io.net
Decentralized physical infrastructure network for clustered AI compute
Key network specifications
Network architecture & ecosystem
io.net is a state-of-the-art DePIN network that aggregates underutilized GPUs from independent data centers, crypto mining facilities, and consumer nodes into unified, low-latency clusters. Built on the Ray distributed computing framework, io.net allows AI engineers to deploy massive GPU clusters for model training, fine-tuning, and scalable batch inference at up to 70–80% lower cost than traditional hyperscalers.
Key strengths
- •IO Cloud clusters disparate geographic GPUs using the Ray framework with automated orchestration
- •Aggregates large volumes of enterprise NVIDIA H100s, A100s, and high-spec consumer RTX 4090s
- •Instant self-service cluster spin-up without sales calls, quotas, or annual enterprise lock-in
- •Transparent on-chain Proof-of-Compute-Capacity verifying real hardware performance
Limitations & tradeoffs
- •Cross-node public WAN communication incurs higher latency than co-located 3200 Gbps InfiniBand data center pods
- •Independent node stability requires cluster-level fault tolerance for extended training runs
- •Managed software ecosystem is actively maturing compared to decade-old cloud ecosystems like AWS SageMaker
Target workloads
io.net GPU pricing table
Sourced from active cloud instances and order books
| GPU model | VRAM | Price / hour | Status | Source note | Audit date |
|---|---|---|---|---|---|
| NVIDIA H100 (80GB HBM3) | 80 GB HBM3 | $1.98 / hr | Last checked: 2026-09-30 | IO Cloud catalog hourly rate for clustered H100 80GB nodes | Last checked: 2026-09-30 |
| NVIDIA A100 (80GB SXM4) | 80 GB HBM2e | $1.20 / hr | Last checked: 2026-09-30 | IO Cloud on-demand benchmark for NVIDIA A100 80GB PCIe/SXM | Last checked: 2026-09-30 |
| NVIDIA GeForce RTX 4090 (24GB) | 24 GB GDDR6X | $0.42 / hr | Last checked: 2026-09-30 | IO Cloud catalog hourly rate for dedicated GeForce RTX 4090 24GB nodes | Last checked: 2026-09-30 |
Head-to-head comparisons
Disclosure: Some links on this page may be affiliate referral links. When you lease compute through these links, DecentralGPU may earn a small referral commission at no additional cost to you. See our affiliate disclosure and pricing methodology.