What Robinhood's tokenized stocks trade against

1. The 195 tokens, and the signature that finds them

Stock tokens are ordinary ERC-20 contracts, which is awkward if you want to count them. Anyone can deploy something called GameStop, and on this chain somebody did. There is a cheaper filter than the name: every real stock token emits a transfer signature that no other token on this chain currently uses, the scaled-UI transfer with its ERC-8056 multiplier. One topic filter narrows the whole chain to a couple of hundred contracts. It is a filter and not a proof of authenticity, for reasons section 6 gets into.

The roster this report counts is the registry's own list plus one: the 194 contracts on the official registry page, and WEEK, the tokenized Roundhill T-Bill ETF at 0xc93a…099f, which is live onchain and missing from the page. 195 names in total. Three counts follow from it and they are all different, so it is worth keeping them apart: 194 of the 195 have ever emitted the scaled-UI signature, because BND is listed but has never minted a unit, and 193 have ever traded. The two that have not are instructive in different ways. WEEK has never appeared in a pool at all. BND has: five pools hold it, the first a BND/USDG v3 pool created on August 10, and not one of them has ever seen a swap. A token can have liquidity infrastructure, a listing, and no existence.

The trading itself is three pool designs, and three topic filters cover all of them: the Uniswap v4 singleton at 0x8366…0951, every v3-style pool including the 128 Algebra pools, and every v2-style pair.

That returns 200,632,972 swap events between genesis and the pin. Joined against the 1,204,799 distinct pools in the creation census (1,204,804 creation events, five of them duplicates), 198.2 million of those swaps (98.8%) fall in a known pool and 195.8 million (97.6%) can be given a dollar value. Every figure on this page counts a swap once and prices it from the chain's own pools, never an external feed. Section 5 covers what that convention does to the totals, and section 9 lists what it throws away.

2. What trades: NVDA, and then everyone else

One ticker is a third of the market. NVDA has traded $706.1M across 2.99 million swaps, and the four biggest names together account for 70% of all stock-token volume. Third place is the odd one: SpaceX, a private company, has done $303.2M here, more than four times what Apple has.

One line of that table deserves its assumption stated, because the assumption is what puts it there. When a pool holds two registry stocks, SPY against QQQ, there is no neutral way to say whose volume it is, and this report credits the whole pool-day to whichever contract address sorts lower. SPY sorts low, so $93.7M of its $310.3M, 30.2% of the ticker's total and 89.8% of the entire stock-against-stock bucket, is volume it shares with a counterparty that gets nothing. Credit both legs instead and SPY reads $318.5M and stays second. Split those pool-days evenly and it reads $267.5M and drops to third, behind SPCX at $303.3M. NVDA's first place is not sensitive to any of this. Second place is entirely a choice of convention, which is the same lesson as section 5 arriving one table earlier than expected.

Almost none of this is old. $1.69B of the $2.21B lifetime total traded in August, and $1.12B of that in the seven days from August 24 to August 30. Restating this report's own narrow series at the August 7 pin of the earlier guide puts lifetime dollar-leg volume at $402.1M. Twenty-three days later, with the method unchanged, it reads $1.40B. That $402.1M is computed here, not a figure the August 7 guide published; section 7 sets its own numbers beside this report's.

The list of tokens grew too. Robinhood announced 100 more stock tokens on August 13, and the chain shows 97 tokens minting for the first time on or after August 10. That batch has since done $174.4M, 7.9% of all volume, in under three weeks, and two of them are already top ten: DJT at $69.8M and the gold ETF GLD at $63.9M. The bigger mover was a token nobody had to launch. SPY first minted on June 23 and had traded $15.6M by August 7, 5% of its eventual total, then did $294.7M over August 8 to 30 and took second place on the table above.

3. What a stock token trades against

Sort the $2.21B by what is on the other side of the pool and the market splits into four, only one of which looks like the equities market it is named after. Trading against the chain's dollars, USDG plus a negligible amount of USDe, is $1.21B. Rotation between registry stocks, SPY against QQQ with no dollar involved, is $104.3M. ETH and WETH pairs come to $187.7M. The remaining $711.2M pairs a tokenized stock with a token the registry has never listed, mostly memecoins launched on the chain itself.

