Robinhood Chain's gas price went up 25 times in 11 days
Thirty-one wallets, all of them funded by the same one, now pay a sixth of every fee on Robinhood's blockchain. They cannot buy their way to the front of the queue, because this chain sells no such thing. What they can do, and did, is make the ride dearer for everyone.
01 — The turnA cent became half a dollar
On 22 August, somebody swapping one token for another on Robinhood Chain paid a bit over a cent for it. It is the sort of charge nobody notices. Twelve days later the same swap cost 48 cents, and the person paying it had done nothing different.
Robinhood Chain is a blockchain that Robinhood built and runs itself. It went live in the spring. Tokens standing for company shares trade on it next to ordinary crypto, so a swap there can be a tokenised NVIDIA share changing hands, and every one of those trades pays the network a fee to be carried.
Nobody changed a price. There is no price to change. The charge is set by the network, second by second, from a single number that rises when the chain is busy and falls when it is quiet. That number has a floor built into the software, and for most of the summer it either sat on the floor or came back to it within two weeks. On 24 August it left, and by 3 September it was 25 times higher than it had been and still climbing.
Robinhood Chain puts out every block, so all of this is on the record. Bitquery indexes the chain from its first block, and we pulled the lot: 54 million blocks, 597 million transactions and the fee attached to each one. We wanted three things. What the price actually did, who created the demand that moved it, and where the money goes once it is paid. The last of those has the shortest answer and the largest number in this piece.
One limit belongs at the top rather than the bottom. We can describe what the price did in complete detail and we can identify who created the demand. We cannot fully explain what joins the two. The settings the chain gives out about its own pricing describe a different machine from the one we measured, and section eight sets out that gap rather than papering over it.
02 — The premiseOne price, and no queue-jumping
A blockchain charges for work. Every transaction takes an amount of work, measured in units called gas, and the sender pays for those units. On most networks the sender pays two things. A base price everyone pays, and a tip they can raise to get served sooner. The tip is what turns a block into an auction. It is why fees spike during a busy token sale, and it is why trading bots fight over who goes first.
Robinhood Chain has the base price and nothing else.
We checked every transaction on four days spread across the chain's life, from July to September. Not one was charged a tip. Every transaction that was not a system message paid exactly the network price, to the wei. There is no express lane sold on the side either. We looked for the auction contract that Arbitrum networks use for that and found none here.
What makes this odd is that people keep trying anyway. On 2 September, three transactions in every five arrived carrying an offer to pay above the network price. The chain took none of them.
| What a transaction pays, 2 September | Transactions |
|---|---|
| Sent that day | 12,324,245 |
| Offering to pay above the network price | 7,418,452 (60%) |
| Actually charged above the network price | 0 |
| Paying a fee the network destroyed rather than kept | 0 |
So the only lever on this chain is the shared price, and everyone moves with it together. That matters for what follows. When a few dozen wallets create enough work to lift the price, nobody gets outbid. The bill simply goes up for every other person on the network, and nobody can opt out or pay their way around it.
Robinhood Chain runs on the Arbitrum Orbit stack and settles to Ethereum, which means it does its own computing and then posts a short record of the results back to Ethereum to be kept. It uses ether for gas. It makes a block roughly every tenth of a second, fast enough that it makes more blocks in a day than Ethereum does in two weeks.
Nobody on this chain can pay to go first. The only price is the one everybody pays together.
03 — The floorIt had come back twice before
The chain will not price gas below 0.02 gwei, a unit so small that a plain transfer of ether at that price costs a hundredth of a cent. That floor is not a guess. Ask the chain and it reports the floor and the surcharge on top of it as two separate numbers.
The price left that floor twice over the summer and returned both times. For three weeks in July it ran at about three times the floor, peaking near six times, then fell back and sat on the minimum for a week. It lifted again in the first half of August, ran at not quite twice the floor for ten days, and again came back. Each time the chain was busier for a while, the price did its job, demand eased and the charge fell back to the floor.
