Firm capacity on bankable terms?
Validate allocation, interconnection, rate structure, escalation, redundancy, and delivery schedule directly with the utility — before any capital commitment.
Happy Valley–Goose Bay · Labrador · 53.3°N
Move the computeto the power.
Labrador North Compute is developing hydro-powered infrastructure for large-scale, latency-tolerant AI training — starting with a disciplined 3 MW pilot on some of the lowest-cost renewable electricity in North America.
Initial pilot
Campus vision
~3.15¢per kWh · renewable
Hydroelectric
Hydroelectric power
Batch AI training
Canadian sovereign infrastructure
Built with Labrador
The thesis
The bottleneck for AI isn't chips. It's power.
Training large models is, in the end, an electricity business. The single largest operating cost is power — and increasingly the binding constraint is simply whether enough clean power exists in one place. Labrador has it, in quantity, at a price few regions can match. The move is to bring the compute to the power, not the other way around.
The project is in development. All rates, capacities, costs, schedules, and performance figures are planning estimates and require confirmation through utility, engineering, commercial, environmental, Indigenous, and financing diligence.
01
Power dominates the bill
Across the life of a cluster, electricity is the largest recurring cost of running AI compute. A structural price edge compounds every hour, on every rack.
02
Training can travel
Model training is latency-tolerant. Unlike consumer inference, it doesn't need to sit beside its users — data arrives in bulk and results leave in bulk.
03
So site it at the source
Put the load where clean power is abundant and cheap. Labrador turns a national power constraint into a durable, domestic advantage.
The power case
The power is already here. In enormous quantity.
Happy Valley–Goose Bay sits at the centre of one of the largest hydroelectric complexes in North America. Against that backdrop, even the long-term campus vision is a modest load.
PilotPhase 1 · our load
3 MW
Campus visionLong-term · our load
300 MW
Muskrat FallsInstalled · ~30 km away
824 MW
Gull IslandProposed · future
2,250 MW
Churchill FallsInstalled capacity
5,428 MW
The point isn't to consume it all. It's that a small, disciplined load can sit next to extraordinary clean-power headroom. Much of Churchill Falls is committed under long-term contract; the near-term path draws on Muskrat Falls proximity and available capacity, with Gull Island and the 2041 contract horizon as long-term headroom.
// ~3.15¢/kWh · Labrador Interconnected System
// 100% renewable hydroelectric
// figures are installed nameplate capacity
The advantage
A structural gap in the operating cost.
Power price is the lever that moves AI economics most. Adjust the pilot size to see an illustrative difference in annual energy spend.
Facility IT load3.0 MW
1 MW20 MW
Labrador rate3.15¢/kWh
Illustrative benchmark12.0¢/kWh
Illustrative comparison of energy cost only, at the rates shown. Excludes demand charges, transmission, facility overhead, escalation, taxes, curtailment, and commercial PPA terms. Actual pricing is subject to a utility agreement.
Illustrative annual energy advantage
$2.30M
26,280 MWh of energy per year
≈ 3.8× lower power cost
What we run
Built for compute that doesn't need a city.
There are two kinds of AI compute. One answers in milliseconds and must sit beside its users. The other trains for days or weeks and needs one thing in bulk: cheap, clean power. We build for the second.
- 01Large-scale model training
Long batch jobs where cost-per-token and throughput matter far more than proximity to the end user.
- 02Fine-tuning & research runs
Schedulable, latency-tolerant workloads that can be planned around power and capacity.
- 03Flexible, grid-friendly load
Batch work can pause and resume — making the facility a cooperative customer on the system rather than a rigid one.
The platform
Designed as a scalable compute campus.
A modular, single-storey core-and-shell strategy lets the facility start disciplined, prove the operating model, and expand in planned phases.
Layer 05
AI compute optimized for batch training
The pilot targets dense GPU infrastructure suited to high-value training and batch workloads. The commercial model can evolve from an anchor tenant or capacity contract toward owned-compute exposure as financing and operating capability mature.
- Modular deployment aligned with contracted demand
- High-bandwidth cluster fabric
- Remote operations supported by local critical-facilities staff
Disciplined scale
Prove the fundamentals, then expand.
Structured to retire the hardest risks — power, connectivity, and the first customer — before committing to the full campus.
Phase 0
Develop
Secure the foundation
Confirm the utility path, site control, fibre options, Indigenous engagement, permitting, commercial demand, and preliminary engineering.
Current priority
Phase 1
3 MW
Commission the pilot
Build a right-sized, expandable facility, validate operating performance and uptime, and deliver contracted AI workloads.
Pilot
Phase 2
9–10 MW
Expand the core
Add white space and electrical capacity within a pre-planned core-and-shell footprint as demand and power allocation grow.
Demand-led
Phase 3
Up to 300 MW
Build the regional platform
Pair large-scale clean compute with new transmission, dedicated fibre, and Canadian sovereign-AI requirements.
Long-term vision
Built with Labrador
Located in Labrador — and built with it.
A large flexible load is exactly the domestic anchor Labrador's clean power has been waiting for. The strongest structure keeps that value in the region: local ownership, local jobs, and durable public benefit.
01
A domestic anchor for clean power. With export routes in flux, a steady local customer turns idle renewable capacity into economic activity at home.
02
An Innu Nation equity partnership is being pursued, so the community holds a genuine stake in the project — not just a footprint on the land.
03
Local capability building in critical-facilities operations, skilled trades, construction, and services.
04
Government as facilitator, not financier — a bankable long-term power agreement and Crown-land access, rather than public equity risk.
Potential regional outcomes
Clean industrial loadCreates a domestic customer for Labrador energy.
Skilled employmentSupports operations, construction, trades, and service roles — quantified conservatively, methodology on request.
Digital infrastructureStrengthens the case for improved fibre and power capacity across the region.
Sovereign AI capacityAdvances Canadian-owned, Canadian-sited compute infrastructure.
Execution discipline
We lead with the hard questions.
The project is compelling only if the core assumptions survive commercial and technical diligence. These are the ones we resolve first.
Reliable, redundant data movement?
Confirm terrestrial fibre ownership, dark-fibre or wavelength options, route diversity, capacity, latency, and interim bulk-transfer logistics.
Who signs the first contract?
Convert broad AI demand into a creditworthy anchor agreement before overbuilding or purchasing rapidly depreciating hardware.
Can the region run 24/7?
Build a realistic staffing, relocation, training, spares, vendor-support, and remote-operations model for a northern site.
Project questions
What stakeholders ask.
Why Labrador instead of a major city?
The project targets workloads where power economics matter more than ultra-low metro latency. That value depends on securing firm hydro capacity and a reliable data-transfer strategy — the two things we resolve first.
Why start with only 3 MW?
A 3 MW pilot is large enough to demonstrate real operating performance and customer value, but small enough to manage capital, utility, staffing, construction, and market risk before scaling.
Will the project own the GPUs?
The path is deliberately flexible. An anchor-tenant or contracted-capacity model reduces early risk; owned compute can capture more of the low-cost-power advantage once the operating platform is proven.
Is the 300 MW campus committed?
No. It is a long-term scenario, not a construction commitment. Expansion would depend on power, transmission, fibre, customer contracts, capital, environmental review, community support, and successful pilot performance.
How does the community benefit?
Through a domestic customer for local clean power, skilled employment, an Innu Nation equity partnership that is actively being pursued, and stronger regional digital infrastructure.
Investor, utility & partner enquiries
Help build Canada's next clean-compute region.
Start a conversation about power, land, connectivity, Indigenous participation, engineering, customers, financing, or workforce development.