What validators do
Each epoch, a validator queries the miners registered on its subnet and scores the responses against whatever the subnet measures — answer quality on a text subnet, data correctness on a scraping subnet, latency and accuracy on an inference subnet. It then submits a weight vector to the chain: one number per miner, saying how highly it rates them. Yuma Consensus combines every validator’s vector into the ranking that sets emissions.
How validators earn
Two streams. The first is emissions on your own staked TAO — the same yield any staker gets, proportional to your position in the subnet pool. The second, and usually the larger, is the take rate: the cut you keep of the emissions generated by TAO other people have delegated to your hotkey.
You set your own take, under a chain-wide ceiling. MaxDelegateTake is 11796/65535 — 18% — and has been capped there since runtime v320; a validator may set less, and many do, to compete for delegation. Because it is a per- validator setting that can change, it is a number to read rather than assume.
See what validators are currently charging →
That shape is what makes validating a service business rather than a capital deployment: beyond your own stake, earnings scale with the size and loyalty of your delegation base, and that scales with reputation.
Validators vs. miners
A miner’s binding constraint is GPU throughput. A validator’s is uptime: miss enough consecutive epochs of weight submissions and you can lose your slot on the subnet, which costs you re-registration and, worse, the delegation base you had built. Validators also frequently run scoring models of their own — judge models, evaluation pipelines, reference datasets — so the compute bill is not zero either.
The entry barrier is also higher. Validating requires meaningful stake before a single emission arrives, where mining requires only a machine and a burned registration fee. In exchange the business scales better: once your reputation is established, attracting delegation grows revenue without growing operating cost proportionally.