3D optoelectronics and co-packaged optics: when solving the wrong problems stalls deployment

TL;DR

This paper discusses how focusing solely on component-level improvements in co-packaged optics can hinder their deployment in AI datacenters, emphasizing the need for holistic system-level design and standardization.

Contribution

It reframes co-packaged optics as an architectural strategy rather than just a component, highlighting the importance of system-level considerations for successful deployment.

Findings

Co-packaged optics reshape system design boundaries.

Trade-offs in integration strategies impact scalability.

Standardization and thermal management are crucial for adoption.

Abstract

The rapid growth of AI and accelerator-driven workloads is forcing a fundamental rethinking of optical interconnect architectures in datacenters. Co-packaged optics and three-dimensional photonic integration have emerged as promising solutions to overcome the energy and bandwidth limitations of electrical I/O. Yet, as optics move closer to compute, packaging, thermal management, and system-level robustness increasingly dominate performance and scalability. Here, we argue that co-packaged optics should not be viewed as a component-level optimization, but as an architectural commitment that reshapes the boundaries between photonics, electronics, and system design. We examine how heterogeneous integration strategies, chiplet-based optics, and emerging packaging platforms redefine scaling laws for AI systems, often introducing trade-offs that are underappreciated in device-centric analyses. Looking forward, we discuss why standardization, serviceability, and thermal-aware co-design will be decisive in determining whether co-packaged optics can transition from early deployment to widespread adoption in AI-scale datacenters.