Vancouver, BC — 2026

Sovereignty
Is Infrastructure.

5
Exits
9.5x
MOIC
$140M+
Raised for energy
Fund I is what happens when that track record gets its own vehicle.
Scroll
02 / Conviction

Capital is timeless.
Conviction is not.

The most important infrastructure being built right now doesn't have a weapons contract. It has a power purchase agreement. AI needs power. Power is sovereignty. Everything else follows.

"You either control your power
or someone else controls you."
Nicholas Roche
GP, Cinderwood Fund I
Nicholas Roche
Vancouver, BC
03 / The GP

Nicholas Roche

Solo GP — Cinderwood Fund I
Selected exits
AdvisorStream → Broadridge
FoxPass → Splashtop
DataSpark → ThreeColts

Growth operator and consultant for five years — scaling companies across sectors, then running their capital raises. 5 private company exits at 9.5x MOIC between 2019 and 2024.

Energy infrastructure advisory: On.Energy ($125M raised), BioEsol ($15M closed, $50M Series A active). 8,600-operator dealflow network built through those mandates.

Nicholas Roche
9.5x
Blended MOIC — 5 exits, 2019–2024
$140M+
Raised for energy companies (advisory)
8,600
Operator network
Vancouver, BC — Solo GP — US & Canada
04 / Mission

Five categories.
One thesis.

01 Grid Resilience
02 Hardware Lock-In
03 Dual-Use Infrastructure
04 AI Physical Bottlenecks
05 Critical Minerals Technology
Index
Fund I
Focus
Seed · Deep Tech · US & Canada
05 / The Problem

The grid wasn't
built for this.

AI → Energy Crisis
Power → National Security
Grid Failure → Sovereignty Failure
06 / Thesis

The bet is
sovereign power.

$5M · Seed · US & Canada · Sovereign Energy Infrastructure
"AI needs power.
Power is sovereignty."
$600B
Hyperscaler capex projected, 2026
165%
Data center power demand surge by 2030
07 / What We Back

What we
back.

01
Energy Infra + Sovereignty
Grid resilience, BESS, BMS, modular data centers, and power delivery with a sovereignty angle.
02
Hardware Lock-In
Churn becomes a physical problem. Nobody rips out an embedded BMS to evaluate a competitor.
03
Dual-Use Infrastructure
Same platform. Different customer. Same moat.
04
AI Physical Bottlenecks
Interconnects, power delivery, thermal — the physical limits on how fast AI can scale.
05
Critical Minerals Technology
AI-powered extraction for copper, lithium, cobalt, nickel. Same hardware flywheel — one layer upstream.
08 / Anti-Portfolio

What we
don't back.

Procurement-Dependent Revenue
Procurement cycles are too slow and exit buyers too narrow.
Pure Climate / ESG Without Sovereignty
ESG marketing cycles drive churn. No physical switching cost means no moat.
Software-Only Without Hardware Moat
Feature parity is a roadmap problem. We back deployment problems.
Full CapEx Ownership
We back the software layer. SPV-structured CapEx with infra fund backing is a feature, not a flag.
Thesis in the wild
Crusoe
Armada
Nscale
Varda
Radiant
These companies proved the thesis before we named it.
09 / Active Pipeline

Deals We're
Already Seeing.

Deal A — Pre-Seed
BESS / Grid Resilience
Battery management software embedded at industrial and municipal facilities. Hardware-first go-to-market, software contract renews on 3-year cycles. First commercial pilot signed.
Thesis fit
Hardware lock-in · Sovereign infra · Canadian domicile
Deal B — Seed
Modular Data Infrastructure
Edge compute deployed at stranded and underutilized power assets. Bypasses interconnection queue. Two hyperscaler LOIs in hand. Margins above 60% at current deployment scale.
Thesis fit
Dual-use · AI bottleneck · Hyperscaler exit path
Deal C — Pre-Seed
Critical Minerals Tech
AI-powered yield optimization for lithium and copper extraction. Sensor hardware deployed at mine site, proprietary geological dataset compounds with every tonne processed. BC-based, federally co-investable.
Thesis fit
Critical minerals · Hardware flywheel · Canada advantage
Companies anonymized. Illustrative examples only — Cinderwood may or may not choose to invest.
10 / The Model

Hardware Deploys.
Software Compounds.

Hardware companies that become software companies after deployment. The BMS ships once. The software contract renews forever. The switching cost lives in both.

01 →
01 — Deploy Once
Hardware creates the beachhead.
02 →
02 — Renew Forever
Software contract follows automatically.
03 →
03 — Switch Never
Replacement requires a new qualification cycle. Nobody mid-contract starts that process over.
How we underwrite it
Stickiness
The stickiness SaaS promised and rarely delivered.
Unit Economics
Premium multiples. Switching costs are physical, not contractual.
The Renewal
Once deployed, the software contract isn't a sales motion. It's a renewal.
11 / Spotlights

Four archetypes.
One moat.

AI Infra
The "Ayar Labs" Model
Solves a hard AI bottleneck — interconnects, inference silicon, or thermal. Hardware embedded at chip or system level makes switching a qualification-cycle problem. e.g. Ayar Labs, Untether AI
Exit → Nvidia, Broadcom, or hyperscaler vertical integration.
BESS / BMS
Embedded Sovereign BESS
A BESS or BMS deployment embedded in a critical facility or hyperscaler campus. The data layer and optimization contract that follows renews forever. e.g. On.Energy, Voltus
Exit → Google, Microsoft, Amazon, Tesla Energy, Siemens, ABB.
Edge Compute / Modular Infra
Modular Data Centers at Point of Power
Deployed directly at the power source, bypassing utility interconnection delays entirely. Armada has proven the model at scale — ruggedized modular data centers across 80+ countries, energy acquisition as the core moat, 70%+ margins despite hardware deployment.
Exit → Hyperscaler needing distributed compute, or Crusoe, CoreWeave consolidation.
Critical Minerals Tech
AI-Powered Extraction
Sensors and software at copper, lithium, cobalt, or nickel production sites. Proprietary operational data compounds on the installed hardware base. e.g. KoBold Metals, Plotlogic
Exit → Tesla, Panasonic, western CATL-alternatives, major mining consolidators.
12 / Where We Operate

Closest to the Deals.
Globally-Minded.

No US fund sees Canadian critical minerals deals before they're priced. No Canadian fund has the Pacific Northwest energy and compute network we do. Vancouver is one hour from Seattle, two from SF. We're in those rooms same-day.

We work with founders based in Paris deploying gigawatts in Mexico. The thesis — sovereign energy infrastructure — is more relevant in Tokyo, Berlin, and Sydney every quarter. We invest in US and Canadian companies. We think without borders.

We invest only in founders we know, like, and trust. That standard doesn't change based on where they're building.

"The best deals don't announce themselves.
You have to already be in the room."
<1hr
Vancouver → Seattle
~2hrs
Vancouver → San Francisco
US + CA
Primary investment geography
Global
Thesis relevance — Europe, Japan, Australia
13 / Portfolio Construction

Concentrated.
High-conviction.

Stage Initial Check Target Own. Focus
Pre-Seed
Technical Risk
 $500k – $750k 5 – 10% Core IP development, first hardware deployment.
Seed
Product Risk
 $500k – $750k 5 – 10% Pilot deployment, first commercial validation.
14 / Fund I — 2026

Build the
Backbone

Capital for the infrastructure that doesn't get a second chance to work.

Request Fund I Memo
Accepting commitments from accredited investors. Minimum: $100K.
Headquarters
Vancouver, BC
Web
cinderwood.vc
Email
Slide
01 — 14