THE PLATFORM

One service portfolio, from the carrier core to the access edge

Buy dedicated metal, managed platforms, networking, and delivery through one portfolio. Train in Core Harbor, serve a region through Middle Harbor, and run near-real-time inference at the access edge, all on a network HarborOne owns.

ARCHITECTURE

Three tiers place compute near demand

Core Harbor

1,024- or 1,152-GPU SuperPODs*

WHERE IT SITS

Carrier hotels and peering exchanges

WHAT IT IS FOR

Large distributed training, fine-tuning, batch inference, and model hosting

Middle Harbor

Regional GPU capacity

WHERE IT SITS

Tier 2 and Tier 3 metros

WHAT IT IS FOR

Regional inference and burst capacity, closer to the people and devices that use it

Edge Harbor

Compact GPUs

WHERE IT SITS

ISP PoPs, central offices, headends, small data centers, and qualified buildings

WHAT IT IS FOR

Near-real-time inference, with an ≈5 ms network-path target at selected sites*

SuperPODs are available in 1,024- and 1,152-GPU configurations. Site power profiles depend on the selected systems and facility design.

≈5 ms is a network-path target for selected sites, not total inference latency. End-to-end latency depends on the access network, routing, model size, serving hardware, queueing, and response length.

Platform services across the stack

GPU infrastructure

B300 SuperPODs, dedicated GPUs, GPU instances, bare metal, and GPU partitioning where supported, on Reserve or On-Demand terms

Cluster access

Kubernetes workloads, Slurm-compatible batch jobs, virtual machines, and containers

Data services

PostgreSQL, S3-compatible object storage, vector databases, cache, and dashboards

Security options

Virtual networks, identity and policy controls, DDoS mitigation, WAF, IPS, and DLP by deployment

Exact service availability varies by Harbor tier, region, hardware, and customer configuration.

Where does your workload belong?

WorkloadHarborOne locationWhy it fits
Large model trainingCore HarborB300 SuperPOD with InfiniBand and high-performance storage
Fine-tuning and batchCore or Middle HarborPut compute near the data, the capacity, and the operating requirements
Regional inferenceMiddle HarborServe metro users from a nearby GPU pool
Near-real-time inferenceEdge HarborQualified edge GPU sites with an ≈5 ms network-path target*

≈5 ms is a network-path target for selected sites, not total inference latency. End-to-end latency depends on the access network, routing, model size, serving hardware, queueing, and response length.

Tell us what you run and where your users are

We map your workload to a tier, design the path, and come back with a plan and a price.