According to Morgan Stanley's July 13 report, CPO adoption in Scale-Up networks is expected after 2029, with only limited deployment in 2028. Market concerns about CPO delays are overblown; the technology requires rebuilding supply chains for packaging, optical engines, and lasers, with Nvidia's Feynman generation serving as the key adoption timeline.
Coppper interconnects can sustain operations for two more years through innovations in PAM4 modulation, DSP, and retiming. AI cluster scaling from 72 GPUs to 576 or 1,152 units is the core driver. In 2026, non-Nvidia Scale-Up ecosystems enter production, with AMD MI400, Amazon Trainium 3, and Microsoft Maia ramping. Astera Labs and Broadcom are positioned as early beneficiaries.