GPU
fabryka.ai price index
10/4/2026, 8:17:56 PM UTCJSON API
All GPU Types · MLPerf-Normalized

GPU Price Index

$2.26-0.73 / -24.3%

Daily USD per H100 SXM-equivalent compute-hour: each GPU's price divided by its MLPerf-derived throughput, so every card type lands on one comparable axis.

P25$1.84
P75$2.88
30D+0.0%
Obs109
Meta Index · class-balanced
69.9-30.1%

Base 100 · $2.40 / H100-eq hr · median of 20 GPU classes (one vote each, listing volume removed).

Confidence Bands

H100 SXM-Equivalent Index

$2.26median / gpu-hour
$0.66$1.81$2.95$4.10$5.25Jun 2Oct 4
min-max observed range p25-p75 confidence band median index

Meta Index History

The class-balanced meta index over time, rebased to 100 at the start of the series. One vote per GPU class, normalized by MLPerf throughput, sourced from ~7 months of Vast.ai daily medians.

Meta Index · class-balanced (base 100)
69.9 · $2.40/H100-eq
-30.1% since base
62738596108Jun 2Oct 4

Historical Index

The public dataset is one MLPerf-normalized, cross-GPU price series with daily distribution bands.

Normalized H100 Index
$2.26
p25-p75 range median
$1.49$2.11$2.74$3.36$3.99Jun 2Oct 4

Variant History

Daily raw median values by GPU class.

Variant Medians
Jun 2 - Oct 4
RTX 3090 RTX A5000 V100 RTX 4080 RTX A6000 RTX 4090
$0.00$0.25$0.50$0.75Jun 2Oct 4

Recent Values

The latest daily index statistics.

DateObservationsMinP25MedianP75Max
2026-10-041091.00751.83702.26332.88474.9000
2026-06-02241.85242.40002.99003.64004.8058

Current Variant Prices

Raw variant prices remain visible below the single normalized headline index.

ClassObservationsMinP25MedianP75Max
RTX 3090 24GB120.12960.16180.17650.17810.5000
RTX A5000 24GB20.16000.18750.21500.24250.2700
V100 32GB20.19000.20000.21000.22000.2300
RTX 4080 16GB20.27000.27250.27500.27750.2800
RTX A6000 48GB80.33000.40370.40370.47230.5300
RTX 4090 24GB120.33440.37560.40220.41720.7400
A100 SXM 80GB120.40300.46540.61410.73481.5900
A100 PCIe 80GB110.40370.51760.80221.24001.5900
L4 24GB10.49000.49000.49000.49000.4900
L40S 48GB90.53560.79000.80150.80221.0900
A100 40GB21.00001.02501.05001.07501.1000
H100 SXM 80GB111.89631.94282.40002.88144.2900
H100 PCIe 80GB41.90001.96752.44002.99003.2900
H100 NVLink 80GB11.95001.95001.95001.95001.9500
H100 NVL 94GB82.37482.57622.83562.96013.1900
MI300X 192GB12.39002.39002.39002.39002.3900
GH200 144GB13.19003.19003.19003.19003.1900
H200 SXM 141GB73.59003.79004.59005.06295.1213
H100 HGX 80GB13.90003.90003.90003.90003.9000
B200 180GB25.98006.18256.38506.58756.7900

Performance Normalization

MLPerf-derived throughput per GPU class (H100 SXM = 1.00×) converts each raw price into a comparable H100-equivalent compute-hour.

ClassPerf vs H100SourceRaw median $/hrNorm. $/H100-eq hrObs
A100 SXM 80GB0.40×MLPerf0.61411.535312
RTX 3090 24GB0.11×spec est.0.17651.604512
V100 32GB0.12×spec est.0.21001.75002
H100 NVLink 80GB1.00×MLPerf1.95001.95001
RTX 4090 24GB0.20×spec est.0.40222.011012
RTX 4080 16GB0.13×spec est.0.27502.11542
B200 180GB3.00×MLPerf6.38502.12832
RTX A5000 24GB0.10×spec est.0.21502.15002
A100 PCIe 80GB0.36×MLPerf0.80222.228311
MI300X 192GB1.00×MLPerf2.39002.39001
H100 SXM 80GB1.00×MLPerf2.40002.400011
RTX A6000 48GB0.16×spec est.0.40372.52318
GH200 144GB1.20×MLPerf3.19002.65831
H100 NVL 94GB1.05×MLPerf2.83562.70058
L40S 48GB0.28×MLPerf0.80152.86259
A100 40GB0.34×MLPerf1.05003.08822
H100 PCIe 80GB0.78×MLPerf2.44003.12824
H200 SXM 141GB1.40×MLPerf4.59003.27867
H100 HGX 80GB1.00×MLPerf3.90003.90001
L4 24GB0.10×spec est.0.49004.90001

Method

Each daily snapshot stores aggregate index statistics.

MetricDefinition
Normalized indexMedian USD per H100 SXM-equivalent compute-hour: each offer's price ÷ its MLPerf-derived perf factor, across all GPU classes.
Perf factorPer-GPU throughput relative to H100 SXM (=1.00×), from MLPerf Inference where available, else estimated from FP16/BF16 specs.
Variant pricesRaw median USD per GPU-hour within each GPU class, before normalization.
SpreadMin, p25, p75, and max show the daily market range.