AI that pays for itself.
AI requests on our network also earn network rewards from the same computation. GPU owners earn more, you pay less, and your data stays yours.
- API
- OpenAI-compatible, open models
- Network
- Pearl mainnet, 1% pool fee
- Privacy
- No prompts stored, no training
Why it works
One computation. Two kinds of value.
Our GPUs answer your request, and the same maths earns PRL, the Pearl network’s coin. That second income is shared three ways.
You pay less
GPUs earn more
We earn on both
The network gets real work
Products
One network, four ways in.
-
GPU Clouds
Same GPUs, two revenues: your customers’ payments and network rewards.
See how -
Developers
Open models behind an OpenAI-compatible API, priced per token.
Read the API -
Private AI
A chat that stores no prompts or answers. No training.
Start chatting -
Vault
A hardware-grade vault for your crypto. Three parts, any two sign.
Meet Vault (opens in a new tab)
The opportunity
Millions of AI GPUs could earn this way.
About 320,000 GPUs mine Pearl today. The world runs about 10 million H100-class GPUs for AI.
- Each H100 or H200 already runs the maths Pearl rewards.
- We bring the software, the pool and our API.
Demo run
Running on Pearl mainnet.
- First GPU ready in
- 101 s
- Rehearsal, 3 H100s, cached weights
- Requests served
- 42,297
- answered in 30 minutes
- Tokens
- 53.3M
- Shares accepted
- 713
- 4 rejected
Requests, tokens and shares: demo run, 5 Oct 2026, 5 H100s, 30 minutes, Pearl mainnet. Ready time: rehearsal the same day, 3 H100s with cached weights.
See the pool liveHow it works
From request to reward.
-
1
You send a request
To our API or chat.
-
2
A GPU answers it
In our network, on Pearl’s mining-enabled vLLM.
-
3
The work also mines
Shares go to our pool; your prompt and the answer are never stored.
-
4
Everyone earns
The pool splits the PRL by work; we keep a 1% fee.