Skip to content

Announcing Liquid Compute — a $15M seed round to build a regulated venue that commoditizes intelligence

Read the announcement

[ The financial frontier for the AI economy ]

Compute has a price.
Now it has a market.

Liquid Compute is building the regulated exchange for compute — price discovery, clearing, and settlement for neoclouds, AI companies, and traders.

Capacity becomes transferable through a physical order book, with derivatives markets on top.

* PMEX Markets is pending designation as a DCM and registration as a DCO by the CFTC.

IndexLastChg
  • LCI-A100-US$1.53down
  • LCI-H100-US$2.69up
  • LCI-H200-US$3.42up
  • LCI-B200-US$5.01up
  • LCI-B300-US$6.66up

USD per GPU-hour · continuous prints from the physical order book

Physical brokerage — liveIndex licensing — institutional counterpartiesCFTC — DCM / DCO applications pendingYC W25 · New York

[ 01 · The asset ]

Oil can be stored in a barrel. Compute cannot.

If a GPU sits idle this afternoon, that capacity evaporates forever. Its value depends on location, configuration, and workload — closer to power markets than to oil, and a direct proxy for access to intelligence.

F1

Bilateral is brittle

Every resale needs a new contract, a fresh credit decision, another payment setup, a separate SLA. As capacity changes hands, liability tangles.

F2

No trusted price

The same GPU rents at three prices depending on how long you commit. The term structure exists — it just isn't standardized anywhere.

F3

The buyer changed

Compute is no longer bought only by its end user. Neoclouds, brokers, and trading firms warehouse it, resell it, and arbitrage it.

Precedent — ICE in energy · CBOE in volatility · Kalshi in events: regulated venues accrued most of the value of the commoditization they enabled.

[ 02 · The market ]

The market stack for compute.

Every trade needs the same plumbing: a trusted price, a place to trade, a way to settle, and a way to finance it. Liquid Compute builds all four.

  1. Data · Indices

    Prices in

    Verifiable physical reference prices. Continuous index prints for the GPUs that actually move the market — from A100 to B300.

  2. Exchange

    Trade

    A physical order book for capacity. Standardized contracts replace bespoke bilateral deals, so positions change hands in one place.

  3. Settlement

    Clear & deliver

    Clearing through a common framework. Delivery is verified, liability stops tangling, and credit is netted — the venue does the plumbing.

  4. Credit

    Finance

    Finance against the forward curve. Term structure lets lenders underwrite cash flows and operators unlock working capital.

[ 02 · The market ]

One chip. Three prices. That’s a term structure.

The same GPU rents at three different prices today depending on how long you commit. That is not a broken market — it is a forward curve, and people have already started trading it.

$10$8$6$4$2$9.00$5.75$3.85On-demandReserved · 3–12 moMulti-year · 36 mo
USD per GPU-hour · illustrative levels, current generation

[ 03 · The products ]

Three ways onto the market.

Every request starts the same way: tell us your side, your chip, your quantity, your start date and your tenor. The desk prices it and shows you what it does to your book.

Physical

Buy / Sell Capacity

Trade blocks of committed GPU capacity against verifiable reference prices — on the side that matches your book.

Operators locking forward revenue, buyers capping future cost, desks running tenor transformation.

Side
Buy or Sell
Chip
A100 · H100 · H200 · B200 · B300 · Other
Quantity
GPU-hours under contract
Start
Desired start date
Tenor
Duration in months
Request access
Cash-settled

Trade GPU Indices

Take a view on the price of compute — long or short — against the US indices Liquid Compute publishes, without touching hardware.

Relative-value traders, hedgers laying off residual risk, anyone expressing a view on a chip generation.

Side
Long or Short
Index
LCI-A100-US → LCI-B300-US
Quantity
Index-referenced hours
Start
Desired start date
Tenor
Duration in months
Request access
Credit

Finance Compute

Convert forecast revenue into contracted revenue, and finance against a residual the market can actually see.

Neoclouds lowering cost of capital, lessors bridging the lease/offtake gap, lenders advancing against hedged fleets.

Structure
Forward sale · hedge · residual protection
Chip
A100 · H100 · H200 · B200 · B300 · Other
Reference
Published index or physical print
Start
Desired start date
Tenor
Duration in months
Request access

On the book today

Supply

Neoclouds & datacenters

Lock in future utilization revenue, move idle capacity into a liquid book.

Demand

AI companies & enterprises

Hedge rising inference costs with standardized instruments.

Trade

Trading firms & market makers

Warehouse, resell, and arbitrage compute with a balance sheet.

Capital

Lenders & credit desks

Underwrite financing against forward prices and verifiable prints.

[ 04 · The mechanics ]

Tenor transformation. The first trade in every young market.

Blocks are trading right now: buy multi-year capacity, resell it shorter, run the book the way a desk runs any carry position. Standardized contracts are what turn an operator’s trade into a trade available to capital — exit by assignment instead of by rack relationships.

Trade pattern and levels per our research, Compute Is Already Trading — figures illustrative, not drawn to scale.

01 · Buy the block

36 months at $3.70

Multi-year capacity clears at a discount — financed operators are structural sellers of term.

02 · Sell the near year

Months 1–12 at $5.50

Shorter commitments clear higher. The first year is sold and the spread is locked.

