The sovereign stack for the underserved AI vertical.
A $255 billion AI inference market that hyperscalers can't economically serve. We acquire stranded gas, build modular generation, and host AI and Bitcoin workloads at energy cost approaching zero.
The Problem
Hyperscalers are optimized for the top of the market.
Massive scale, frontier models, five-nine SLAs, premium pricing. The structural gap is everything below that — cost-sensitive, deployment-sensitive, sovereignty-conscious workloads that need 'good enough and always on,' not 'best possible with scheduled maintenance.'
- Managed SLMs for customer-specific data — regulatory, legal, medical, manufacturing, retail
- Private hosted facilities that eliminate per-token cloud costs and data sovereignty issues
- Low-cost, environmentally friendly energy supporting energy-conscious hardware stacks (NPU-class chips, Apple Silicon, edge ASICs) at a third of GPU cluster power profiles
- Low-cost inference model hosting that makes effective agent deployment economically viable
The Approach
Build where the market is going at a fraction of owning where it is now.
Vertical Integration Moat
Acquire and build sovereign energy sources from stranded oil and gas assets for off-grid electricity production.
No Stranded Assets
Modular 1 MW AI centers at 30–40% below cost, with 2 MW Bitcoin underlay so every electron is monetized 24/7.
Category Creation
Define a new industry segment — Energy Native Computing — for cost-sensitive AI workload buyers.
Module Architecture
Four module types. One integrated stack.
| Module | Capacity | Function | Revenue Role |
|---|---|---|---|
| Modular Gen Sets | 2.5 MW each + 1 MW battery + 1 MW diesel backup | Captive on-site generation, 99.999% uptime target | Eliminates grid dependency; enables power arbitrage |
| AI Data Centers | 1 MW per module | SLM hosting & inference compute | Primary revenue driver, ~50% discount to hyperscaler pricing |
| Bitcoin Data Centers | 2 MW per module | Hash rate generation / load balancing | Baseline cash flow; absorbs excess generation |
| Fiber Connectivity | Scalable | Dedicated low-latency network backhaul | Enables private cloud & sovereign AI deployments |
Why Our Economics Work
Near-zero long-term cost. Asymmetric by design.
Near-Zero Long-Term Leasehold Cost
Near-Zero Long-Term Electricity Cost
No Legacy Data Centers / No Grid Dependency
Rapid Delivery, HyperFlexibility, Low Cost Provider
Computational Arbitrage
Timeline to Market
Built, not theorized.
- Q1–Q2 2026
Develop 2–3 stranded gas sites; deploy first modular units
- Q3–Q4 2026
Demonstrate Tier 1 'Economic Engine'
- Q3–Q4 2026
Demonstrate distributed training on live model
- Q4 2026 – Q1 2027
Sign first Inference / SLM customers
See the active site portfolio.
100 MW of properties under due diligence across the US and Canada. Start with Project Foothills.