Unit Economics & Cost to Serve
The whole model rests on one fact: the expensive thing is built once, and then licensed many times at almost no extra cost. This section traces every cost number to where it comes from — the four things we build — shows how each stream's cost is calculated, and explains why the margin widens as the business grows. Every figure here is a management estimate from the cost-to-serve model, built bottom-up from the bill of materials; there is no signed revenue yet.
The cost is built from four things
Every cost in this business traces to four cost centers. One is built once and shared; the other three are deployed in the field, in amounts that depend on what each sale has to prove.
- ROOT, the shared backbone, built once and run at a flat annual cost.
- Sensor nodes, drones, and environmental-DNA sampling — the marginal kit deployed per job.
Keeping these separate is what prevents double-counting: ROOT covers the fixed calibration layer on our own concession; the node, drone and eDNA cost centers count only the extra units deployed on a client's land.
ROOT, the shared backbone
ROOT is the one large cost, and it is fixed. It is built bottom-up, not estimated top-down.
- Build cost (CapEx): $19.3M. The pieces: two flux super-towers, one on terra firme and one on floodplain ($6.1M); the data-fusion platform ($3.0M); the AI models and digital twin ($2.3M); the bio-reference and environmental-DNA labs and libraries ($3.8M); external data and recognition libraries ($0.8M); the research campus near Iquitos with the reference node mesh and calibration plots on our own concession ($3.0M); and contingency ($0.4M). A deliberately narrow build — one tower and a small reference library — would be about $13.5 to $15M; the $19.3M is the fuller version.
- Run cost (OpEx): $3.15M a year, flat. Tower upkeep, the platform run-team and cloud, model retraining, library upkeep, satellite imagery, and the upkeep of the campus and concession. It is financed mostly without dilution: $13.5M grants, $3.0M development-bank debt, $2.8M equity.
- Amortisation is gated on the accounting treatment. Under the base case the grant nets against the asset, so the depreciation reaching the profit line is near zero; under the no-grant case the $19.3M depreciates straight-line at roughly $2.4M a year at peak. The treatment is pending the IFRS auditor.
The single most important property: ROOT is built once on our own land and does not grow when we sell more.
The marginal kit: nodes, drones, eDNA
- Sensor nodes, built in-house in four tiers. ECHO, a dense mini-node for sound, microclimate and soil, $90. HYDRO, for water level in flooded forest, $175. NEXUS-Hub, an edge-AI concentrator with cameras, a CO₂ sensor and a tree-growth band, $600. NEXUS-Core, the reference station with a full soil profile, a methane sensor and the highest-grade AI, $1,300. Running cost is attrition (10 to 15% a year) plus maintenance and connectivity, $25 to $142 per node per year, and nodes last about four years. We build them ourselves, which is the gap between our at-scale cost and the retail price of the same parts.
- Drones. A Skyeton Raybird fixed-wing aircraft carrying survey-grade LiDAR: about $3.9M to acquire, and $600k a year to fly. One aircraft can scan about 35 million hectares a year. Flown at full utilisation, that is about $0.04 per scanned hectare.
- Environmental DNA. $200 per sample, covering field collection, sequencing in our own laboratory, tamper-evident chain-of-custody, and bioinformatics. The eDNA laboratory and the reference library are built once, inside ROOT, so only the per-sample cost lands here.
Density: why each stream costs what it costs
The cost to serve a stream is its density — how much kit per 100,000 hectares — multiplied by the per-unit costs above. Each stream has its own density, set by what it has to prove.
- IS1, jurisdictional verification, is deliberately lean: 8 ECHO and 6 HYDRO nodes per 100,000 hectares, a drone over 20% of the area twice a year, and no eDNA. Degradation is a topology signal that needs sparse, well-placed reference points, not a full grid — which is why it costs only about $0.024 a hectare to serve.
- IS3 carbon is a fuller grid: 40 ECHO, 12 Hub, 4 Core and 12 HYDRO nodes, with a drone over 30% twice a year. IS3 biodiversity adds an acoustic and camera layer and 240 eDNA samples per 100,000 hectares. More signals mean more density and a higher cost — $0.10 a hectare for carbon, $0.65 for biodiversity.
- IS5, the national account, is the densest of all: 40 ECHO, 20 Hub, 8 Core, a drone over 35% twice a year, and 150 eDNA samples, because an official ecosystem account needs deep condition, not a floor signal.
- IS2, the data license, deploys nothing new. It reuses what ROOT already produces, so its cost is essentially its share of ROOT alone — which is why its margin is the highest.
- IS4 builds on the IS3 deployment with a single light drone pass, so it carries little new field cost of its own.
The drone is cheap because we cluster jurisdictions
A drone's cost per hectare is set by how full we fly it, not by the drone itself. A single jurisdiction loads one aircraft only 20 to 40%, which would push the cost to $0.07 to $0.13 a hectare; and flying a rented drone instead of our own would cost $1.50 per scanned hectare, roughly 37 times more, which would destroy the IS1 economics.