The third break came on 24 August. The base fee last touched its floor on that day and has been above it every hour since.
The blue band is how much work the chain was doing. The line is what it charged. Through July and early August the two rise and fall together and the line always comes home. From 24 August it breaks away from everything before it.
04 — The breakThree times the work, eighty times the money
Demand roughly tripled between 22 August and 3 September. The price went up 25 times, and the money the chain took went up 82 times.
That is not an error. It is how this kind of pricing is meant to behave. Below a threshold the charge stays at the floor, and above it the charge climbs steeply, so a modest rise in demand produces a very large rise in price. The design exists to make a crowded chain dear, fast. It did.
Two months of fees sit in that chart and the final ten days dwarf all of them. Robinhood Chain has taken about $23 million in fees since its first block, and seven tenths of that arrived from 24 August onwards.
| 22 August against 3 September | Change |
|---|---|
| Work done, billions of gas units a day | 1,091 → 3,393 |
| Price of gas, gwei | 0.020 → 0.511 |
| Fees collected | $54,701 → $4,503,705 |
| Blocks produced | 861,465 → 842,360 |
The last row is the quiet one. The chain made fewer blocks on the busy day than on the calm one. Block production slipped by about two in every hundred as the load went up.
05 — The causeEight addresses did four fifths of it
Extra demand on a blockchain usually spreads itself thinly across thousands of users, and this did not. We took every transaction over the period and grouped the gas by the contract each one was sent to. Eight addresses account for four fifths of the whole increase, and by 3 September they were using more gas than the rest of the chain put together.
Three of them do most of the work.
The first is a swap router at 0x65050a9b…, which took 1.7 million transactions on 3 September and paid a quarter of every fee the chain collected that day. It has 35,408 senders, which sounds like a crowd until you look at how they behave. Wallets sending between 101 and 1,000 swaps in the day are one in every eight of its users and nearly six in every ten of its transactions. Sampling an hour of its trades, the pairs come back even, roughly as many swaps one way as the other, routed through a tokenised NVIDIA share and a pair of meme tokens. That is the shape of arbitrage, which we have measured before across five chains and which lives by doing the same round trip thousands of times for fractions of a cent.
| The swap router, 3 September | Value |
|---|---|
| Transactions | 1,700,887 |
| Fees paid | 448.9 ether ($1.10 million) |
| Share of everything the chain earned | 24.3% |
| Wallets sending 101 to 1,000 swaps that day | 4,263 (59% of its transactions) |
The second is a settlement contract that fills orders on a trader's behalf. Each of its transactions burns about 790,000 gas, roughly what 38 plain transfers would cost, and 31 wallets send nearly all of them. Those wallets are not rivals. One of them funded 29 of the other 30 directly, which is why we treat them as one.
| The 31 wallets, 3 September | Value |
|---|---|
| Transactions sent | 695,062 |
| Fees paid | 283.6 ether ($692,000) |
| Share of everything the chain earned | 15.4% |
| Funded by a single wallet | 29 of 31 |
The third is account abstraction, the standard that lets somebody use a wallet without holding gas themselves. Robinhood Chain carries real traffic through both ERC-4337 entry points, about 650,000 bundled wallet jobs a day, and those bundles paid $424,000 in fees on 3 September. This is a different kind of demand from the other two. Somebody is running consumer wallets on this chain at volume.
Everything else barely moved. Plain sends, token launches and the chain's own perpetual exchange used about as much gas at the end of the period as at the start. What is squeezing the chain comes down to three pieces of trading plumbing and the software feeding them. One application taking over a whole network is a pattern we have seen before, most plainly when Polymarket grew into most of Polygon.
06 — The billWhat everybody else pays
There were 382,546 wallets paying fees on the chain on 3 September and almost none of them run trading software. They pay the same price as the wallets that caused it. Sorting a crowd that size by what each address does is the same exercise we ran when we asked how much of Polymarket is bots.
