Operations Overview

Operations

March 29, 2026

Integrated view of core operational processes, cost efficiency (~92% gross margin, ~85% EBITDA), structural advantages, and risk map for 1for.ai BTM hydro GPU infrastructure

Icon
$45M
Total CAPEX
$1.0M
Annual OPEX
~92%
Gross Margin (EBITDA ~85%)
288
GPUs Deployed
<1.10
PUE Target
~8%
Breakeven Util.
01Infrastructure Stack
Power — BTM Hydroelectric
1 MW behind-the-meter · $0 electricity · No BESS
$1.06M
Compute — NVIDIA Rubin Ultra
4× GR200 NVL72 · 288 GPU + 144 Grace CPU · 10.4 ExaFLOPS FP8
$26.5M
Networking — InfiniBand NDR800
Fat-tree topology · inter-GPU fabric
$4.8M
Storage — All-NVMe
~1.5 PB usable capacity
$2.1M
Cooling — DLC + River Water
Mountain river 4–12°C · PUE < 1.10 · No mechanical chillers
$1.6M
Civil — Prefab Containers
Modular, seismic-rated · ~18mo build
$1.6M
Connectivity — Fiber + Starlink
Dual-path redundancy · global reach from Day 1
$0.5M
CAPEX Distribution — $45M Total
58.9%
10.7%
Compute $26.5M
Network $4.8M
Storage $2.1M
Power $1.06M + Cooling $1.6M
Civil $1.6M
SW + Connect $2.0M
Contingency ~$5.3M
02Cost & Time Efficiency
$1.0MANNUAL OPEX
NVIDIA Maintenance40%
Insurance + Connectivity18.5%
Software Licenses15%
Contingency / Other14.5%
Remote NOC12%
Electricity$0
Time Efficiency: 1for.ai vs Industry
Site Development
1for.ai
~18 mo
Industry
24–36 mo
Client Onboarding
1for.ai
2–4 wk
Industry
4–12 wk
CAPEX Payback
1for.ai
~3.5 yr
Industry
4–7 yr
Gross Margin (EBITDA ~85%)
1for.ai
~92%
Industry
20–40%
03Structural Advantages
01
Energy Cost — Structural Zero
JV co-ownership of hydro generation. Internal transfer = $0 on P&L. Industry benchmark: $0.08–0.15/kWh. Accounts for 50–70pp gross margin gap vs industry.
$0/kWh vs $0.08–0.15 industry
02
Zero Carbon — Source-Verified
100% hydroelectric, behind-the-meter. No RECs, no offsets, no grid-mix averaging. Structural compliance with EU AI Act, SEC disclosure, ESG mandates.
0g CO2/kWh · No audit friction
03
Data Sovereignty
Non-EU, non-US, non-Five Eyes jurisdiction. EU Association Agreement. No hyperscaler vendor lock-in. Neutral compute for sovereign AI programs.
Georgia · Singapore wrapper ~2030
Moat Defensibility Assessment
BTM Hydro JV ($0 electricity)
Very Low
Mountain River DLC (PUE <1.10)
Very Low
Neutral Jurisdiction (Georgia)
Moderate
Latest-Gen NVIDIA Hardware
Low
Remote NOC / Lean Ops
High
Scale = replicability risk · More dots = harder to replicate · Green = structural moat
04Operational Issues & Risk Map
Critical: 1
High: 4
Moderate: 4
Low: 1
NVIDIA Single-Source GPU Dependency
No alternative at performance tier. BIS export controls may restrict supply. Mitigated by allocation agreements + Singapore jurisdictional optionality.
Critical
BIS Export Control Escalation
Georgia status under US export controls could change. Existential risk — no structural mitigation. Singapore entity provides procurement optionality.
High
Client Concentration
1–2 clients per facility. Multi-track revenue model distributes risk. Anchor LOI required before FID.
High
Pre-Revenue Credibility Gap
No operating history at launch. Site visit program + beta PoC (Q3 2028) + founder direct engagement.
High
Founder Dependency
Single founder across BD, strategy, ops. COO/CTO hire planned Year 2. Advisory board pre-Series A.
High
GPU Generation Transition
7yr depreciation + modular rack refresh. Inference less generation-sensitive than training workloads.
Moderate
Remote NOC Reliability
Dual connectivity + UPS + per-rack DLC isolation. Local emergency response team contracted.
Moderate
Georgia Jurisdiction Perception
Singapore holding wrapper ~2030. EU Association Agreement. Neutral jurisdiction as feature.
Moderate
USD CAPEX Exposure
$45M in USD. Revenue in USD/EUR = natural hedge. GEL OPEX base structurally favorable.
Moderate
Hydro Variability
Eliminated by design — GPU infra sized to HPP minimum guaranteed output, not peak or average.
Low