Skip to content
[01]
Rule no. 01

Published, not transacted. The front page of the physical layer of AI — companies, capital, and compute, sourced and dated.

Cadence
Saturday issue. Daily index. Continuously updated directory.

Published, not transacted.
The front page of GPU infrastructure.

Weekly analysis, company profiles, live GPU pricing, and a directory of 2,633+ companies across the stack that makes AI physically run.

Trusted by teams at NVIDIA · CoreWeave · Nebius · Brookfield · Bloomberg

H100 SXM · on-demand · median
$2.49
USD / GPU-hr
range $1.57 – $10.14
Coverage
2,633
Companies tracked
37+
Pricing providers
364+
News sources
3,750
Newsletter subscribers
[02]
Rule no. 02 · Rate floor

The lowest verified on-demand rate for each flagship accelerator family, sourced from public provider pages.

Unit
USD per GPU-hour. On-demand only. No reserved, no spot.
Lowest verified on-demand · USD / GPU-hr
H100
$1.90
Hyperstack
GB200
$2.55
WhiteFiber HPC
B200
$2.55
IonStream
H200
$2.06
IonStream
A100
$1.15
Denvr Dataworks
MI300X
$1.71
TensorWave
Coverage
6
Families tracked
37+
Public providers
Full index
[07]
Rule no. 07

One email a week. Saturday morning. The deals that moved, the deployments that shipped, the capital that landed.

A weekly briefing for the physical layer of AI.

Every Saturday: deals, deployments, and capital moves across AI infrastructure. In-depth company profiles. Live pricing data. Read by teams at NVIDIA, CoreWeave, Nebius, Brookfield, and hundreds more.