One caution before the table, because these two cuts are easy to mistake for each other. What a stock trades against is not the same question as which pools you agree to count. The dollar and ETH rows together, $1.40B, are the narrow counting convention of section 5; the $1.21B dollar row on its own is not that series and should not be compared to anyone's published total.

At 32.1% the memecoin bucket is a third of the venue, and its top pool is not small. AI, Artificial Inu, at 0x2e8c…1e18, has traded $91.3M against NVDA in that single pairing. Were AI a stock token, that would rank it fifth on the table above, between GME at $223.1M and DJT at $69.8M. The rest of the ranking of counterparts falls away fast: the counterfeit GME of section 6 at $40.4M, microduck at $15.0M, PAIR at $13.1M, SAYLORMOON at $12.6M. Behind those are 10,169 more counterpart tokens, of which 111 have cleared $1M. Note what is in second place there. The largest memecoin counterpart after AI is not a memecoin at all, it is a token impersonating one of the stocks in the table above.

If you are measuring this market, plan for that split. A third of the flow has no dollar leg and no equity leg, so it never shows up in a USDG-denominated query, and pricing it at all means falling back to the stock leg's own pool price. That choice is what section 5 is about, and it moves the total by hundreds of millions.

4. The weekend that ran without new supply

A venue that gated creation to market hours might be expected to go quiet when the underlying market closes, and through mid-August this chain behaved that way. Every Saturday-Sunday pair from July 25 to August 15 traded at 0.32x to 0.51x the average of its five prior weekdays. Then it flipped. August 22-23 came in at 0.95x, and August 29-30 at 1.96x, with Saturday's $222.0M and Sunday's $270.6M the two largest stock-token days the chain has had. Sunday's leaders were not curiosities either: SPY traded $76.3M, NVDA $26.7M, DJT $22.2M, GME $20.7M.

The primary market underneath that weekend goes quiet, and issuance leaves a log event, so the gap is countable: of the 27,966 mints and 2,078 burns in the chain's entire history, the number falling on a Saturday or Sunday is zero. Every unit of every stock token was created or destroyed on a weekday.

Nothing in the contracts requires that. The stock tokens are beacon proxies over one shared Stock implementation, checked on NVDA, GME, SPY and AAPL and identical for all four, at 0xb354…5ae2, whose mint is guarded by a minter role, a pause flag, and a blocklist. There is no time condition in it, and the word timestamp does not appear anywhere in that verified implementation source. The issuer can mint on a Saturday. It has just never done so.

So the $492.6M that traded on August 29 and 30 moved entirely in the secondary market, with no new supply arriving to meet it until Monday. The prices did drift, and less than the setup suggests. Taking each token's volume-weighted price in its USDG pools by UTC day, NVDA went from $223.84 on Friday to $218.79 on Saturday, down 2.3%, and recovered to $219.21 on Sunday; SPY went $771.48, $772.45, $776.42, up 0.6% across the two days. Neither token had a multiplier update in the window, so those are price moves rather than an accounting artifact. A weekend on this chain is not a hard peg to Friday's close, but over the largest weekend it has had, the drift was a couple of percent in either direction.

Worth keeping the two apart, though: what is measured here is Saturday and Sunday pool prices against Friday's, on one chain, over one weekend. The reference market was closed the whole time, so there is no independent price to compare against and nothing here establishes what the drift would have been had it been open.

5. Why your number will not match anyone else's

"DEX volume for stock tokens" is not one quantity. On the same logs, on the same day, three defensible conventions give three answers. Count only pools with a dollar or ETH leg and the cumulative total at the end of August 26 is $937.0M. Add every pool with no dollar or ETH leg, which is the memecoin pairs plus the stock-against-stock rotation from section 3, priced at the stock leg's own same-day USDG price, and it is $1.39B. Count both legs of every swap instead of one and the dollar-leg figure roughly doubles again, to about $1.87B.

