Register your GPU as a DCS worker — H100, H200, A100, L40S, RTX 5090, RTX 4090. Bid on jobs in the reverse-auction lane. Earn USDC on Base mainnet, settled per-job. 97.5% to you, 2.5% to network. Your trust score builds with every successful run.
pip install dcs-worker (NOT PUBLISHED — no such package on the registry, checked 30 Aug 2026) then dcs-worker login
One-time Base mainnet mint (gas covered by us). Soulbound to your wallet — non-transferable identity.
Minimum hourly rate you'll accept. Below this, you don't bid. Above, the auction picks the lowest bid.
Worker accepts winning bids automatically. Container runs in isolation. Output uploaded + signed.
Settlement on Base mainnet within 60s of job completion. USDC to your wallet. Withdraw any time.
Clean history → Probation → Verified → Trusted tier. Higher tier = better job preference + lower commission.
Rates fluctuate with demand. Floor prices below are typical lows — actual bids run higher.
$18-24/hr avg
For: fine-tuning, frontier inference, large-batch render. Highest demand.
$15-22/hr avg
For: same workloads as H200, slightly lower memory. Always busy.
$10-14/hr avg
For: medium-batch training, vision workloads, inference at scale.
$3-5/hr avg
For: video render, image generation, smaller inference workloads.
$1.20-2.40/hr avg
Consumer cards. Best for: SDXL generation, small inference, render farms.
$0.40-0.90/hr avg
Best-effort tier. Hobbyist + edge use cases. Lower job priority.
1× RTX 4090, 16h/day idle, Verified tier.
~$580/mo
after 2.5% commission, before electricity.
1× H100, 24/7, Trusted tier.
~$11,000/mo
after commission. Roughly 1× H100 cost amortization in 2-3 months.
8× H200 cluster, 24/7, Trusted tier.
~$95,000/mo
at sustained 80% utilization. Operator runs as a business.
Numbers assume current demand. Past performance ≠ future. Electricity cost not included. Some markets (cheap power) materially better return profile.
Receipt-signed. Settled in USDC. Trust score builds automatically.