The typical transaction and the average one moved apart as the price climbed, which is a clue in itself. The average is dragged up by the heavy contracts in the last section. The middle is what an ordinary user meets.
| One transaction on the chain, 22 August → 3 September | Change |
|---|---|
| The typical transaction | $0.006 → $0.20 |
| The average transaction | $0.007 → $0.41 |
| The dearest 1 in 100 transactions | $0.05 → $3.46 |
| Gas used by the average transaction | 147,713 → 305,571 |
Broken down by what people are actually doing, the pattern is clean. Anything with a fixed cost in gas went up by about the same multiple as the price itself. The swap went up nearly twice as much again, because the swaps themselves grew heavier over the same stretch.
| What each action cost, typical case | 22 Aug → 3 Sep |
|---|---|
| Send ether to somebody | $0.0010 → $0.024 (23x) |
| Send a token | $0.0023 → $0.053 (23x) |
| A bundled wallet operation | $0.024 → $0.51 (21x) |
| Settle somebody else's order | $0.025 → $0.60 (24x) |
| Swap through the main router | $0.012 → $0.48 (42x) |
For a trader the fee only means something measured against the size of the trade. The typical swap through that router on 3 September was worth about $73 and paid roughly two thirds of one percent of that in gas. Two swaps in every five paid more than one percent. Fifteen thousand of them paid more in gas than the trade was worth.
Timing helps a little and not much. Averaged over the last week of the surge, the dearest hour of the day costs not quite twice the cheapest, at its dearest in the early afternoon UTC and at its lowest around one in the morning. Nobody trading in reaction to a price move can wait twelve hours for a cheaper block.
Failed transactions carry the same charge as successful ones. On 3 September the chain collected about $254,000 from transactions that reverted and delivered nothing, more than it earned from every transaction of any kind on 22 August. Yet the failure rate itself halved over the same stretch, from one transaction in five on 21 August to one in ten on 3 September. When gas is nearly free, software fires on spec and eats the losses. When it costs real money the same software gets careful. It is the clearest sign in the data that anybody reacts to the price at all.
07 — The marginWhere the money actually goes
None of this money is destroyed. Chains like Ethereum burn part of every fee, which takes it out of supply and hands the gain to everyone holding the token. Robinhood Chain burns nothing. Every wei goes into two accounts the chain names when you ask it.
We watched both balances for three minutes while the chain was busy. The floor goes into one account. The entire congestion surcharge goes into the other. At the price on the day we measured, that meant 99 parts in every 100 of what users paid landed in a single account, which holds about 6,500 ether.
Set against that is the chain's own bill. A network of this kind has to publish its results back to Ethereum, and that costs real money, paid in blobs, the cheap bulk storage Ethereum sells for exactly this purpose. Robinhood Chain is a heavy buyer. On 3 September it posted 5,703 batches, three blobs each, touching more than half of all Ethereum blocks that day. Across the whole of Ethereum there were 15,652 blob transactions from 164 senders, and Robinhood Chain sent 36% of them.
| Robinhood Chain, 3 September 2026 | Amount |
|---|---|
| Fees collected from users | $4,503,705 |
| Paid to Ethereum for posting the day's data | $396 |
| Paid to Ethereum for proving the day's results | $2 |
| Ratio | about 11,400 to 1 |
That ratio is not profit. Hardware, staff and everything else a network needs cost money no ledger records, and none of it is in these figures. What the ledger does show is that the single largest cost a network of this design is supposed to carry, the bill for putting its data on Ethereum, is currently a rounding error against what it charges for the service. Blob space is very cheap at the moment, so the same day's posting would have cost more in a busier market for it. It would not have cost anything close to $4.5 million.
Users are not charged for that part at all. The chain sets the price of Ethereum data to zero in its own settings, which is why a plain transfer on Robinhood Chain costs exactly 21,000 gas and not a unit more. Everything a user pays is for the work, and the work is done by Robinhood.
08 — The gapWhat we could not explain
The chain gives out the settings that set its own price, and they do not describe what it is doing.