Two public numbers from that week make useful calibration points. Robinhood posted $1.5 billion on August 26. On the wider convention this chain reads $1.39B that evening and crosses $1.5B the next day, so the two are a definition and a day apart. Messari reported an all-time high of over $100 million for August 28. That one is a floor, not a point, and it does not discriminate: the narrow measure here gives $101.3M for that day and the wider one gives $186.5M, and both clear $100M. It is worth stating plainly because a floor quoted next to a precise-looking series is exactly how two dashboards come to disagree without either being wrong.

Against DefiLlama's published all-protocol daily volume, fetched August 31, the all-protocol series here runs at 0.927x to 0.986x on every day from August 25 to 30, reconciling from below, which is what you want from a measure that excludes rather than guesses. Do not read that as agreement in general. Across July 1 to August 30 this series runs above DefiLlama's on 40 of the 61 comparable days, 22 of them by more than 10%, and July totals $17.2B here against their $14.3B. The gap is widest in the launch fortnight, when their coverage of the chain was still starting up and a low denominator makes the ratio meaningless. The late-August week is the part of the comparison that carries weight, and it is the part these figures rest on.

6. The counterfeit GME trades against the real one

The August 7 measurement flagged a token calling itself GME at a non-registry address as a warning about canonical address lists. Three weeks later it is an economy. The counterfeit, symbol GME and name "GameStop" at 0xc236…4ba3, first appeared in a pool on July 23. Since then 409 pools have been created around it, 314 of them have traded, and it has cleared $111.0M of priced volume against the canonical token's (0x1b0e…153e) $223.1M. Its volume splits almost in thirds: $40.4M against the canonical GME (36.4%), $36.4M against dollars (32.8%), $34.2M against ETH (30.8%). The first of those is the strange one. Six of the 409 pools pair the fake directly with the real token, five of them have traded, and those five account for the largest of the three slices.

That has a consequence for the canonical number, not just the fake one. Those five traded pools have a registry stock leg, so this page's method prices them and counts them, which means $40.4M of the canonical GME's $223.1M, 18% of it, is volume traded against its own counterfeit. Filtering on the registry address is necessary and it is not sufficient. The address tells you which token moved; it does not tell you what the other side of the pool was.

For anyone computing "GME volume on Robinhood Chain" the lesson is exact. The two contracts share a symbol and a name, both emit ordinary transfer events, and only the address separates them. The registry page says as much ("a token with a matching name/ticker but a different contract address is not a Robinhood Stock Token"), and the scaled-UI signature from section 1 sorts these two today: the canonical token emits it and this counterfeit does not. Every stock figure on this page passes that test, so the $70.6M the counterfeit trades against dollars and ETH is in no stock-token total anywhere else here. It is not invisible though: those are ordinary priced pools, so that $70.6M is inside the chain's $33.54B all-protocol volume and inside the denominator of the share line in section 2. The $40.4M is the other exception, and it is in the stock totals for the reason just given.

Do not promote that filter into a proof of authenticity, though. An event signature is just a topic hash, and nothing stops a spoof from declaring the same event and emitting it, exactly as this one already emits ordinary transfers. The signature is a cheap first pass that the counterfeits deployed so far happen to fail; the registry's addresses, or deployment provenance you have checked yourself, are what actually settle authenticity. Key your analytics on symbols and you will count this counterfeit into the float today. Key them on the signature alone and you will count the first spoof that bothers to copy it.

7. Transactions, pools, and float

Underneath the trading are the chain's own numbers, worth restating against the August 7 measurement: same method, same dataset, twenty-three days apart.

