Compute for EquityAI runs on compute.
Compute runs on energy.
Equity is illiquid.
Until now, none of them could trade with each other. This is the whole story of the clearing layer for the AI economy — premise to plan, scene by scene.
~12 min readThe premise
Three of the most valuable assets of the AI era are energy, compute and equity. Each is enormous. Each is priced in its own silo. And none of them can settle directly against the others.
A startup cannot pay a datacenter in equity. A datacenter cannot bank AI upside instead of cash. An investor cannot hold a position denominated in compute. The pipes between these markets simply do not exist — and that absence is the entire opportunity.
The status quo is friction stacked on friction
Watch the cash flow through today's market. Every arrow is a layer of friction, markup and lost upside.
Four layers, and the equity sold at step one never touches the energy at step four. The trade that should be one hop is four hops of leakage.
The thesis: three orphan markets, one clearing house
It is not a capital problem — there is an estimated $490B of demand to build AI infrastructure. It is a coordination problem. The markets that should connect are kept apart because no neutral institution prices both sides, becomes counterparty to each, and clears the trade.
Energy
Abundant, cheap, stranded from AI upside.
Compute
Priced against cash, never against ownership.
Equity
The most valuable, least liquid asset of the three.
How it works — four steps
Originate
Providers deposit verified energy & GPU credits, tokenized as standard assets.
Price
The Oracle values the contribution and the equity it converts to.
Clear
Startup gets compute; provider gets a Compute-SAFE; we custody under ADGM.
Settle
Credits and positions become tradable on a transparent secondary market.
The instrument: the Compute-SAFE
A SAFE whose consideration is metered compute instead of cash. A provider commits a capped amount of compute, drawn over a window, that converts to equity at the startup's next priced round — bounded by a valuation cap, sweetened by a discount, and protected by an SLA-suspension clause.
Economically it behaves like a SAFE. Legally, it is a security — and we operate the market for it under ADGM/FSRA rules, not around them. Read the standard term sheet →
The engine: a cross-asset Oracle
The IP at the center. It prices a megawatt-hour against a GPU-hour against a slice of equity, in real time, from depreciation curves, utilization and market indices.
Because we own no compute, the quote is neutral — the precondition for a price both sides can trust. Try the live Oracle →
The market is already mapped
This is not a market we are imagining — it is one we have catalogued. Three thousand entities across tokenized compute, AI funds, and compute demand, searchable and filterable today.
1,000 tokens
DePIN / compute / AI networks.
1,000 funds
VCs, family offices, sovereigns.
1,000 demand
AI companies that need GPUs.
Why now
The problem becomes visible — founders burn raised cash on GPUs; GPU-backed debt appears.
The gap is named — compute-as-an-asset enters the mainstream conversation.
Compute is priced against cash — CME×Silicon Data and ICE×Ornn launch compute futures. We launch the equity leg.
The full marketplace and secondary market open.
Why Abu Dhabi — first
This company can only be built in one place first. The ADGM runs the world's most advanced framework for tokenized securities under English common law; the FSRA RegLab gives a clear path from sandbox to full license; the emirate pairs the cheapest energy on earth with a multi-gigawatt compute cluster; and sovereign capital — MGX, Mubadala, G42 — is already building non-dollar AI rails.
Nowhere else co-locates the regulator, the energy, the compute and the capital within a few kilometres. So we relocate and build on the ground. The regulatory path →
The road, and the ask
We do not need a thousand participants on day one. We need two great ones, then ten great deals, then a market. Phase 1 (2026): hand-brokered fat deals under RegLab. Phase 2 (2027): standardize and open to a vetted cohort. Phase 3 (2028): open marketplace and secondary.