03 · Carry the tail

$1.80 locked

Months 13–36 are the position: payment for holding re-let risk and obsolescence risk the short buyer refuses.

04 · Lay off the risk

Short the index

A cash-settled short against the unsold tail converts the carry into a managed book instead of a bet.

[ How the hedge works ]

A contract that fixes a price for a future period.

Nobody delivers a GPU. Nothing about your operations changes. At the end of the period the agreed price is compared to the printed index and the difference settles in cash. You already hold the exposure — the hedge is how you put it down.

  1. S1Your exposureThe renewal you haven't priced, the utilization you haven't sold.
  2. S2Fix the levelA hedge struck at $4.40 for the period you need.
  3. S3The index printsLCI publishes continuously for the chip, for the whole period.
  4. S4Cash settles the differenceAgreed price vs printed index. No GPUs change hands.
LongHurt when prices fall
  • Operators with unsold capacity — hours not yet under contract
  • Lessors and owners at lease end — hardware value tracks the rental price
  • Anyone whose offtake is shorter than their financing

Lock in a price you can sell at

ShortHurt when prices rise
  • Enterprises and AI companies — renewal cost unknown, budget due today
  • Buyers who cannot commit long — short contracts mean constant repricing
  • Anyone growing into more compute — future cost base rises with the market

Lock in a price you can buy at

Case · An enterprise fixing what compute will cost

Your renewal is twelve months out and the budget is due now. You need 128 GPUs for a year — about 1.12 million GPU-hours.

Where the market landsUnhedged billHedged at $4.40
Falls to $3.80$4.26M$4.93M
Flat at $4.40$4.93M$4.93M
Rises 25% to $5.50$6.16M$4.93M

Your procurement is untouched: run your RFP, pick your provider. The hedge only fixes the level. No GPUs arrive — it settles in cash against a published index.

[ 05 · The proof ]

The record so far.

A $15M seed, pending CFTC applications, a live physical brokerage, and published indices to settle against.

$15MSeed round — Chemistry & Firstmark
YC W25Combinator batch, New York
5US indices published, A100 → B300
2CFTC applications pending — DCM & DCO
LivePhysical brokerage — blocks trading today

Backed by

  • Brainchild Holdings
  • UFO Holdings

Working with

[ Before / after ]

Bilateral today. A market tomorrow.

AspectToday · bilateral OTCThe venue
Opening a positionNew contract, fresh negotiation, another counterparty reviewStandardized contract — enter and exit by assignment
CreditA fresh credit decision on every tradeOne framework — clearing through the venue
Payments & settlementAnother payment setup, tangled chains of liabilityNetted and settled by the market
Price discoveryEvery price is private; the same GPU rents at three prices and nobody publishes whyPublished reference indices, verifiable physical prints
Service levelsA separate SLA per deal, operational reach requiredDelivery verified against defined performance metrics

Participants keep their counterparties and their procurement. The venue replaces the plumbing, not the relationships.

“Trading is not what happens after a market matures. Trading is how it matures.”

Stanley Lee · Chief Product Officer

The instruments are new. The trades are not.

[ 06 · Access ]

Asked before onboarding.

What does “pending DCM / DCO” mean?

PMEX Markets has pending applications before the CFTC for designation as a Designated Contract Market and registration as a Derivatives Clearing Organization — the same institutional-grade framework that governs traditional commodities. Regulatory outcomes are not guaranteed, and nothing on this page is an offer to sell or a solicitation to buy any security or commodity interest.

What is actually traded?

Standardized contracts on AI infrastructure, both cash and physically settled, starting with compute. Before a futures market can launch, the industry must standardize exactly what is being delivered — which is why verifiable infrastructure, defined performance metrics, and credible physical reference prices come first.

What is an EFP market?

Exchange for Physical. Clearing transactions through a common framework lets any participant with a balance sheet take a position on compute without becoming a cloud operator, a credit underwriter, or an AI company.

Who can participate?

Trading firms, datacenters, cloud providers, lenders, and everyone else trading or financing capacity. The venue is onboarding now — request access and the team will be in touch.

Where do reference prices come from?

Index prints from the physical order book — continuous prints for the accelerators that actually move the market, from A100 to B300. The US indices Liquid Compute publishes are licensed today to institutional counterparties, and physical brokerage is live — blocks are trading now.

Does anyone deliver a GPU?

Physical capacity trades in the book; the cash-settled market needs no racks at all. A hedge fixes a price for a future period and settles in cash against the printed index — your procurement, your providers and your operations stay exactly as they are.

What does it take to start?

An ISDA with the counterparty, a licence for the index it settles against, and a view on how much of your exposure you want fixed. Most desks start with a slice rather than the whole book — which is the right instinct. Hedge the volume you are confident about; a hedge fixes price, not volume.

[ Now onboarding ]

Request access to the compute market.

  • Trade against verifiable index prints, not rate cards.
  • Standardized contracts instead of bespoke bilateral deals.
  • Clearing and settlement handled by the venue.

Onboarding trading firms, datacenters, cloud providers and lenders now — tell the desk your side, chip, quantity, start date and tenor.

Access request
Interest

By submitting you agree to the access request privacy terms.