Subnet MetricsIssue 307 September 2026

The week on chain

The week showed a wide performance split: RedTeam led at +43.7%, while MVTRX fell -36.2%. The single most interesting result was verified burning: 6 subnets burned 520 TAO and 41,291 alpha, led by Gradients with 269 TAO. This matters because 13 other subnets claimed buybacks but recorded no burn, highlighting the gap between stated activity and direct chain evidence. If verified burns continued at this level, token removal would remain concentrated among few subnets, while allocators would still need to distinguish on-chain execution from unverified claims.

Panel depth
569
Subnets
128
Burned this week
520τ
Top-5 share
32.3%
Median root prop
0.156
Finding 01 · Verified burns

Verified alpha burns

Every burn_alpha and recycle_alpha extrinsic on chain this week, read exactly: 41,291 α ≈ 520.2τ (~$135,964) across 6 subnets. No attribution, no estimate.

SubnetNameα burnedτ valueUSDBurnsWallets
SN56Gradients16,397269.1τ$70,32711
SN49Nepher Robotics17,233135.6τ$35,43521
SN8Vanta2,07357.4τ$14,9971691
SN11TrajectoryRL5,30048.7τ$12,73110615
SN64Chutes1017.4τ$1,9383322
SN62Ridges1862.1τ$5368520
Alpha valued at the panel's latest price for that subnet. A burn call that asks for more than the wallet holds is capped by the chain and excluded here rather than guessed.

13 subnets that publicly claim revenue-funded alpha buys burned nothing this week: SN4 Targon, SN19 Blockmachine, SN21 AdTAO, SN33 ReadyAI, SN34 BitMind, SN41 Almanac, SN50 Synth, SN51 Lium, SN53 Engy, SN54 Yanez, SN93 Bitcast, SN110 Green Compute, SN118 Ditto. A claim without a burn is not false: alpha bought and held in a treasury never shows here. It is unverified, and the burn is the one form of buyback the chain proves.

Alpha burned per day, valued in τ
Every successful, uncapped burn_alpha and recycle_alpha extrinsic, by UTC day, over the last 90 panel days. Each subnet's alpha is valued at the panel's latest price for that subnet, so the height is comparable across days at today's prices, not the price on the day.
Finding 02 · Concentration

Where subnet value sits

The five largest subnets hold 32.3% of subnet market cap and the ten largest 46.8%. Over the last 90 days the top-5 share moved +4.7% (from 27.6% on 2026-06-10) and the top-10 share +5.2%.

Top 5 shareTop 10 share
Share of total subnet market cap (root excluded) held by the five and ten largest subnets, one reading per panel day.
Finding 03 · Root

How much of validation still comes from root

The median subnet's root proportion is 0.156, -0.016 against 0.172 on 2026-06-10. Root proportion is the share of a subnet's stake weight that comes from root (netuid 0) rather than its own alpha; as it falls, the subnet's own token carries more of its validation.

Median root proportion
The median across all subnets of the root share of stake weight, one reading per panel day.
Finding 04 · Movers

Biggest movers

Market-cap change over the week, excluding subnets under 2,000τ where a large percentage move carries no information.

Subnet7dMarket cap (τ)
SN61+43.7%70,319τ
SN91+38.4%11,539τ
SN103+34.1%3,835τ
SN28+21.4%107,082τ
SN117+20.0%8,787τ
SN79-36.2%19,056τ
SN92-30.1%2,461τ
SN84-21.3%2,918τ
SN90-14.0%9,450τ
SN50-11.3%27,650τ

How these numbers are made

One snapshot per day, unbroken since dTAO opened, plus a direct chain read of owner wallets for the buyback figures. Nothing here is estimated, modelled or sourced from a disclosure. It is what the chain showed. Where a figure has a known limit, the limit is stated next to it rather than in a footnote.