Data · Indices

The reference price for GPU compute.

Transparent, published benchmarks for B100, B200, H100 and H200 — the settlement basis the entire market prices against.

Request index accessLog in
H100 · Spot / GPU-hr
$2.57+0.60%
MARKETLASTBIDASK
H100$2.57$2.56$2.57
H200$3.28$3.27$3.28
B100$4.55$4.55$4.56
B200$5.90$5.89$5.91
Designed to be cleared through a CFTC-registered clearinghouse (DCO)ILLUSTRATIVE
01 — What the index gives you
TRANSPARENT
Published methodology

Every input, weighting and calculation window is documented and versioned — no black box between spot and settle.

COMPREHENSIVE
Full accelerator coverage

B100, B200, H100 and H200 on-demand and reserved, normalised to a common $/GPU-hr unit.

DISTRIBUTED
API, feed & fixings

Real-time API, streaming feed and a daily settlement fixing your desk and systems can reference directly.

02 — Methodology

How a compute reference price is built.

01
Collect

Aggregate transacted and quoted rates from qualifying venues and providers.

02
Normalise

Adjust for region, term and instance type to a common $/GPU-hr basis.

03
Filter

Trim outliers and stale prints inside a defined observation window.

04
Publish

Release the fixing over API and feed with a full audit trail per print.

4Accelerator classes indexed
DailyPlanned settlement fixing
<250msStreaming feed latency target
v2.1Draft methodology version
Downstream — the rest of the market
LAYER 02
Exchange →

Trade futures and options priced off the index.

LAYER 03
Settlement →

Convert a contract into delivered, running GPUs.

LAYER 04
Credit →

Finance buildout against contracted compute.

Price your book against the index.

Request access