CPO could be one of the biggest shifts in AI infrastructure.

As AI clusters scale, traditional pluggable optics might just be a temporary fix. Future systems will need higher bandwidth, lower latency, and better power efficiency.

Some key challenges for CPO are thermal management, as AI chips generate massive heat while optical components are highly temperature sensitive, glass substrates requiring advanced manufacturing and precision via drilling, and laser integration where balancing heat, efficiency, and optical performance remains difficult.

Once reliable CPO solutions are commercialized, demand could accelerate rapidly, with supply becoming the main constraint.

Key players to watch include $Marvell Technology(MRVL)$ , $Amkor Technology(AMKR)$ , NVDA, $Taiwan Semiconductor Manufacturing(TSM)$ , $Lumentum(LITE)$ , INTC, and AAOI.

The next AI infrastructure race won't just be about compute, but also about moving data faster and more efficiently. Advanced packaging, optical engines, lasers, cooling, and optical switching could become critical growth areas.