Become a provider
Connect anything from a single RTX 3090 to a rack of H200s. The MZGrid agent benchmarks your hardware, runs customer workloads in isolated containers and gets you paid in MZG every time a job settles.
Providers on the network
64
Network utilisation
31.1%
GPU-hours · 24h
1,913
MZG locked by providers
860.7K
Earnings calculator
Pick a GPU, how long it’s online and how busy you expect it to be. The breakdown follows the on-chain settlement exactly: 7% protocol fee, then 80% liquid and 20% locked.
Price range shown is 0.85–1.20× the catalogue reference, where live offers typically sit. Current network median for this GPU: 150.50 MZG/h.
Estimated per 30 days
Gross MZG charged to customers
58,890 MZG
390.0 billed GPU-hours × 151 MZG
Job cap = 25 MZG + 4 × locked balance, enforced by ProviderRegistry.
Illustrative estimate only — not a forecast or promise of earnings. Actual demand, utilisation and prices vary, all amounts are in MZG and their value changes with the market. Electricity, hardware and gas costs are not included.
Why MZGrid
Customers pre-fund escrow before you start. You are paid for every second of verified usage the moment a job settles.
Start with zero capital at risk. Your collateral builds automatically from 20% of your own earnings.
Completion rate, uptime, benchmark consistency and tenure feed a 0–100 score that wins you more work.
Customers deploy from anywhere; the scheduler weighs proximity so regional capacity is valuable.
Community, Pro and Datacenter tiers let customers find the right trade-off — and let you price accordingly.
Leave whenever you want. Your lock unlocks exactly 10 days after you request exit, once jobs are done.
Requirements
Install
# 1. Install the provider CLI
npm install -g @mzgrid/provider-agent
# 2. Create or import your key, pick a region and prices
mzgrid-provider init
# 3. Detect GPUs and run the FP16 benchmark
mzgrid-provider detect
mzgrid-provider benchmark
# 4. Register on-chain (ProviderRegistry) and with the coordinator
mzgrid-provider register
# 5. Start serving jobs
mzgrid-provider startCheck your lock, earnings, reliability and running jobs any time with mzgrid-provider status, or open the provider dashboard. The full guide, including a systemd unit and the container isolation model, is in the installation docs.
Lock & exit
Every settlement adds 20% of your share to your lock. The lock raises your job cap, signals reliability to customers, and comes back to you in full 10 days after exit — unless it is slashed for misbehaviour.
Day 0
requestExit()
Status becomes Exiting. No new jobs are scheduled; running jobs must finish.
Days 0–10
Challenge window
Locked MZG can still be slashed for abandoned, offline or fraudulent jobs.
Day 10+
withdrawLocked()
With zero active jobs, the full remaining lock is transferred back to the provider.
Max deposit per job = 25 MZG + 4 × locked
| Locked MZG | Max job deposit | ≈ H100 SXM hours* |
|---|---|---|
| 0 | 25 MZG | 0.03 h |
| 250 | 1,025 MZG | 1.4 h |
| 1,000 | 4,025 MZG | 5.4 h |
| 5,000 | 20,025 MZG | 26.7 h |
| 25,000 | 100,025 MZG | 133.4 h |
| 100,000 | 400,025 MZG | 533.4 h |
*At the 750 MZG/h catalogue reference price.
Slashed MZG: 50% to the affected customer, 50% to the treasury.
Abandoned job
Accepted but not completed by its deadline
min(locked, 2 × deposit)Offline during job
No usage samples for more than 10 minutes while running
min(locked, 1 × deposit)Verification failed
Canary output mismatch or falsified usage reports
100% of lockedReputation
Your 0–100 reputation score is recomputed continuously and drives 35% of the scheduler’s ranking.
Pinecrestfield Systems
sg-central · 1,153 GPU-h
Quillongate Systems
us-central · 1,032 GPU-h
quiet-otter-compute
us-west · 852 GPU-h
Dovetailpoint Compute
us-west · 713 GPU-h
Kelpiebay Works
us-central · 575 GPU-h
FAQ
Can’t find an answer? Ask in the #providers channel on Discord.