Openreach's optical upgrade: what 800G means for AI data center buyers

Openreach is expanding its optical network with 800G capacity, which could impact AI data center buyers. The deployment, announced on October 7, includes Ciena's Waveserver equipment using WaveLogic 6 technology to deliver 400G and 800G traffic over dedicated wavelengths. This creates a national 800G service, intended for data center interconnection, private cloud networks, and links between cable landing stations and data centers.

The upgrade introduces new considerations for infrastructure buyers. The key is understanding how the 800G service performs between specific sites used by the buyer. Line rate, customer-facing service rate, and application throughput are distinct aspects that should be evaluated separately. Simply having a higher line rate does not guarantee performance in the broader system.

When evaluating the 800G service, define the transfer that matters for the specific AI workload. Is it moving training data, replicating checkpoints, or serving requests from another site? Each task has different completion times and behaviors under interruptions. Ask the provider about the offered interface, service boundaries, and committed performance for the endpoints in your order.

A high-capacity port is beneficial only when part of a complete path. The order should specify the locations being connected and the facilities through which the service reaches the customer's equipment. Address fault investigation responsibilities, route diversity, and shared risks between services. Test the traffic you intend to carry, rather than relying solely on link-up success. Establish how long transfers take, what happens during connection failures, and how the application behaves during recoveries.

Remember that an optical service may remove one bottleneck but may introduce others in the data-loading or processing pipeline. Monitor failover plans, including who is responsible and how recovery is confirmed. Clear ownership of monitoring and escalation is crucial, especially when multiple suppliers are involved.

Ultimately, the practical value for AI deployments lies in the actual service performance at the endpoints, the path's resilience, and measured application behavior. Keep these factors in mind when making purchasing decisions, alongside the advertised rate and bandwidth specifications.