AI Grid: Inference Router

Set the latency budget, sovereignty constraint, and demand for an AI inference workload. Watch the eligible-region set collapse around your constraints.

Workload Demand 100 req/s Model tier required

Constraints Latency budget (from UK egress) 120 ms Sovereignty Headroom required 20%

Headroom is the spare capacity a region must have before it's eligible — loaded regions are excluded.

Strategy Cheapest — pack onto lowest $/M tokens first Fastest — pack onto lowest-latency first Balanced — score by latency × price Resilient — spread across ≥3 jurisdictions

Eligible regions —

Weighted latency —

Weighted price —

Jurisdictions used —

—

Read the full AI Grid insight →