Tech giants form optical interconnect alliance to solve AI infrastructure's data bottleneck

Major tech players unite to tackle AI infrastructure challenges

Nvidia, AMD, Broadcom, Meta, Microsoft, and OpenAI have launched the Optical Compute Interconnect Multi-Source Agreement (OCI MSA), a consortium designed to standardize optical interconnect technology for AI infrastructure 1

1

. The alliance represents a pivotal shift toward a hyperscaler-driven open ecosystem, enabling the development of a multi-vendor supply chain for optical scale-up interconnects 2

2

. By aligning on an open optical interconnect specification, these industry leaders are addressing the AI data bottleneck that threatens to limit the growth of next-generation AI workloads.

Breaking through copper's physical limitations

Traditional copper-based solutions are reaching their physical limits as AI models advance toward super intelligence. Copper is inherently a lossy, resistant medium that requires enormous amounts of power to send data over distances at high speeds 1

1

. Pushing electrical signals to high speeds results in signal degradation and unsustainable levels of power consumption. An optical physical layer (PHY) can overcome this electrical resistance challenge, allowing for higher-speed data transmission while stabilizing power usage 1

1

. The OCI MSA aims to develop a PHY capable of delivering up to 3.2Tbps and beyond, dramatically expanding bandwidth density and system scalability 2

2

.

Building a protocol-agnostic foundation for AI data centers

The OCI MSA's specification is architected to be power, latency, and cost optimized, combining non-return to zero (NRZ) modulation and wavelength division multiplexing (WDM) optical technology 2

2

. This shifts the connectivity paradigm from a module-centric to a silicon-centric model, enabling tighter integration of silicon photonics with compute and networking silicon. Vivek Raghunathan, CEO and co-founder at Xscape Photonics, noted that "optical cables and the silicon photonics technology already exist when it comes to connecting different switches as part of a pluggable transfer ecosystem" 1

1

. The new standard enables these physical chips to effectively travel across the same lanes, allowing AI data centers to theoretically run NVLink for Nvidia chips and UALink for AMD hardware while using the same underlying optical infrastructure at higher data transfer speeds than copper can realistically allow 1

1

.

Enabling plug-and-play multi-vendor deployments

By establishing an interoperable optical interface protocol, the OCI MSA enables a "plug-and-play" ecosystem that allows hyperscalers to disaggregate any top-tier processor unit (XPU) engines and top-tier scale-up switches through a common optical physical layer 2

2

. The consortium has introduced a scalable roadmap with high-density interfaces, including the 200Gbps OCI GEN1 and the 800Gbps GEN2 BiDi technology, supporting various form factors including pluggable, on-board, and co-packaged optics 2

2

. This standardization dramatically reduces integration risk, shortens development cycles, and provides the full AI rack supply chain with a clear, de-risked path for multi-generational deployments. Without an effective rulebook, each vendor design would become proprietary and siloed, creating barriers for large-scale AI workloads that run multiple ecosystems 1

1

.

Strategic implications for AGI and future AI development

Industry leaders emphasized that the OCI MSA is a critical response to the soaring power and bandwidth demands of next-generation AI interconnects. Broadcom and OpenAI underscored the technical and strategic importance of this transition for the path toward Artificial General Intelligence (AGI), with OpenAI stressing that reaching AGI requires the massive network bandwidth and extended reach that only a standardized optical scale-up solution can provide 2

2

. Meta and Microsoft noted that shifting from traditional electrical backplanes to advanced optical protocols is foundational for decoupling AI cluster growth from current physical limitations, ensuring high-performance compute domains remain scalable and cost-effective 2

2

. The specification is available at the consortium's website, providing transparency for the broader industry. While optical cables offer significant advantages in data bottleneck resolution, they also present challenges including failure rates, increased heat output, and higher overall costs 1

1

. The OCI MSA's standardized approach aims to mature this nascent technology while driving down costs through an open ecosystem that ensures interoperability across different packaging platforms, effectively de-risking optical supply chains in the process 1

1

.