Chutes (netuid 64, built by Rayon Labs) does something unglamorous: it lets anyone deploy a model behind an HTTP endpoint and pay per call, with miners supplying the compute and validators checking that the responses came from the model that was claimed. No new architecture, no research breakthrough. Just hosting.

That is exactly why it matters. Most Bittensor subnets are markets for a capability that has to be scored subjectively — is this a good summary, is this a plausible forecast, is this image on-style. Chutes is a market for something metered: requests served, latency, uptime. When output is countable, incentive gaming is harder and emissions get more stable. In our sample set, SN64 carries 5.2% of network emissions, the largest single share, across 281 validators.

The number that tells the story: 18.1% APY

Chutes has the lowest delegation APY of the 36 subnets we track — 18.1%, against a set average near 24%. Newcomers read that as a weakness. It is the opposite. Delegation yield on a subnet is a function of emissions divided by stake. When a subnet is widely believed in, stake piles in, and the yield compresses. A low APY on a high-emission subnet is the market saying "this is the safe one."

Compare it with SN77 Liquidity at 29.4%. That is an 11-point yield premium. You are collecting it because SN77 earns from market-making spreads on alpha tokens, so a drawdown in alpha prices hits the token you hold and the revenue that funds your yield at the same time. The yields are not comparable instruments.

What the flows show

SN64 turned over $6.18M in 24 hours against a $96.4M market cap — a 6.4% daily turnover ratio, the healthiest in the set. For a delegator this is the least appreciated metric on the page: it is the difference between being able to rebalance a five-figure position in one clip and having to leg out over a week. SN89 Vericore, by contrast, turns over $240K against a $4.4M cap. Same asset class, entirely different exit.

Price action has been unspectacular and that fits: +1.9% on the day, +5.4% on the week, tracking TAO more than any subnet narrative. If you want SN64 to double, you are really asking for TAO to re-rate.

The bear case we take seriously

Two things could break the thesis. First, emissions are set by consensus weight, not contract. A 5.2% share today can be reweighted, and the largest subnets have the most to lose from governance shifts because they are the obvious source of budget for newer ones.

Second, serverless hosting is the most competitive thing in AI infrastructure. Chutes is not competing with SN19 Nineteen for Bittensor mindshare; it is competing with the price per million tokens at centralised providers who are cutting it every quarter. The subnet wins if decentralised supply is cheaper or more censorship-resistant. It loses slowly and quietly if it is neither.

Verdict

Chutes is a core holding, not a trade. If you are building a delegation book, it is a defensible 25–40% anchor: the deepest liquidity in the set, the largest emission share, a validator count that makes single-operator failure unlikely, and a yield you should expect to stay boring. Pair it with one higher-APY satellite — SN56 Gradients or SN37 Taoverse are the reasonable candidates — rather than trying to make SN64 do a job it is not built for.

What would change our mind: emission share dropping below 4%, or daily turnover falling under 3% of cap for a sustained stretch. Both would mean the "safe one" premium is eroding.

TAO Compound is an analytics tool, not financial advice. Subnet figures referenced above come from the sample data set in this build; the live TAO price shown elsewhere on the site is from CoinGecko.