dTAO at 18 months

Seventy percent of the original top twenty is gone

Rank a subnet by market cap on the day dTAO opened, take the top twenty, and follow that exact cohort forward. It does not decay gently. Half of it was gone inside three months.

The steep part is the first quarter — twenty down to ten by mid-May 2025 — after which the survivors are unusually durable. That shape matters more than the endpoint: it says early dominance under dTAO carried almost no information, but dominance that survived the first quarter carried a great deal. A ranking taken at launch was close to noise. The same ranking taken ninety days later has held for over a year.

Value spread out, then stopped spreading

dTAO was meant to decentralise value away from a handful of incumbent subnets. It did — for nine months. Then it stopped, and it has not resumed.

The top five held 51.6% of non-root market cap at launch and 30.4% by November 2025. Since then the line is flat: today it is 32.1%, marginally more concentrated than the trough. The subnet count doubled from 64 to 128 across the same window, so the flat line is not a shortage of new entrants — it is new entrants failing to take share from the leaders.

The root-stake drain, in one line

The median subnet’s root proportion fell from 0.994 to 0.159. This is the clearest single measure of what dTAO actually did.

At launch, stake was almost entirely root-held. Eighteen months later the median subnet holds about a sixth of that. The decay is smooth and still going — there is no floor visible in the data yet. That trajectory is the context for the v450 root-basket change now redirecting root emission into subnet alpha: it lands on a network that has already moved most of the way, not one being pushed at the start of the journey.

Two subnets in thirty are buying their own alpha

The prevailing story is that revenue-funded alpha buybacks are becoming standard practice. We scanned the top thirty subnets by market cap against the chain for the week to 31 August. Two showed any buyback flow at all.

The method is deliberately narrow. For each subnet we take the owner coldkey straight off the metagraph — not a wallet someone told us about — pull that wallet’s trades, and keep only the ones touching that subnet’s own alpha. Net alpha acquired is the buyback. No disclosures, no interviews, no estimates: if a team is buying, the chain says so, and if it is not, no amount of narrative fills the gap.

Vanta alone is 97% of the tracked flow. Read that carefully before treating buybacks as an ecosystem trend: on this week’s evidence it is one team’s policy, not a norm. The honest caveat runs the other way too — a team buying through a wallet we cannot tie to the subnet is invisible here, so this is a floor on buyback activity, never a ceiling.

Get the next one

One issue a week, built from the same panel. No tracking pixels, no third party holding the list, one-click unsubscribe on every email.