Compute for EquityWe take 1.5% when a trade clears.
A regulated clearing house. We process trades between energy, compute and AI equity, and take about 1.5% per settlement — at an 85% gross margin, owning no GPUs.
~9 min readThe unit of trade
Everything we monetize is one repeatable event: a deal. An AI startup needs compute. A datacenter or energy producer deposits energy and GPU-hours. Our Oracle prices the contribution and the equity it converts to. A Compute-SAFE is issued, custodied under ADGM, and cleared.
We are the neutral party in the middle of that event — and we charge a thin fee for making it happen safely. The whole business is: how many deals clear, and at what size.
Four revenue lines
By Year 3, revenue mix (hover a slice):
| Line | Basis | % of Y3 rev |
|---|---|---|
| Clearing fee | 1.5% of GMV (cleared volume) | 78% |
| Origination fee | 0.5% of supply listed | 12% |
| Treasury float | 3.0% on average custody balance | 7% |
| Oracle data | Subscriptions to the cross-asset index | 3% |
Unit economics, per deal
Take the average deal — $250,000 of GMV:
| Clearing fee (1.5%) | $3,750 |
| Origination fee (0.5%) | $1,250 |
| Treasury float (12-mo custody) | ~$2,500 |
| Revenue per deal | ~$7,500 |
| Gross margin (capital-light, no inventory) | 85% |
| Contribution per deal | ~$6,400 |
Volume math
At a 2% blended take, revenue is a clean function of cleared volume:
| Target revenue | GMV needed | Deals / year |
|---|---|---|
| $1M ARR | $50M | 200 |
| $10M ARR | $500M | 2,000 |
| $100M ARR | $5B | 20,000 |
The market is large enough that these are throughput problems, not demand problems.
Network effects
Each side reinforces the other — the flywheel of every great marketplace:
More supply tightens the Oracle's price; a tighter price attracts more demand; more demand pulls in more supply. A secondary market makes every position more valuable — and every new deal sharpens a proprietary cross-asset dataset no single lender or cloud can replicate.
Defensibility — three moats
Regulatory
An ADGM/FSRA RegLab wedge for tokenized Compute-SAFEs — slow to copy, and exactly what unlocks sovereign and institutional capital.
Data
The Oracle + a 3,000-entity mapped market. Every cleared deal improves the cross-asset price.
Network
Multi-sided clearing plus a secondary market — liquidity no bilateral counterparty pair can offer.
Comparable models
Clearing venues are capital-light, take-rate businesses. Our take is higher because we price a nascent market with real complexity — not a commoditized order book:
| Venue | ~Take rate | Market |
|---|---|---|
| NYSE | ~0.003% | Mature equities |
| CME | ~0.03% | Mature futures |
| Coinbase | ~0.5% | Crypto |
| Compute for Equity | ~1.5% | Compute-for-equity (nascent) |
The path to $100M ARR
$100M revenue ÷ 2% blended take = $5B GMV ÷ $250k average deal = 20,000 deals/year ÷ 250 working days = 80 deals a day. Against a GPU-rental market growing from $9.8B to $47.2B by 2033 and a $490B financing gap, that is throughput a clearing platform can reach.
What each side pays
| Party | Origination | Clearing | Secondary |
|---|---|---|---|
| AI startup (buys compute) | — | on settle | — |
| Provider (lists supply) | 0.5% on listing | on settle | — |
| Investor (trades SAFEs) | — | — | 2nd-leg clearing |
The startup — the party we most want to attract — pays nothing to receive compute. Revenue comes from the side that captures upside, and from liquidity events.