increases may moderate, but earnings can still sustain elevated levels; share price volatility and low valuations often occur because markets are already pricing based on the late-cycle stage. Simultaneously, concentrating solely on a memory leader means betting on both how long the shortage lasts and when the market stops applying late-cycle discounting, making timing difficult to master.

Within the AI supply chain, it is not just memory that is strengthening. Optical communications are also drawing interest, indicating a positive market outlook on AI infrastructure demand. The beneficiaries span the entire capital expenditure spectrum: computing power, foundry services, equipment and packaging testing, memory, as well as cloud and end-user applications. Memory can represent a lagging segment in pricing, but it need not be the sole direction. Shifting the focus from "guessing the cycle peak" to "covering the multiple bottlenecks through which capital expenditure flows" aligns better with forecasts: the cycle remains intact, while peak timings vary across segments.

Under these circumstances, capturing AI development opportunities requires selecting quality targets across different links of the industry chain. 易方达(香港)Solactive亚洲半导体精选指数ETF (03486), 易方达(香港)富时人工智能精选指数ETF (03489), and 易方达(香港)香港交易所科技100指数ETF (03456) cover multiple nodes that benefit from the AI supply chain, targeting the entire chain rather than a single asset. Among them, the 易方达亚洲半导体ETF focuses more on Asia's semiconductor supply side, making it more sensitive to manufacturing and packaging testing; the 易方达人工智能ETF spans the global AI supply chain, incorporating global computing power and applications, which helps diversify risks associated with any single segment's cycle. The 易方达(香港)香港交易所科技100指数ETF corresponds to Hong Kong's technology ecosystem, offering investors a more accessible option to seize opportunities.