Every Bittensor subnet is a marketplace for a specific kind of machine intelligence: inference, training, data, trading signals, robotics, storage. Miners compete to produce the work; validators score it; emissions reward both. Since dTAO, each subnet also has its own token and its own price. So evaluating a subnet is evaluating a small economy. Here is what to actually look at.
1. What does it produce, and does anyone want it?
The first filter is brutally simple: does the subnet produce something a real customer would pay for? Some subnets ship products with revenue; others are research bets; some are little more than a whitepaper. A subnet with real usage (API customers, an app, enterprise deals) has a demand story beyond speculation. A few even use revenue to buy back and burn their alpha, directly linking usage to token value.
2. Emissions share and net flows
Under dTAO, emissions follow staking flows. Two readings matter:
- Emission share: what fraction of network emissions the subnet currently earns. Zero or falling share means the market is voting against it with its TAO.
- Net flows (7/30 days): is TAO entering or leaving the subnet's pool? Sustained outflows pressure both the alpha price and future emissions. Momentum is real in an economy where capital allocates resources.
3. Pool depth (your exit door)
The alpha/TAO pool is where you enter and exit. A subnet whose pool holds very little TAO can show a beautiful price, one you cannot realize, because selling any meaningful size moves the market against you. Before sizing a position, ask: how much TAO of depth is on the other side?
4. Ownership concentration
If a handful of wallets hold most of a subnet's alpha, they can swing its price at will. Concentration measures (like a Gini coefficient, or how many entities control half the supply) tell you how much you are exposed to someone else's exit. Broad, organic ownership is the healthier signal.
5. Immunity and deregistration risk
Subnet slots are scarce. A new subnet enjoys an immunity period; after that, the weakest subnets (lowest sustained price among non-immune slots) can be deregistered when a new one registers. Staking into a struggling, post-immunity subnet carries a real tail risk that no APY figure reflects. Check where a subnet stands in that queue before chasing its yield.
6. The team is shipping (or not)
On-chain metrics lag reality. A subnet with an active repository, regular releases and a responsive team can recover from a rough quarter; an abandoned one cannot. Development activity is the leading indicator; flows are the confirming one.
Putting it together
No single metric decides. The strongest setups combine a real product, positive flows, decent pool depth, distributed ownership and an active team. The weakest combine a high advertised APY with outflows, a thin pool and silence from the builders. The yield is the bait, the illiquidity is the trap.
Reading all of this across 128+ subnets is work. That's the point of TaoScale: every subnet page shows the live data, and our curated strategies (by sector, market cap, revenue, or team picks) encode this analysis so a single tap spreads your TAO across subnets that pass the filters. Non-custodial, always.