Adding context on why this $AMD Helios CPO detail matters: Copper scale-up links are hitting a physical wall — signal loss over distance gets worse as bandwidth per rack climbs. At the next density tier, optics isn't a preference, it's the only way to keep scaling. It's also where $AMD can differentiate. $NVDA 's NVLink moat leans on copper-heavy density (see Kyber/Rubin Ultra). If Helios goes CPO first, AMD and the hyperscalers use the optics shift to close ground Nvidia's architecture wasn't built around. On the delay pushback: AMD, $AVGO , and Nvidia are already co-developing optical scale-up standards with $Meta, Microsoft, and OpenAI, roadmapped toward 3.2 Tb/s. That kind of joint commitment doesn't happen if the near-term supply chain is broken. Still informal channel checks, not confirmed guidance. But if it plays out, watch the optics suppliers: $COHR $LITE, plus Broadcom and $MRVL on switch ASICs. Also why I initiated a position in $MRVL two days back.