Most dTAO risk talk is about price: alpha goes up, alpha goes down. But Bittensor carries a sharper, more binary risk that rarely makes it into an APY screenshot: a subnet can be deregistered (replaced entirely) and its slot handed to a newcomer.
Why slots are scarce
Bittensor caps the number of active subnets. That scarcity is a feature: it forces competition for the right to exist, the same way emissions force competition for the right to earn. But it means registration is a zero-sum event: when a new subnet registers and no free slot exists, an existing one must go.
How immunity works
A newly registered subnet gets an immunity period: a protected window during which it cannot be deregistered, whatever its performance. This is the network's grace period for building (bootstrap miners, ship something, attract stake).
Immunity is also why "new subnet" strategies exist at all (ours included, on /invest): during the window, you get pure upside exposure to a launch without the replacement tail risk. The flip side arrives on schedule: the day immunity ends, the subnet joins the eviction pool and is judged like everyone else.
How the axe falls
When a registration comes in and no slot is free, the protocol removes the weakest non-immune subnet, in essence the one the market values least, judged on its sustained (moving-average) alpha price rather than a single day's print. Consequences worth internalizing:
- It's automatic. No committee, no warning shot. The trigger is someone else's decision to register.
- "Cheap" can mean "condemned." A rock-bottom alpha price on a post-immunity subnet is not automatically a bargain. It may be the queue for the exit. Low price + outflows + no shipping team is the profile to avoid, however seductive the APY.
- The floor for that trade is unfriendly. Deregistration mechanics have evolved across runtime upgrades, but the working assumption should be brutal: stake parked in a deregistered subnet realizes whatever the unwinding gives it, which for a dying alpha is little.
Keeping the tail out of your portfolio
- Know each position's immunity status. Immune, safely mid-pack, or hovering near the bottom of the non-immune table? That third state deserves an exit plan, not more averaging down.
- Watch sustained price, not spot. The eviction metric smooths over days, so a subnet living near the bottom of that ranking is at risk even after a bounce.
- Diversify the binary away. A deregistration inside a 20-subnet basket is a bad line item; inside a 2-subnet portfolio it's a disaster. Basket strategies exist precisely to make tail events survivable.
- Yield-chasing screens must include this. Any subnet whose APY looks free is usually charging you somewhere (we covered where), and the deregistration queue is the most expensive "somewhere" there is.
Scarcity of slots is what keeps Bittensor's subnet quality honest. It is also a real, dated, binary risk sitting inside otherwise smooth-looking positions. Price it, or hold a portfolio where it can't hurt you.