Own your GPU servers. Colocate them in Grant County, WA — where Columbia River hydro power from Grant County PUD makes electricity costs 40–60% lower than Seattle. Same 10–20ms latency. Real bandwidth included, not nickel-and-dimed by the gigabyte. Racks deliver up to 17.3kW on 208V 3-phase power, air-cooled. Owner-operated facility with real people on the phone.
GPU colocation in Washington state isn't hypothetical for us. We have a paying customer running dedicated GPU hardware right now.
Our first GPU colocation customer colocates a dedicated GPU server cluster at our Quincy facility on a 5kW dedicated power circuit, 24-month term. The deployment runs continuously on Columbia River hydro power with 50 Mbps of included bandwidth per kW — 250 Mbps (about 80 TB/month) at this customer's 5kW, with no per-GB fees.
This customer evaluated Seattle-area colocation options before choosing Quincy. The decision came down to power cost and flexibility: the same 5kW deployment in a Seattle facility would cost roughly 2–2.5× more per month. Grant County PUD hydro power changes the economics of running GPU hardware around the clock.
A single 8-GPU server draws 3–6kW continuously. At that power density, where you plug in matters more than almost any other infrastructure decision.
Grant County PUD delivers Columbia River hydroelectric power — consistently among the cheapest commercial electricity rates in the US. Microsoft, Amazon, and Google all built in Quincy for the same reason. You get the same power advantage at a fraction of hyperscaler scale.
10–20ms latency from Quincy to Seattle on Grant County PUD fiber. Sufficient for AI inference serving, model APIs, and remote management. Connections are 1 Gbps, and each kW includes 50 Mbps of sustained bandwidth (about 16 TB/month per kW) — never billed by the gigabyte.
Call the owner at 509-906-1250. Unusual power configurations, non-standard circuits, custom cooling requirements — we work these out directly, not through a ticket queue. If it's physically feasible, we'll figure it out with you.
Running LLM inference, image generation, or ML model APIs 24/7? Power cost at full utilization dominates your TCO. Colocate your inference hardware in Grant County and your per-request cost drops dramatically versus Seattle or cloud GPU instances. 12-month term, or month-to-month at $210/kW.
Training runs that go days or weeks hit full power draw the entire time. That's exactly when cheap power matters most. Colocate your training cluster — GPU servers, NVMe storage, and InfiniBand interconnects — in a facility that doesn't penalize you for running hot and heavy around the clock.
Visual effects, 3D rendering, and compute rendering workloads are power-hungry and latency-tolerant. Grant County is ideal: cheap power, fast fiber to Seattle, real bandwidth included for your render output. Custom rack configurations for high-density GPU arrays available.
University labs, research groups, and simulation workloads that run for days at full draw. Grant County PUD hydro power keeps long-running GPU jobs affordable. Bring your own cluster and we handle power, cooling, and bandwidth.
For GPU workloads specifically, the comparison isn't close. Here's what to expect:
| Factor | Columbia Colocation (Quincy, WA) | Typical Seattle-Area Colo |
|---|---|---|
| Power rate (5kW GPU) | $200/kW/mo all-in | $400–600+/kW/mo typical |
| Monthly cost (5kW example) | ~$1,000/mo | ~$2,000–$3,000+/mo |
| Bandwidth | 50Mbps/kW incl., no per-GB fees | $0.05–$0.10+/GB common |
| Latency to Seattle | 10–20ms | <5ms |
| 3-phase power for GPU | Available (208V 3-phase) | Available at most facilities |
| Custom circuit configs | Yes — call the owner directly | Yes, but through procurement |
| Minimum term | 12-month term; month-to-month +$10/kW | Often 12-month minimum |
| Who answers the phone | The owner | Sales / support team |
Seattle-area facility rates are estimates based on publicly available information. Actual rates vary by facility and configuration.
Customer-owned servers only — we do not rent servers or GPUs.
| Power per rack | Up to 17.3kW — 208V 3-phase, 60A circuit, 48A continuous |
|---|---|
| Single-phase | 30A single-phase PDUs available |
| PDUs | APC and Raritan |
| Cooling | Air-cooled, hot/cold aisle. Liquid cooling not offered today; being evaluated. |
| Bandwidth | 1 Gbps connection; 50 Mbps per kW included (about 16 TB/month per kW), no per-GB fees |
| Pricing | From $200/kW/mo all-in. $0 setup. |
| 1 × 8-GPU node | H100/H200-class, about 10kW each, plus networking and storage |
|---|---|
| 1 × 8-GPU node | B200-class, about 14kW — fits within a single rack |
| 2–3 × 4-GPU servers | About 5–6kW each |
| 4–8 × GPU workstations | RTX-class at roughly 1.5–3kW each |
No hidden line items. One rate covers power, rack space, and 50 Mbps of bandwidth per kW reserved — billed on actual usage, not per gigabyte.
GPU deployments vary widely. Tell us your GPU models, server count, estimated power draw, and timeline — we'll give you a real number fast.
Get a Custom GPU Quote →GPU model, server count, estimated power draw, cooling requirements, timeline — the more you tell us, the faster we can give you a real number. No sales scripts, no runaround. You're talking to the people who run the facility.
Our first GPU colocation customer went from inquiry to signed agreement in under a week. That's how we operate.