RETN has launched a new backbone route between Tallinn, Estonia and Cēsis, Latvia, strengthening connectivity between Northern and Central Europe. The route was validated under real-world conditions when a fiber break disrupted the primary backbone path during pre-service testing in late 2025.

During the incident, RETN engineers rerouted more than 40 DWDM backbone channels across multiple European segments onto the newly built Tallinn – Cēsis line within 60 minutes.

In Layer 1 environments, where restoration typically depends on physical fiber repair, outages can last days – particularly on subsea routes. Despite the scale of the rerouting, key performance indicators, including latency and jitter, remained stable throughout.

The Tallinn – Cēsis backbone forms part of RETN’ s strategic expansion programme, designed to enhance route and supplier diversity across the region. The new route delivers improved resilience by adding an additional terrestrial path linking Northern Europe, including Finland and the Nordics, with Central Europe. It also reduces reliance on single fiber suppliers.

The project includes a new Core Point of Presence at Greenergy Data Centre, Tallinn, enabling RETN to provide its full portfolio of network services locally.

With capacity of up to 40Tbps and new DWDM spectrum available, the route supports future growth and rapid traffic rerouting during disruptions.

“ Backbone networks must be designed on the assumption that outages will occur,” said Tony O’ Sullivan, CEO, RETN.“ This route strengthens both route and supplier diversity, ensuring traffic can be shifted quickly without compromising performance.”

ESTONIA

Gcore has integrated NVIDIA Dynamo into its AI inference portfolio to deliver high-performance, cost-efficient AI inference as a fully managed service

The integration delivers significant GPU efficiency gains, enabling up to six times higher throughput and two times lower latency through a fully managed, one-click deployment.

NVIDIA Dynamo is an open-source inference framework designed to accelerate and optimize large-scale generative AI and inference workloads. It addresses key challenges organizations face when running inference at scale, including GPU underutilization, static resource allocation, memory bottlenecks and inefficient data transfer.

Gcore delivers Dynamo as a fully managed solution, preoptimized for leading inference models. Customers can activate Dynamo with a single click in the Gcore Customer Portal without managing routing, KV cache logic or GPU scheduling. The integration is supported across private cloud, hybrid and on-prem environments through Gcore Everywhere AI and Everywhere Inference.

By disaggregating prefill and decode, applying KV cacheaware routing and leveraging NIXL for efficient inter-node communication, Dynamo increases GPU utilization and reduces wasted cycles. More requests are processed on the same infrastructure, lowering cost per token and improving ROI at scale.

LUXEMBOURG www. intelligentcio. com

INTELLIGENT CIO EUROPE

33