The debate over whether co-packaged optics (CPO) will soon make traditional pluggable optical transceivers obsolete in AI data centers got a reality check this week. Analyst Serenity relayed comments from a photonics-industry CEO, made on an industry podcast called "Market Insights: Photonics & AI," arguing pluggables have roughly another decade of relevant life left, not months. The exchange started with a reader asking whether Applied Optoelectronics (AAOI) is in trouble as CPO, which mounts optical modules directly next to a processor's chip package instead of plugging them in at the edge of a server, gains traction. Serenity's answer: not exactly. AAOI is also developing CPO technology, it's simply further behind on commercializing it than some peers, so lagging on one architecture doesn't mean the company is being left out of the market entirely.
The CEO's broader framework, as relayed, is that the industry is heading toward multiple coexisting architectures rather than one clean winner. Alongside pluggables and CPO, a newer category called near-package optics (NPO), which sits closer to the chip than a pluggable but doesn't go as far as full co-packaging, is described as a major opportunity over the next five years. That means capital and R&D spending across the optical supply chain is spreading across at least three architectures rather than concentrating on CPO alone.
CPO will soon make pluggable transceivers obsolete as AI clusters scale up.
Pluggables stay commercially relevant for roughly a decade, coexisting with CPO and NPO rather than one winning outright.
The binding constraint across all three paths, per the CEO, isn't architecture choice but manufacturing capacity and raw material availability, likely a reference to the industry's tight supply of continuous-wave (CW) lasers used to power optical modules. That suggests demand for optical components is outrunning the industry's ability to build them, regardless of which technology wins. This is secondhand color from a podcast relayed through a tweet, not a company roadmap or earnings guidance, so it should be read as directional industry sentiment rather than a confirmed timeline for AAOI or any single supplier.
Why it matters · It pushes back on the assumption that CPO will quickly sideline pluggable optics suppliers like AAOI, and shows optical component supply, not architecture choice, may be the real bottleneck for AI data center buildouts.
Worth asking · If pluggables, CPO and NPO all coexist for a decade, which optical suppliers actually win?