Consumer fintechs are turning themselves into settlement layers. The interesting question is not how fast their chains run, but who gets to keep the value

 

On September 4th a blockchain that was just over two months old processed 14.28m transactions and collected ~$5.44m in revenue in a single day, the applications built on top collected ~$5.6m, much more than the applications on both Ethereum and Base. It was not built by cryptographers. It was built by a retail brokerage best known for zero-commission trading and a fondness for confetti animations.

Robinhood Chain went live on public mainnet on July 1st 2026, five months after a testnet that cleared more than 200m transactions. It is an Arbitrum Orbit deployment running the Nitro stack, settling to Ethereum, using blobs for data availability and ETH for gas. Blocks arrive every 100 milliseconds. There is no network token, and one sequencer, operated by Robinhood, which orders every transaction.

The growth has been unusual even by the standards of a market that discounts novelty quickly. Daily decentralised-exchange volume reached $1.72bn on September 4th, with cumulative turnover past $50bn inside seven weeks. Deposits in the chain’s DeFi applications climbed from under $5m in late June to about $830m, with a further $2.5bn bridged. Stablecoin supply approached $1bn, and Ethena’s USDe overtook Robinhood’s own Paxos-issued USDG as the largest of them. Since late July the chain has been recording more daily active users than Base chain. 

Two qualifications matter more than any of those figures. The first is composition. Robinhood built the network for tokenised equities; what arrived was a launchpad economy. More than 80% of the first $9bn of volume came from memecoin speculation, and by mid-August tokenised real-world assets had fallen to about 6% of chain value, from roughly a third in early July. On the network’s busiest day, a launchpad, a trading bot and Uniswap produced almost 90% of application revenue. The second qualification is that every metric so far has been recorded while Robinhood pays the gas for qualifying swaps made through its wallet. That subsidy expires on September 29th. October will be the first honest month without the subsidy. 

 

The licensable chain

 

Robinhood is not an outlier but a template. Kraken has Ink, Sony has Soneium, Uniswap has Unichain, Stripe has Tempo, and Coinbase has Base, which by some measures became the only rollup to turn a profit in 2025, earning around $55m after data costs and revenue sharing. Coinbase launched its own tokenised American equities natively on Base on August 24th for non-American users, on a bespoke token standard, with roughly 50 DeFi protocols lined up to support them.

What is being deployed here is not scaling technology. It is market structure. A consumer broker with an order-routing business, a wallet, a stablecoin relationship and a few tens of millions of verified customers can now own the execution venue, the settlement layer and the custody interface at once, while leaving the venue itself formally permissionless. Robinhood ended June with 28.4m funded customers and $369bn of platform assets. Its group revenue hit a record $1.31bn in the second quarter even as its crypto line fell 38%. The chain is the answer to that second number.

The rollup is no longer a scaling product. It is a distribution product wearing a scaling product’s clothes.

 

Where the money stops

 

Follow the fee and the stack looks very different from the way it is usually drawn.

Start at the bottom. Ethereum sells data availability, and for two years it sold it at close to nothing. In the first year after Dencun, blob transactions paid something on the order of $8m in total, against the roughly $34m a month rollups had been spending on calldata beforehand. Fusaka, live since December 3rd 2025, was as much a commercial repair as a technical one: EIP-7918 floors the blob base fee at a fraction of the L1 execution base fee, so data can no longer be sold at 1 wei while the base layer is busy. Fidelity’s research estimated the mechanism would have generated roughly 24,600 additional ETH between Dencun and October 2025. Even so, one recent estimate put the base layer’s capture at 4.9% of the economic value generated across its application layer in the second quarter. 

Move up to the stack providers. Arbitrum’s Expansion Programme charges chains that settle outside Arbitrum One 10% of protocol net revenue, split 8% to the DAO treasury and 2% to the Developer Guild. This is the cleanest institutional revenue share in the sector, and ARB rallied by double digits when it was confirmed. On DefiLlama’s measure, Robinhood Chain’s net chain revenue over the 30 days to August 31st was about $4.04m, however the last week that has followed saw an explosion in revenue, recording ~23 million in just the last 7 days. The Arbitrum DAO’s share of that is a 10% cut, this easily marks the single most successful chain launch of the cycle. The market repriced ARB by considerably more than the discounted value of that annuity as a result.

Optimism’s version of the same trade has already been tested. Base paid the greater of 2.5% of revenue or 15% of onchain profit, contributed the overwhelming majority of Collective sequencer fees, and in February announced it would move to its own codebase. OP fell 28% in two days. The OP Stack is MIT-licensed, which is the point: a fee share that survives only while the licensee finds it convenient is not an annuity, it is a courtesy. Arbitrum’s licence has contractual teeth that Optimism’s arrangement lacked, but the incentive it creates for a successful licensee is identical. 

Then the operators. Coinbase keeps Base’s sequencer margin as corporate revenue and has no token to dilute it. Robinhood keeps gas fees and gives up a tenth. Neither, however, captures the application layer sitting on top: the launchpads, bots and DEXs on Robinhood Chain earned in a day roughly two-thirds of what the chain itself earned in a month. The sequencer is a toll booth on a road whose shops keep their own takings.

 

Value at the front door

 

The implication is that value is migrating to the top of the stack, where the customer is, and that it is being captured in two currencies rather than one. The first is fee revenue, which is real, cyclical and currently dominated by speculation. The second is the regulatory perimeter, which is where consumer brokers hold an advantage no infrastructure provider can license.

Consider how the Robinhood product is actually built. Stock Tokens are debt securities issued by Robinhood Assets (Jersey) Limited, tracking a listed share without conferring ownership or votes, priced by Chainlink oracles, held in self-custody, usable as collateral across DeFi. They are unavailable to Americans. The SEC staff statement of January 28th 2026 confirmed that the wrapper does not change the instrument, and drew the distinction that matters here: an issuer-sponsored entitlement is one thing, a linked debt instrument that provides price exposure only is another. Robinhood chose the second. Enforcement runs through the issuer and the wallet interface; the chain underneath stays open, and anyone can list anything on it.

That is the model in miniature. Listing standards migrate from the venue to the interface. The custody perimeter migrates from the broker’s omnibus account to a wallet whose screening happens at the edges. The result is a permissionless network whose flagship product is geofenced, curated and legally conventional, wrapped around an unpermissioned long tail that generates most of the fees. Both wolves, as Robinhood’s crypto chief puts it, get fed.

 

The Apollo Crypto View

 

For anyone underwriting this stack, the variable has changed. Total value locked was always a poor proxy; fee generation, and specifically the durability and mix of it, is the number that matters.

Three tests follow. The first is the subsidy cliff on September 29th: no data yet exists for this network in a world where users pay their own gas. The second is mix. Stock Token volume as a share of total DEX turnover is the cleanest available measure of whether a brokerage chain is becoming financial infrastructure or a better-branded casino. The third is the enforceability of fee share. Base’s exit is the precedent every licence-fee valuation now has to clear, and the Arbitrum arrangement should be underwritten as a royalty on someone else’s business, not as equity in it.

Our reading is that distribution wins, and that it wins twice over: once through fees, and again through a regulatory perimeter that neither the base layer nor the stack provider can replicate. That argues for owning exposure where fee capture is contractual and the customer relationship is proprietary, which today means listed equity more often than tokens. It argues for treating stack-provider tokens as call options on licensee success with weak enforcement, priced accordingly. And it argues for watching Ethereum’s newly floored blob economics closely, because a settlement layer that has finally started charging for its scarcest product is a more interesting asset than the one that gave it away.