In CY26Q2, the capital expenditure (Capex) of the four major North American cloud providers maintained rapid growth. Giants like Google and Amazon further revised their full-year guidance upward. Notably, Amazon AWS delivered performance that exceeded market expectations across revenue, operating margins, and backlog. Amazon CEO Andy Jassy directly addressed concerns regarding the ROI of massive AI investments, stating that the breakeven point for server and network equipment is "less than three years".
CITIC Securities points out that a virtuous loop of AI investment and return has begun to form, significantly easing market worries about the sustainability of massive AI #Capex. As cloud provider earnings continue to validate how AI drives business growth, AI cluster scales will expand further. Optical interconnects, as a vital link in cluster networks, will see sustained high-speed growth driven by increased GPU ratios, port speed upgrades, and the trend of "optical-over-copper," reinforcing the long-term prospects of the optical communication sector.
Meanwhile, domestic institutional funds in China are boosting their equity exposure during market adjustments. Data from Simu Pai Pai Wang shows that stock private equity funds have increased positions for four consecutive weeks, pushing overall positions to a nearly 220-week high. Mega private equity firms have signaled bullishness through intensive self-purchasing, totaling over 1.4 billion RMB since July. Analysts suggest that as risks dissipate, the value of equity asset allocation is becoming prominent, though the short-term market will likely remain volatile with structural opportunities.
[AgentUpdate Depth Analysis]: The robust Capex growth and clear under-three-year ROI timeline from North American cloud giants provide a solid financial foundation for the entire AI value chain. For the AI Agent ecosystem, this infrastructure boom is critical. Next-generation Multi-Agent systems require ultra-low latency and massive bandwidth to handle real-time reasoning and cross-agent communication. The expansion of high-speed optical interconnects directly mitigates network bottlenecks in large-scale cluster environments. Compared to static LLM APIs, dynamic Agentic Workflows demand continuous, intensive compute. The continuous optimization of underlying hardware efficiency and declining cost-per-token will accelerate the transition of AI Agents from experimental enterprise tools to ubiquitous, autonomous production forces in the global economy.
