Deploy GPU applications on supported community hardware and cloud capacity
Cloud inference platform for deploying and benchmarking robot policies on GPUs
Stage
Startup
Startup
Founded
2026
2026
Based in
SA
US
Pricing model
Paid
Free trial
Pricing
One published price per GPU model in USD per GPU-hour. Hosts keep 90% of completed rental fees and Petabyte keeps 10%. A reservation holds funds, with final charges and returned balance following settlement after the run completes.
Starter: $0/month with $5 free credit monthly, up to $200 usage; Team: $200/month with $100 free credit, up to $2,000 usage. GPU rates: H100 80GB $6.52/hr ($0.001811/sec), A100 80GB $4.21/hr ($0.001169/sec), A100 40GB $3.46/hr ($0.000961/sec), RTX PRO 6000 $3.38/hr ($0.000939/sec), L4 24GB $0.80/hr
Key features
Choose from available GPU hardware with published hourly rates
Review cost estimates before deployment starts
Start from ready-made template applications
Monitor workload lifecycle states, logs and outputs
Connect your own GPU hardware as a host
Escrow-based payment system for secure transactions
Support for inference model requests
GPU routing across seven US regions with fallback card support
Inference API for deploying policies with one command
Automatic simulation benchmarking on LIBERO, RoboCasa and RoboTwin
Teleoperation from console with no cold start on handoff
Access to vision-language and decision models on same API key
Pay-per-second billing for GPU, CPU and memory
Scale to zero or maintain minimum workers per deployment
Library of policy templates including DROID, LIBERO and GR00T
Teleoperation integration for remote control
Automatic regression detection before deployment
What makes it different
Escrow payment ensures work is delivered before host is paid
Transparent pricing: all rates published upfront with no surprises
Minimal transaction fees: hosts keep 90% of rental earnings
Automatic refunds for failed rentals on platform side
One API key for policies, benchmarks and all vision-language models
Automatic simulation benchmarking on actual target environments
No GPU provisioning or reservation required
Billed only for actual compute used, by the second