Exascale

NeoCloud & Exchange · Spot · Forward · Futures

The Global Exchange for AI Compute

Exascale is the exchange where data centers sell compute capacity, AI labs buy it, and traders make the market — one venue for GPU spot, futures and compute credits.

04ExchangeOpen account
03AI PlatformGet API keys
02GPU CloudPrice a cluster
01Data Center NetworkList capacity
Exascale H100-SXM IndexUS-EAST
2.1908USD / GPU-hr▼ 2.43%

Bids

2.190569603
2.190059600
2.189574291
2.189033292
2.188554130
2.188036258
2.187517841

Asks

2.191549309
2.192084999
2.192573452
2.193024784
2.193542177
2.194038632
2.194539103

Built to clear capacity across the world’s clouds & neoclouds

The market

A price discovered, not negotiated.

Capacity is still bought the way commodities were before they had exchanges — bilaterally, at a price set by the relationship, on terms neither side can hedge. Buyers over-contract rather than risk being short; operators hold racks they have no orderly way to sell. Both are the same missing institution. Exascale quotes both sides of it on one continuous book.

H100-SXM · quoted by regionclearing 2.1440 USD/GPU-hr
Hover a region to inspect its spread and basis

The stack

Megawatts at one end, tokens at the other.

Capacity is contracted and metered as megawatts, sold on as GPU-hours, served as tokens, and quoted on a book that prices the two units beneath it. Enter at whichever unit your work is already denominated in; what you stop needing goes back on the same book.

Layer 04Buy / Sell

Exchange

A central limit order book over the two units below it, cleared and settled in dollars.

Who
Traders
Why
To hedge and arbitrage
What
Spot · forwards · perps
Open the terminal
  1. TradersBuy / Sell
    04

    Exchange

    GPU-hours and AI credits

    Spot · forwards · perps
  2. AI companiesBuy side
    03

    AI Models

    Every modality, one schema

    AI tokens
  3. AI companiesBuy side
    02

    GPUs

    Train, tune and serve, all sites

    GPU-hours
  4. Data centersSell side
    01

    Data Centers

    Someone else’s hardware, our meter

    Megawatts

Three sides of the market

Supply, demand, and the price between them.

01 / SUPPLY
Data centers & neocloud operators

Meter the racks you are not selling.

List spare capacity the moment it frees up and let the market price it. Turn stranded power and dark racks into continuous, metered revenue.

  • Per-second metering & payout
  • Proof-of-capacity attestation
  • Dynamic pricing by region & SKU
02 / DEMAND
AI enterprises & frontier labs

Buy the run, not the reservation.

Provision training and inference capacity at the market clearing price — no year-long lock-ins, no idle reservations. Burst on spot, secure runs on futures.

  • Instant multi-region provisioning
  • Lock forward price & capacity
  • One settlement account, any partner
03 / LIQUIDITY
Traders & market makers

Take the price, or make it.

Make markets, arbitrage regional price gaps, and hedge exposure with GPU futures and compute credits. One book for spot, futures and credits, settled T+0.

  • Spot, futures & options
  • Fungible, transferable credits
  • FIX / REST / WebSocket APIs

Instruments

From the next hour to the next quarter.

Spot for capacity needed now, forwards that fix a price and reserve the metal, perpetuals that take the price without the delivery, and credits that hold a balance against any partner. Four instruments, one unit of account, one settlement.

Credits

Compute credits

Prepaid, escrowed dollar balances spendable against any partner on the venue — transferable, and redeemable at the clearing price.

Escrowed · settled in USD
Spot

Buy compute now

Immediate provisioning priced per GPU-hour, matched at the best available bid or ask.

Settle T+0 · priced per GPU-hour
Forwards

Lock the curve

Fix price and reserved capacity weeks to quarters ahead, and take delivery of the capacity itself at expiry. Hedge a planned training run against demand spikes.

Dated tenors · physical delivery
Futures

Perpetual futures

No expiry and no delivery — a cash-settled position on the GPU-hour price, held for as long as funding is paid. Take or hedge a view without ever touching capacity.

Perpetual · cash-settled · no delivery

Settlement pipeline

From order to settlement without leaving the venue.

An order crosses on the book, capacity is allocated on a partner’s metal, usage meters to the second against an auditable ledger, and funds clear the same day. Five stages, one system — not five integrations.

01

Order

Post a bid or ask to the central book via console or API.

Console or API
02

Match

The engine crosses orders at best price, price-time priority.

Price-time priority
03

Provision

Winning capacity is allocated on the partner facility's metal.

Allocated on partner metal
04

Meter

Usage is metered per-second against an auditable settlement ledger.

Auditable ledger
05

Settle

Funds clear T+0; credits reconcile against escrow automatically.

T+0 clearing

Market structure

The mechanics a counterparty asks about first.

What makes a market is not the introduction but what stands between the two parties when a trade goes wrong: who carries the credit, where the money sits, which price is authoritative, and what is actually delivered.

Central counterparty

Every match is novated to the venue, so neither side carries the other’s credit. One exposure to clear, not a web of bilateral ones.

Segregated escrow

Client funds sit apart from the venue’s own balance sheet and are reserved per trade, not commingled and drawn against later.

One reference price

A single continuously discovered price per accelerator class and region, published to every participant on the same terms.

Delivery or cash

Forwards deliver the capacity itself; perpetuals settle in cash against the index. The same book prices both.

Clearing & custody

Exascale is the counterparty to both sides.

Every match is novated: the venue becomes buyer to the seller and seller to the buyer, escrows the credits, and guarantees settlement. A lab never fronts capital to an operator it has not underwritten, and an operator is paid for metered work whether or not the buyer turns out to be good for it. Capacity is verified by attestations pulled from the partner’s own metal.

Central counterpartyNovated at matchSegregated escrowProof-of-capacity
Order bookCentral limit, price-time
Clearing modelNovated to the venue
Client fundsSegregated, reserved per trade
Capacity proofCryptographic attestation
MeteringPer-second, auditable

Live order book

The book, cumulative by price.

Every resting bid and ask on H100-SXM. Hover the curve for price, size, and distance from mid.

Bids 5,259 Asks 6,746
H100-SXM · US-EAST · cumulative GPU-hr
mid 2.1440
43.8% bid / 56.2% ask · book imbalance

The exchange, end to end

What has to be there before the first trade.

Multi-region supply

Capacity aggregated across operators, regions and continents.

Accelerator-agnostic

NVIDIA and AMD classes quoted on one book, and new silicon as it reaches the market.

Central limit order book

Continuous two-sided markets, matched on price-time priority.

T+0 settlement

Same-day clearing, netted and settled in US dollars.

Segregated escrow

Counterparty funds held apart and reserved per trade.

Per-second metering

Usage billed to the second on an auditable ledger.

Proof-of-capacity

Cryptographic attestations pulled from partner metal.

Programmable

REST, a WebSocket feed, and a FIX 4.4 gateway.

Request access

Whichever side you are on.

Onboarding for data-center operators, AI enterprises, frontier labs and trading desks, in rolling cohorts.

Open a trading accountBuy computeList your capacity