Bittensor after eTAO: three scenarios for a decentralised AI network still living on subsidies
Bittensor now lets subnet owners set their own reward splits. The bigger test is whether subnets can earn real revenue before emissions shrink again.
AI & CRYPTOALTCOINSANALYSIS
Hana Kimura
10/28/20263 min read
Bittensor has spent 2026 rebuilding the economics of its decentralised AI marketplace one patch at a time. The latest change, eTAO, went live around October 21. It lets the owners of each subnet decide how that subnet's token rewards are divided between themselves, the validators who score work and the miners who perform it. The move looks technical, but it touches the central question hanging over the network: can its subnets sell enough AI services to pay their way once token emissions stop doing the job?
How the incentive machine works
Bittensor is a network of specialised subnets, each focused on a task such as model inference, training, data scraping or prediction. Since the launch of Dynamic TAO in February 2025, every subnet has its own alpha token, and stakers decide which subnets receive TAO emissions by moving capital into them. A model called TaoFlow, introduced in November 2025, ties those rewards more closely to net inflows. The first halving in December 2025 cut daily issuance from 7,200 to 3,600 TAO, on a fixed maximum supply of 21 million.
Until now each subnet's alpha was split on a fixed formula: 18% to the owner, 41% to validators and their stakers, and 41% to miners. Under eTAO, owners can take between 0% and 33%, validators can receive up to 50% and miners must get at least 33%, according to the announcement summarised by KuCoin News. Owners can revise the split roughly once a week. Total emissions to a subnet do not change; only the division does.
The subsidy gap
The reason this matters is money. In a March report, Pine Analytics estimated that Chutes, one of the largest subnets, received about 14.4% of all emissions, worth roughly $52 million a year at the time, while its external revenue was estimated at $1.3 million to $2.4 million annually. Pine put total demand-side revenue across the network at only $3 million to $15 million a year, and noted that most figures are self-reported and unaudited. The firm also calculated that the ten largest subnets capture about 56% of emissions and that around 70% of TAO supply is staked.
Bittensor's core developers have shown they are willing to intervene. On June 22, emissions were halted on 57 of the network's 128 subnets that had inactive code, burned miner rewards or showed signs of self-mining and gaming of TaoFlow, according to reporting by OurCryptoTalk. Operators can apply for reinstatement through weekly reviews. Together with eTAO, the clean-up suggests a shift from rewarding every subnet to rewarding subnets that do something.
Three scenarios for the next year
In the optimistic case, eTAO helps the best subnets compete. Owners of subnets with paying customers can direct more alpha to miners, attract better hardware and improve the service, which in turn wins more paying users. Inference subnets that already sell access through public APIs are the most likely candidates. If several subnets can publish verifiable revenue that grows faster than their emissions shrink, the network's valuation stops resting on narrative alone.
A middle path is consolidation. Stakers, already concentrated in the top subnets, keep moving capital to a small group with proven traction, and pruning continues at the bottom. Bittensor would then look less like an open marketplace of hundreds of experiments and more like a handful of AI businesses sharing one token. That is not necessarily bad for TAO holders, but it weakens the decentralisation story.
The pessimistic case turns on the next halving, which Pine Analytics expects in late 2026 or 2027 depending on issuance. A further cut in emissions would reduce what miners earn from subnets without customers. If prices do not rise to offset it, miners may leave and services may get worse just as competition from centralised providers intensifies. Pine argued that, without subsidies, Chutes would be 1.6 to 3.5 times more expensive than centralised rivals.
What to watch
Three indicators will show which path Bittensor is taking. The first is how owners use eTAO: a shift towards larger miner shares signals investment in quality, while owners raising their own cut would signal the opposite. The second is disclosure, since independently verifiable subnet revenue is still rare. The third is the timing of the next halving and how subnet token prices react ahead of it. Bittensor remains the largest experiment in paying for AI work with a token. Over the next year the market will find out how much real demand stands behind it.