Ask it for its capacity limit and it reports the stock Arbitrum figure of 7 million gas a second. It has been running at 40 million for a week. Under the standard formula that gap would take the price to 25 times the floor in about a minute rather than 11 days, and the chain would never have sat on the floor through early August at twice the stated limit. Ask it for the backlog the formula prices from and it returns zero while charging 70 times its floor.
One of three things is true. The settings were changed recently and we cannot see the old ones, because the public node keeps no old state and returns an error for any block more than a few thousand back. Or the chain runs a different pricing rule from the published one. Or the figures it reports about itself are stale. We could not tell them apart, and we are not going to guess in print.
What we can say from the record is the behaviour. The price stayed on the floor whenever the chain ran below about 20 million gas a second. It rose above that, it has been pinned near 40 million since the end of August, and it last touched the floor on 24 August.
Whether it comes back down is a question about demand rather than about the chain. Both earlier episodes ended when the traffic thinned out. This one has not thinned yet.
The other limits are ordinary. Every figure here stops at 3 September, the last complete day. Fees are changed into dollars at each day's ether price, within one percent of the rate the chain's own records imply. The Ethereum blob bill is the one figure here we estimate: we summed the execution half exactly and sampled 24 batch receipts across the day for the blob half, and pricing every batch at the highest rate we saw would still leave the ratio above 8,000 to 1. We have not named the companies behind the router or the settlement contract, because neither is verified on a public explorer and we could not work out who runs them from the chain alone.
Anyone who wants to check any of this can query the same records through the Robinhood Chain API, the DEX trades API or the MCP server. It is the same chain where we counted tokenised-share trading and meme-token launches against Solana earlier this year, on a network that was then almost free to use.
| The record | Address |
|---|---|
| The swap router | 0x65050a9b…62c40dc |
| The settlement contract the 31 wallets use | 0xccc88a9d…f1c315be |
| The wallet funding 29 of those 31 | 0xf70da97812cb96acdf810712aa562db8dfa3dbef |
| The account receiving the congestion surcharge | 0xbc5c3a7adecf54d34169fd90dbd1b7d3142df067 |
| The account receiving the floor | 0x5a2b80a9b7effc06129bd5462d77bc20a8a59be7 |
| Robinhood Chain's batch poster on Ethereum | 0xdaa52608…8687f4 |
| The contract on Ethereum it posts to | 0xbd0d173e…33ba96 |
This analysis covers Robinhood Chain from its first block on 30 April 2026 to the end of 3 September 2026. Every fee figure is derived twice, once from block-level gas and price and once from the fee recorded on each transaction, and the two agree on every day tested. Daily figures exclude 4 September, which was incomplete when we measured. Median figures use exact quantiles rather than the approximate kind, which move by a few percent between runs.
Dollar conversions use each day's average ether price from an internal reference series. On 3 September that series gives $2,440.77 against $2,456 implied by the chain's own converted records, a difference of 0.7% that applies to every dollar figure here.
The Ethereum posting cost combines an exact sum of the execution charge on all 5,703 batches with an estimate of the blob charge from 24 sampled receipts spread across the day. Blob prices moved by a factor of 15 within that day. Pricing every batch at the highest rate seen would put the day's total near 0.22 ether rather than 0.16.
The two contracts described in section five are identified by what their code does and by the events they emit. Neither is verified on a public explorer, and nothing here should be read as a statement about any named company's conduct. The 31 wallets are treated as one because a single wallet funded 29 of them, which is a statement about the flow of ether and not about who owns them.
Gas measured as a share of trade value covers successful swaps through the main router on 3 September and uses the larger side of each trade in dollars. The pricing settings quoted in section eight were read live from the chain's own precompiles on 4 September 2026. The public node serves no historical state, so we could not read the same settings at an earlier block.
Run this kind of analysis on your own data
Every figure in this investigation came from Bitquery's archive of Robinhood Chain and Ethereum: blocks, transactions, fees and the raw event records behind them, across 40+ networks. The same data powers trading desks, risk teams and on-chain research.