Robinhood Chain's record week was already three days gone when it was announced
The chain's fees peaked on 4 September and were falling when the scorecards went out, and almost all of them came from one launchpad buying back its own token.
Robinhood Chain's best week was already over by the time anyone wrote it up.
DefiLlama posted the two-month scorecard on 8 September. Total value locked of $900 million, a stablecoin float over a billion, weekly app revenue of $28.5 million. For a chain whose first dollar of TVL landed on 2 July, that is a fast start by any measure anybody uses, and the crypto press ran it as one.
So we pulled the same daily fee series off DefiLlama's own API (it is public, which is the genuinely good thing about this corner of finance), because a weekly total tells you the size of a thing and never its direction.
The line peaks on 4 September and then falls for three straight days, which is the part the scorecard could not show.
The victory lap was running three days behind the chain
Fees topped out at $24.5 million. By 7 September the chain was doing $15.5 million a day, down about 37 percent from the peak, and both DefiLlama's post and The Block's write-up went out after that had already happened.
Nobody hid anything here. Weekly numbers trail by construction, and a victory lap for last week is an ordinary thing to run. But if you read either post as a description of the present, you read it wrong.
The users did not arrive. The extraction did
The Block's account is worth reading closely, because it contains the mechanism rather than the total. Fees over the prior seven days came to roughly $25 million, against $1.4 million the week before. Daily active accounts averaged 396,000 over that same week, and were lower than the week before it.
Fewer people, then. Much more money.
Fees per active account went from $0.13 in mid-August to $15.90 by early September, which is a 122-fold rise in what the chain takes from each user. We did that division ourselves; the number is not in anyone's post, and it is the only figure here that really describes what changed.
That is what a launchpad cycle looks like. It is not what a chain finding its product looks like.
One protocol produced the whole number
Pons is the chain's leading token launchpad. On 3 September it generated nearly $6 million of fees by itself, more than Pump.fun or Hyperliquid managed on the same day, and its token touched an all-time-high valuation above $970 million two days after that.
The tokenomics explain the rest. Roughly 80 percent of the revenue Pons generates goes into buying its own token back, and more than a quarter of the supply has been burnt so far. Fees buy the token. The token going up attracts launches. Launches produce fees.
Do you see the fragile part? The Block's own analysts named it, which we appreciated in a piece that was otherwise a good week's news: fees are "merely a lagging function of how many new pools exist to trade against."
Two trackers, two answers, one chain
A small housekeeping note, because both figures are now circulating as though they were the same measurement. DefiLlama puts weekly DEX volume at $10.4 billion. The Block puts it at $12.4 billion, over a slightly different window with different rules about which protocols count (nobody agrees on what a DEX is either).
Neither is wrong. Neither is comparable to the other, either.
What we would watch instead of TVL
Our expectation is that chain-wide daily fees average under $10 million by the middle of October. That is not a dramatic call, given the series was at $8.7 million as recently as late August, and it is really a bet that this launchpad cycle behaves like every launchpad cycle before it. What would prove us wrong is a rising count of new token launches through September, rather than rising fees on the pools that already exist.
TVL will not tell you either way. It is stickier, it rose about a quarter in the same week, and it does that for reasons that have little to do with launchpad volume.
Robinhood built a chain. A launchpad built the numbers. Those are not the same company.
