The co-packaged optics (CPO) market is entering a breakout phase as AI networking hits a wall with traditional interconnect technology. New projections point to explosive growth ahead, with CPO expected to expand at a 142% CAGR from 2026 to 2030 (excluding ELS). The reason is simple: as AI clusters scale and next-generation 3.2T data rates become the norm, copper-based connections and conventional pluggable transceivers are being pushed beyond their practical limits.
In today’s large AI deployments, pluggable transceivers still dominate most scale-out network designs because they’re familiar, broadly supported, and fit existing infrastructure models. But the 3.2T era changes the math. Power draw, heat, and signal integrity challenges rise quickly as bandwidth demands climb, and the efficiency gap between pluggables and integrated approaches becomes harder to ignore.
That’s where co-packaged optics steps in. CPO integrates optical engines much closer to the processing silicon, rather than relying on longer electrical paths and removable modules at the edge of the system. By shrinking the electrical reach and moving optical connectivity nearer to compute and switching silicon, CPO can significantly improve power efficiency while delivering higher bandwidth density. For AI networking, that translates into more scalable architectures and a better chance of maintaining performance without runaway energy costs.
Momentum is already building on the commercialization side. Major vendors including NVIDIA and Broadcom have begun deploying CPO in switch products, a meaningful signal that co-packaged optics is moving from a promising concept into real-world, revenue-generating deployments. Their interest highlights where the industry is headed: more efficient, higher-bandwidth networking designed specifically for the needs of rapidly growing AI clusters.
Looking forward, the biggest driver of CPO adoption is expected to be scale-up networking, not just traditional scale-out designs. As standards and ecosystems mature—supported by industry frameworks such as OSI MSA—scale-up CPO is projected to accelerate rapidly and surpass scale-out applications before 2030, ultimately becoming the dominant segment of the CPO market.
In short, co-packaged optics is emerging as a critical technology shift for AI infrastructure. As data rates surge and power efficiency becomes a defining constraint, CPO offers a practical path to higher bandwidth density, improved energy performance, and more resilient next-generation switching platforms—exactly the combination hyperscalers and AI platform builders need for the years ahead.
