io.net review

The published floor is $2.10 to $3.50 for a GPU hour, so entry cost sits well above the €0.63 median for the brands ranked on this measure. Real cost depends on the chip you pick from H100, A100 and RTX 4090 and how long the job runs.

The listed models are H100, A100 and RTX 4090. That covers common datacenter parts for training plus a consumer card for lighter inference, and it is a shorter list than the range at ComputePrices.com.

Hourly is the only billing model, listed as hourly. Monthly and reserved terms are not among it, so a job that runs for days costs the same per hour as a job that finishes quickly, with no commitment discount.

Capacity is spread across 130 locations, far above the 4 median for providers in this group. More locations mean more placement choice, though the useful spread depends on where your users and data sit.

Support runs on help centre and community, meaning a help centre and community channels only. There is no phone or chat route listed, which matters for teams treating a stalled training run as an incident.

The listed billing model is hourly alone, with no free allowance stated next to it in the way other providers list one. Budget for paid GPU hours from the first run rather than planning around a free trial.

The operator is io.net, inc., a company founded in 2022. The capacity is aggregated from many locations rather than run from a single owned data centre, which shapes how you plan support and commitment.

It fits work where chip choice and location spread matter more than entry price. If the lowest cost per GPU hour is the priority, a broader marketplace such as ComputePrices.com offers a longer chip list to compare against.