Imagine a World Without Bandwidth Constraints
Announcing a cross-site networking solution that eliminates bandwidth as a constraint on deployment, scaling, and infrastructure location.
Commercially available Q3 2026
Unmatched Performance
Four Structural Advantages
Resilience Without the Tradeoff
Both sites carry live traffic at the same time. If one site fails, the other keeps running. No failover delay, no recovery window, no idle capacity sitting unused.
Compliance Built Into the Architecture
Sensitive data never leaves the site where it originates. Only model gradients cross the inter-site link. Compliance is structural, not a policy layered on after.
Operational Flexibility Across Workload Types
Run a single job across the full cluster, or split into independent workloads that burst across sites as demand shifts.
A Scaling Path That Does Not Require Starting Over
Additional sites join the same logical cluster as demand grows, using the same networking approach already in place. Nothing gets rebuilt.
Use Cases
Repurposing existing assets: Metro sites that are stranded or underutilized for AI workloads become viable when aggregated through a distributed cluster architecture. The investment is in interconnect and compute, not in new land, power, or construction.
Hybrid partnerships: Telcos have relationships across every enterprise vertical. DGC enables them to connect enterprise infrastructure to their own metro fabric, creating managed AI capacity that enterprises cannot build themselves. Shared training, data locality compliance,and enterprise-to-provider distributed training all become viable.
Owning the last mile for compute: The AI transport layer in DGC is built on Telco-native infrastructure. No translation layer between the AI platform and the network it runs on. That is a differentiation Telcos can sustain, because it is built on infrastructure they already operate.
Institutions running AI for trading, risk, and compliance face simultaneous pressure from BCBS 239, DORA, and MAS TRM requirements for operational resilience and geographic redundancy.DGC delivers resilient AI compute purpose-built for financial workloads: train proprietary models without third-party exposure, run real-time inference with automatic site failover, and meetstructural redundancy requirements that contractual SLAs cannot replace.
Research institutions need supercomputer-class compute for genomics, drug discovery, and clinical AI, while facing strict data movement restrictions across patient cohorts and sites. DGC allows each site to hold distinct patient cohorts while contributing to a single federated training run. Patient data stays at its originating site. Only model gradients cross the link. This is privacy-by-architecture, not privacy-by-policy.
.svg)
