A network of GPUs. Not a single data center.
Independent providers contribute capacity. Developers route requests to the region and price point that fits each workload.
- Multi-region
- Policy routing
- Failover
Global pools
Regional capacity, one API.
Policy tags pick region, latency tier, and price ceiling, automatically.
EU pools
Keep workloads inside approved jurisdictions.
US + LATAM
Serve nearby hosts across both continents.
APAC edge
Cut round-trip time for real-time apps.
Failover
Backup routes when a node goes quiet.
Routing
Send traffic where it belongs.
Policy tags pick region, latency tier, and price ceiling. Fallback routes keep apps online when a host goes quiet.
Nearest host
Prefer the closest available GPU for interactive apps.
Region lock
Keep inference inside approved jurisdictions.
Price ceiling
Cap spend per request without manual switching.
Failover
Backup routes when a node goes quiet.
Visibility
See the network breathe.
Live maps of regional utilization, latency, and demand, so you know where capacity is moving.
