IA On Demand
On-demand GPU and inference for Latin America, delivered over the same measured network this site publishes. Fixed price in BRL: the global GPU auction is our problem, not yours.
THE PHYSICAL ARGUMENT
Every token generated outside the continent pays a distance toll before your user sees it. Physics does not negotiate. Routing does.
- Teresina → Miami, measured: the toll paid before inference even starts…
- Teresina → São Paulo, measured: where this inference will live…
What we are building
GPU by the hour, fixed price in BRL
On-demand compute, no minimum contract. Spin up, train, infer, shut down: pay for what you used, in BRL, with a nota fiscal.
Managed inference
API endpoint for open models, served from inside our network. The token is born next to your user.
Dedicated instances
Reserved GPU in our datacenter, behind our AS. Your weights, your data, on national soil.
Under construction means under construction: no invented figure will appear here. When this page shows a number, it will resolve through a source, like every other number on this site.
THE ENGINE
An intelligent buyer, not a panel reseller
When you request capacity, our engine queries the global GPU market (multiple providers, plus our own datacenter) and contracts wherever capacity is available at that moment. The dollar-denominated auction stays on our side of the invoice. On yours: a fixed price, in BRL, with a nota fiscal.
Same doctrine as our international cloud. The invoice is the product.
You choose the geography
Sensitive workload? A dedicated instance in the datacenter behind this network. Your data does not cross a border to generate a token, verifiable on the looking glass. Flexible workload? The engine hunts capacity worldwide and your price stays fixed, in BRL.
EARLY ACCESS
The first customers shape the product with us
Early access by invitation, in order of arrival. Tell us what you want to run, whether training, fine-tuning or inference, and we build the capacity queue from real demand.