Hey, we're Alejandro Soumah and Affaan Mustafa, co-founders of Itô Markets.
Overview
Compute is becoming an asset class. It is also the largest cost line in AI, and nobody can price it forward.
Itô is the agentic exchange for compute: an agent-run OTC desk for sourcing, selling, hedging, and contracting GPU compute. We buy capacity forward, hedge it, and deliver one flat fixed rate for the term.
Ask: if you buy GPU compute, run a datacenter or neocloud, or lend against GPU contracts, we want to talk. Reach us at [email protected].
The problem
Compute breaks the budget from both sides. Buyers cannot plan their largest cost line. Sellers cannot show lenders revenue certainty.
AI startups have two workarounds today, and both fix the price by locking cash and capacity before demand is known. On-demand GPU prices swing violently month to month, so founders overpay for flexibility. Reserved contracts demand one to four year commitments and annual prepayment that seed to Series A companies cannot carry. No founder should have to over-commit capital to reservations they may never use.
Datacenters have one workaround, and it mostly fails:
Lenders demand revenue certainty on $10M–$25M contracts over three to four years.
Near-term demand is financeable; months 30–60 are not, because nobody buys compute that far forward.
Insurance and reinsurance wraps are the only product on offer for that tail, and they mostly fail there.
What the market needs is a forward market: a desk that prices the forward and transfers the tail to parties who want the other side of it.
Our solution
Itô is one desk serving both sides of the market. Buyers get compute on better terms. Suppliers get their capacity sold and their tail risk hedged.
The client pays one fixed GPU-hour rate under one SLA. The desk sits between the client and the whole market: floating supply from a pool of 10+ providers, hedges laid off with exchange-cleared venues and hedge counterparties, and financing partners behind the desk.
Selling a fixed rate means the desk loses if compute prices rise, so we pay a counterparty to take that risk. If spot rises, the hedge pays: the client rate holds and the desk stays flat. If spot falls, supply gets cheaper: the hedge costs only its premium and the rate holds.
Buyers get firm quotes on better terms.
Describe the workload, GPU type, count, region, term, and the agent structures the deal, sourcing comparable quotes side by side across the whole supplier pool.
Suppliers get sold, hedged, and financed.
The contract is generated from the selected quote, and the position lands in the trade terminal where the desk hedges it. GPU-secured credit is originated at 70–80% LTV, priced off our tape.
Run an RFQ.
Your request is collected once and sent instantly to 20+ compute providers, across Slack, email, WhatsApp, and phone so quotes come back to you in minutes, not days.
Traction
$350k
Notional Volume Traded
3
Hedge Fund Partners
10+
Compute Providers
W1
W2
W3
Vision
The end state is the Amazon of compute. Amazon started as an intermediary, used the flow to price everything, and then bought the underliers. We run the same play. Now: the agentic intermediary, market maker, and architect of the market. Next: when the debt cycle turns, the desk already prices everything and its risk is offset, so it buys capacity at distressed prices. Then: own datacenters and the underlying assets, and replicate the desk across power, inference, and semiconductors.
The market is $750B+ in data center capex, with $180B in rented GPU compute by 2030. It is following the sequence every physical commodity followed: indices first, then cash-settled futures, then the physical desks that complete the market. In rates, in power, in agriculture, the desks that arrived at this stage kept the flow permanently. That is the seat we are taking while it is still empty.
Team
CIO – Alejandro Soumah ran the custom credit baskets desk at Goldman Sachs, growing it from $30M to $120M. The structured-products and custom basket machinery he built there maps directly to the desk Itô is building.
CEO – Affaan Mustafa was a founding engineer at a prediction-markets exchange scaled to $45M in volume, and built ECC, the 18th-largest GitHub repository in history. He has spent years turning messy technical workflows into products people actually use.