So we do not size IS1 on a single jurisdiction. We group the 44 Amazon jurisdictions into 14 full-drone packages across about 590 million hectares — 17 aircraft at about 72% average utilisation — which brings the drone cost down to about $0.04 a hectare. Loreto alone loads an aircraft to 73%. Clustering is what makes the $0.20 per hectare jurisdictional license profitable, and sizing the cost on clusters rather than one jurisdiction is what keeps it from being overstated.
How ROOT is shared across the streams
ROOT is a single shared cost, allocated across the five streams by a fixed share that sums to the $3.15M a year:
| Stream | Share of ROOT | ROOT cost per unit (2035) |
|---|---|---|
| IS1 | 40% | $0.004 per licensed hectare |
| IS2 | 30% | $9,450 per client |
| IS3 | 8% | $0.051 per hectare |
| IS4 | 12% | $16,435 per bond |
| IS5 | 10% | $45,000 per jurisdiction |
Each share is a fixed dollar amount divided by a growing volume. As coverage and clients ramp, the ROOT cost carried by each unit falls toward zero: by 2035 ROOT adds only about four tenths of a cent to a licensed hectare. This is the engine of the whole model.
The moat is the gap between two hectare counts
There are two hectare numbers in this business, and they must never be conflated.
- The cost base is the concession: about one million hectares of our own land, where ROOT is calibrated. It drives the $19.3M build and the $3.15M a year, and it stops growing once the calibration footprint is built out around 2029.
- The revenue base is licensed coverage: the third-party jurisdictions that buy the MRV licence, ramping from 35 million to about 295 million hectares, at almost zero marginal cost per added hectare.
Build the calibration once on a million hectares, then license the measurement across hundreds of millions at almost no marginal cost. Fixed cost divided by growing coverage is operating leverage, and it is why the margin rises rather than holding flat.
What it costs to serve each stream
| Stream | Unit | Cost to serve | Price | Gross contribution |
|---|---|---|---|---|
| IS1 · Jurisdictional Verification | per hectare / year | $0.024 | $0.20 | 88% |
| IS2 · Enterprise Data License | per client / year | $10,200 (base) · $25,200 (+eDNA) | $300,000 · $550,000 | ~96% |
| IS3 · Monitoring-as-a-Service | per hectare | $0.10 (carbon) · $0.65 (bio) | $0.50 · $1.50 | 57–79% |
| IS4 · Outcome Verification | per $100M bond / year | $106,435 | $150,000 | 29% |
| IS5 · Natural-Capital Accounting | per jurisdiction | $370,000 | $525,000 | 30% |
Gross contribution is price minus the cost to serve, where the cost includes each stream's allocated share of the $3.15M ROOT running cost. It is a contribution margin, before full overhead, sales and marketing, and depreciation. It is not net profit.
The data streams — IS1 and IS2 — carry very high contribution because they reuse the already-built model with almost no new field work. The two thin streams, outcome verification and accounting, are field-heavy and earn only about 30%; they are carried for the relationships and credibility they open, not for their margin. Blended across the portfolio, the gross contribution is about 86%.
Why the margin widens: 37% to 71%
Early on, the fixed $3.15M a year dominates, because there is little coverage to spread it over. As licensed coverage ramps, that fixed cost spreads across hundreds of millions of hectares, and the operating margin rises. In the model, the EBITDA margin moves from about 37% in the early years to roughly 71% by 2035, where EBITDA is about $77M on revenue of $108M. The margin expansion is not an assumption layered on top; it is the arithmetic of a fixed cost meeting a growing revenue base.
Tax, royalties and working capital
- Tax is low by law. Operating in the Amazon region under Peru's Law 27037, the corporate income-tax rate is 5% through 2048, against the standard national rate.
- The environmental-DNA layer carries a 15% access-and-benefit-sharing royalty on its revenue, under the Nagoya Protocol, reflected in the IS2 and IS3 biodiversity costs.
- Working capital reflects who pays. Government and jurisdiction clients pay slowly (about 120 days), corporate data clients faster (about 60 days), and vendors are paid at about 45 days. The model carries this drag explicitly.
The honest read
The margins above are contribution margins, struck before full company overhead and before sales and marketing, so they are not the same as net profit at each line; the 71% figure is the modelled EBITDA margin at scale, not a margin available today. Every per-unit cost is a conservative at-scale own-build estimate carried with a low-to-high band, the volumes are management estimates, the company is pre-revenue, and the whole structure is gated on the first signed contract. The amortisation treatment is unconfirmed until the IFRS auditor rules. What is structural, and not an assumption, is the shape: four cost centers, one of them fixed and built once, and a near-zero cost to sell the result many times.