THE AGENT TAX

Make inference latency zero. The task still takes 70 seconds.

One agent task, 1,000 calls, identical hardware. The only difference is where the GPUs sit.

GPUs 2,400 miles away70 ms RTT
GPUs across town · Perimeter3 ms RTT
2,400 miles · task time70,000 ms
Across town · task time3,000 ms
The distance penalty23×

The task: 20 turns, 50 calls each; every call, model or tool result, is one round trip. And both trends point the same way: agents keep getting longer (METR, 2025: the length of tasks AI can complete doubles about every 7 months) while compute keeps getting cheaper (Stanford HAI, 2025: inference cost fell about 280x in 18 months).

HOW WE MEASURE IT

Compute shrinks. The wire does not.

The task above is 1,000 round trips: 20 agent turns, 50 tool or model calls each. Every call pays the same wire cost no matter how fast the GPU answers, 70ms for a GPU 2,400 miles away, 3 ms for one across town. We dial compute per call from 750 ms down to 0 and hold turns, calls, and hardware fixed, so distance is the only variable left on the clock.

METR, 2025

7 months

The length of task an AI agent can complete on its own doubles about every 7 months. Longer tasks mean more calls, not fewer.

STANFORD HAI, 2025

280×

Inference cost fell about 280 times in 18 months. Compute keeps getting cheaper. The round trip does not.

THE STACK

Power, compute, and delivery: engineered as one system.

Perimeter Scout

Rapid site assessment using the Perimeter Scout platform. Handshake to deployment in 6 to 12 weeks.

Perimeter Grid

Rapid-scale, flexible deployments using the Perimeter Grid system.

Perimeter Relay

Optimized multi-region inference using the Perimeter Relay router and load balancer.

Compute, closer.