Arelion introduced upgraded 400G Ethernet Virtual Private Line (EVPL) services as part of its AI Direct portfolio, targeting enterprises, neocloud providers, hyperscalers, and wholesale operators building distributed AI infrastructure. The new service enables high-capacity private interconnection between AI data centers, GPU clusters, cloud regions, and hybrid cloud environments across Arelion’s global IP backbone.
The offering adds dynamic prioritized routing capabilities designed to automatically reroute traffic around network disruptions in real time using alternate backbone paths. Arelion said the service combines EVPL, Cloud Connect, and IP Transit through a Multiple Services on One Port (MSOP) architecture, allowing customers to consolidate multiple connectivity services over a single physical interface. The platform also supports jumbo frames with MTU 9400 to improve throughput efficiency for large-scale AI data transfers, including model synchronization, training datasets, and inference workloads.
Arelion positioned the upgraded EVPL platform as a connectivity layer for emerging agentic AI deployments and distributed AI compute architectures. The service includes network-protected and prioritized Class of Service (CoS) options for secure, low-latency transport across geographically dispersed environments. The company said the platform can support sectors requiring real-time performance, including financial services, streaming, gaming, and cloud AI infrastructure. Arelion currently serves more than 2,000 customers across nearly 130 countries through its global backbone and cloud connectivity ecosystem.
- 400G EVPL services added to Arelion’s AI Direct portfolio
- Targets hyperscalers, neoclouds, enterprises, and wholesale providers
- Dynamic prioritized routing automatically reroutes traffic during network events
- MSOP architecture supports EVPL, Cloud Connect, and IP Transit on one port
- MTU 9400 jumbo frame support improves large-scale data transfer efficiency
- Designed for AI training, inference, GPU cluster interconnect, and hybrid cloud connectivity
- Supports secure transport of proprietary AI datasets and model updates
- Operates across Arelion’s global backbone spanning nearly 130 countries
“As demand grows for scalable, highly resilient connectivity to support AI training and inference across distributed environments, customers need faster, more flexible ways to interconnect their infrastructure,” said Johan Ottosson. “By combining dynamic routing, secure connectivity and multi-service delivery over a single port, we help customers efficiently connect cloud, data center and neocloud deployments as AI workloads continue to evolve.”
🌐 Analysis: The launch reflects a broader shift toward dedicated AI interconnect services as GPU infrastructure becomes geographically distributed across hyperscale campuses, colocation facilities, and regional inference clusters. Carriers and backbone providers increasingly position Ethernet-based private networking services as alternatives to traditional wavelength or MPLS offerings for AI data mobility, especially where flexibility and cloud adjacency are priorities.
