ex5.ai — Rent GPUs. Run any model. Ship AI.

COMPUTE INDEX

RTX4090 $0.21 ▾1.2%RTX5090 $0.41 ▴0.8%L40S $0.55 ▾0.4%A100 $0.64 ▾2.1%H100 $1.18 ▾3.0%H200 $1.49 ▾1.7%EX5CI 142.7 ▾2.3%sample data

// EX5 COMPUTE INDEX

142.7▾ 2.3% 24h

The market price of AI compute, in one number — the volume-weighted average $/hr per effective FP16 TFLOP across every verified host on the ex5.ai market. Indexed to 100 = Jan 2026. Falling index = compute getting cheaper.

7d ▾4.1%30d ▾9.8%YTD ▾28.4%ATH 198.2sample data

Rent below index →GET /index/ex5ci — free API

EX5CI — 90 DAYSdaily close · sample

May 19 · 158.3Aug 17 · 142.7

Class indices

volume-weighted · verified hosts onlysample data

CLASS

AVG $/HR

$/TFLOP·HR

LISTED GPUS

Consumer (4090/5090)$0.29$0.00081▾3.2%1,840Chart

Workstation (L40S/6000)$0.61$0.00172▾1.9%412Chart

A100 class$0.64$0.00205▾4.4%396Chart

H100 class$1.18$0.00119▾5.1%608Chart

H200 class$1.49$0.00075▾2.8%224Chart

EX5CI composite—142.7▾4.1%3,480Chart

METHODOLOGY

Rolling 24h volume-weighted mean of executed rental prices, normalized by benchmarked FP16 throughput per machine (not spec-sheet TFLOPS). Outliers beyond 3σ excluded. Published every 60s; history is free over the API.

WHY IT MATTERS

Benchmark your cloud bill against the open market, time your training runs, or settle compute contracts against a neutral reference — the way energy markets settle on a spot index. Hosts price against it too.