The race to build ever-larger artificial intelligence supercomputers has hit a hard physical limit: copper cannot reliably carry data fast enough over the distances required. Now, six technology giants have moved decisively to solve that problem with a new open standard for optical interconnects within AI data centres.
The Optical Compute Interconnect (OCI) Multi-Source Agreement (MSA) group announced its formation, led by founding members AMD, Broadcom, Meta, Microsoft, NVIDIA and OpenAI. This consortium marks a departure from how industries typically standardise technology. Rather than waiting for consensus among dozens of players, these companies are using what's known as a Multi-Source Agreement structure to move with speed.
The urgency reflects a genuine technical bottleneck. Traditional electrical transmission using copper cables faces physical limitations and will struggle to support the massive data movement required by next-generation AI infrastructure, as a result, optical transmission technologies are gaining greater importance. At the ultra-high speeds modern AI systems demand, electrical signals over copper degrade rapidly with distance, effectively limiting copper interconnects to distances of less than one meter.
The OCI standard begins modestly before scaling aggressively. The OCI connectivity technology will define a common PHY based on NRZ signaling and wavelength-division multiplexing (WDM), starting at four wavelengths x 50 Gb/s (200 Gb/s per direction) and scaling all the way to 800 Gb/s per fiber. Over time, the roadmap is expected to expand both wavelength counts and signaling rates, targeting 3.2 Tb/s per fiber and beyond as the ecosystem evolves.
What distinguishes this effort is its architecture-specific focus. OCI targets a very specific architectural layer of AI systems: short-reach links that connect accelerators and switches within a scale-up domain. This precision matters. MSAs enable select companies to align on electrical/optical interfaces and build interoperable products quickly, without the lengthy consensus processes typical of classic organisations. The OCI MSA will enable AMD, Broadcom, and Nvidia to build interoperable short-reach interconnections for Microsoft, Meta, and OpenAI.
Meta's VP of Hardware Systems noted that the appetite for technology to address the power and cost constraints impacting AI cluster design is real and imminent, and encouraged adoption of the OCI protocol to decouple the need for larger scale-up domains from the limitations of electrical backplanes in high performance AI clusters. The power efficiency gains are substantial: optical solutions can match copper's cost per connection while delivering superior bandwidth density and dramatically greater reach.
The standard also supports three different implementation approaches. The technology will support pluggable optical modules, on-board optics, and co-packaged optics (CPO) integrated directly with compute silicon. This flexibility lets designers choose the approach that best fits their systems as the technology matures.
Industry observers note a strategic dimension. NVIDIA is involved in this consortium with other vendors while notably absent is Intel. The formation signals that NVIDIA sees interoperability as strategically important despite its dominance in AI accelerators, while Intel's absence underscores the chip maker's diminished role in this particular technology ecosystem.
The broader context reveals why this timing matters. Hyperscalers are building data centres of staggering scale: ChatGPT3 was trained on a cluster with approximately 1,000 accelerators that required 2,000 optical interconnects, while ChatGPT4 was trained on a 25,000-accelerator cluster with about 75,000 optical interconnects. Future models will dwarf even these numbers, making the interconnect challenge not peripheral but foundational.
For Australia's technology sector and the broader Asia-Pacific region, the OCI standard has indirect but real implications. Investment in optical interconnect components, photonics expertise, and data centre infrastructure increasingly depends on understanding these underlying architectural shifts. The move toward optical scale-up interconnects represents a generational change in how AI computing clusters are designed, with ripple effects across component suppliers, manufacturing partners, and data centre operators across the region.