From June 1 to 5, GTC Taipei and Computex 2026 were held intensively, and NVIDIA coincidentally announced the RTX Spark super chip, officially entering the PC processor market by integrating CPU and GPU on the same chip sharing unified memory. This means AI computing power is shifting from "cloud-centric" to "end-user widespread," and TSMC's shareholder meeting also emphasized the same trend—AI is moving from generative queries to agent command execution, with token consumption levels expected to increase significantly.

But the more critical validation window is coming up: on June 8, Apple’s WWDC will focus on whether edge-side AI can catch up; on June 24, NVIDIA’s annual shareholder meeting and Micron’s earnings report will correspond to AI factory and storage HBM demand. The entire June is a key period for the AI industry chain.

Additionally, silicon photonics technology has entered full-scale mass production. NVIDIA Spectrum-X Ethernet solution introduces CPO packaging, integrating the optical engine directly next to the switch chip, greatly reducing latency and power consumption—optical interconnects are becoming the new bottleneck for computing clusters, and the resilience of this industry chain is also worth monitoring. #分享美股交易赢英伟达股票