Morgan Stanley: CPO adoption has been pushed back to 2029; copper interconnects can be maintained through 2028. According to Morgan Stanley’s July 13 report, CPO in the Scale-Up network is expected to begin deployment after 2029, with only limited deployment in 2028. Market concerns about CPO delays are being exaggerated; the technology requires a supply chain for packaging, optical engines, and laser reconstruction, and Nvidia’s Feynman generation timing serves as the key adoption schedule. Copper interconnects can be kept running for another two years through innovations such as PAM4 modulation, DSP, and retiming. Expanding the AI cluster size from 72 GPUs to 576 or 1,152 GPUs is the core driver. By 2026, non-Nvidia Scale-Up ecosystems will enter mass production; AMD MI400, Amazon Trainium 3, and Microsoft Maia are accelerating ramp-up. Astera Labs and Broadcom are positioned as the earliest beneficiaries.