Disappointing Earnings Trigger Sharp Correction
US tech giants Alphabet, Meta, Amazon and Intel recently reported earnings, with some profits, cash flows and guidance missing market expectations, triggering sharp stock corrections as the market refocused on the ability to monetize AI investments. Among them, Alphabet shares fell below the annual moving average after earnings and formed a death cross, weakening the medium- to long-term technical outlook. Berkshire Hathaway's holdings in Alphabet have shrunk by about $7 billion from the end-June peak, about NT$224 billion, while Meta shares have been consolidating around the annual moving average.
Customers Concentrating Toward Large Clouds
Institutions point out that Musk previously warned that more than 200 emerging cloud computing vendors in the market face bankruptcy risk, including Nebius, CoreWeave and others. Such firms are themselves high-risk customers, and once their funding chains break, AI computing demand will concentrate with the well-funded four major cloud service providers: Microsoft, Google, Amazon and Facebook. Hon Hai, Quanta and Wistron were already tier-one contract manufacturers for these large vendors, and after the customer reshuffle bad-debt risk will fall, leaving long-term orders and profits better secured.
Computing Focus Shifts to Inference Applications
Over the past 2 years the market heavily purchased chips for large-model training, and now as large vendors such as Microsoft and Amazon face pressure to monetize AI applications, infrastructure is accelerating its shift toward inference applications. To reduce reliance on expensive Nvidia chips, North American vendors are expanding investment in self-developed inference chips, such as Google TPU, which will drive demand for TSMC advanced-process foundry services, as well as long-term business opportunities for ASIC design service firms such as MediaTek, Alchip and Global Unichip.
Hardware Upgrades and Long-Term Positioning Opportunities
The global data center expansion wave has not stopped despite the stock correction, with large vendors' capital expenditures still set to hit new highs this year and capacity remaining tight. As Nvidia's high-power cabinets enter volume production and generations change, manufacturing bottlenecks remain in areas such as liquid cooling, high-specification power supply units and high-bandwidth memory, and Taiwan makers that can overcome yield challenges and enter the high-end component supply chain will retain advantages in long-term product pricing and gross margins. Investment trust institutions estimate that driven by AI demand, Taiwan's technology supply chain will still post double-digit growth in overall corporate profits this year, with the medium- to long-term bullish foundation intact. The current sharp drop mainly reflects deleveraging and valuation correction, and when valuations return to reasonable levels, it will instead provide long-term value investors with positioning opportunities.