The transaction count gives the growth a scale. Only 1.38 million transactions existed before the July 1 public launch. The 100 millionth came on July 18, the 250 millionth on August 4, and the 500 millionth on August 26. That count includes the chain's own system transaction (ArbOS type 0x6a, one per block, and the chain's 100ms cadence puts that near 865 thousand blocks a day once it is running at full rate), and its share of a day's transactions swings far more than the raw count suggests: from 3.2% on August 12 to 46.2% on July 5, median 9.6% across the 61 days. That share is blocks divided by transactions, so it assumes exactly one type 0x6a per block rather than counting them; on a quiet day the chain is mostly talking to itself. The busiest day, August 12 at 26.9M transactions, was not a stock-trading day: stocks were $30.9M of $480.8M total DEX volume. What it was is not something this data settles. Pool creation, the one launch proxy available here, puts August 12 at 18,069 new pools, a little above the median day of roughly 16,700 and less than half the July 21 peak of 37,872.

The dollar system grew with the trading: the USDG float went from $342.7M at the end of August 7 to $448.5M (6,269 mints against 6,124 burns lifetime), USDe from $242.3M to $324.3M, and the combined $772.8M float is the cash leg for the $1.2B of dollar-paired stock trading in section 3. The equity side stayed small in float terms even as it exploded in volume terms. Netting every mint against every burn and marking the 46 tokens with liquid USDG pools at their own latest pool price values the tokenized-equity float at $52.4M, led by NVDA at $9.4M and SPY at $5.9M. That is a floor, since 149 of the 195 tokens have no liquid dollar pool to price them, and against that floor the volume is a trading venue's rather than a custody venue's: August's $1.69B is 32 times the $52.4M float, about 7.5 times a week, and the last seven days alone are 21 times it. Corporate actions kept running through it all: 15 multiplier events across nine tokens (AAPL, ASML, COST, CRWD, DELL, MU, ORCL, SGOV, WEEK) in the window. Those are events, not distinct actions. Two are duplicate emissions of the same change, and one WEEK update on June 23 is reversed fifteen minutes later, so about twelve actions produced the fifteen logs.

8. The answer, and what would move it

By value, tokenized stocks on Robinhood Chain trade mostly against dollars. On the $2.21B measured here through August 30: $1.21B (54.7%) against USDG and USDe, $711.2M (32.1%) against tokens the registry has never listed, $187.7M (8.5%) against ETH and WETH, $104.3M (4.7%) stock against stock.

Count trades instead of dollars and the answer inverts. Of the 15,483,914 swaps with a stock leg, 7,368,515 (47.6%) are against a token the registry has never listed and only 5,763,720 (37.2%) are against dollars, with ETH at 10.6% and stock-against-stock at 4.6%. The memecoin trade is the typical trade on this chain; the dollar trade is the large one. Both sentences are true of the same swaps, and which one a dashboard reports is the difference between calling this an equities venue and calling it a memecoin venue with equities in it.

Either way, a market whose largest single non-dollar pairing is a memecoin against NVDA, at $91.3M, is not behaving like the equities market its tickers are borrowed from. Whether that changes is not something this window can tell you.

Three things would move that answer, and each is measurable on the same logs. The memecoin share is concentrated: drop the Artificial Inu pair alone and the non-registry bucket falls from 32.1% to 29.2% of a $2.12B total while dollars rise from 54.7% to 57.1%, so a single counterpart token carries roughly three points of the headline. The batch announced on August 13 as 100 tokens, of which the chain shows 97 minting for the first time, has done $174.4M in under three weeks against a lifetime total that is 76% August, so a second batch would rewrite the ticker ranking before it rewrites the counterpart mix. And the dollar share is a function of how much USDG is on the chain: the $772.8M combined float grew $187.8M since August 7 while August's dollar-paired stock volume was $928.0M, or 1.20 times that float. Those are a stock and a flow and the ratio is not a turnover figure, but the dollar leg of every trade has to come from somewhere, so the float is the number to watch.

What this report does not settle: whether any of it is retail, since an address is not a person and no attribution was attempted; how much is wash or bot flow beyond the $17.19B single-day loop and the per-swap ceiling excluded in section 9; and what the tokens are worth against their underlying equities, which needs an offchain price feed this report deliberately never used. The counterpart split is measured. Its causes are not.

9. Method, exclusions, reproduce it

The window first. Every series ends at block 50,419,085, found by interpolation search on block-header timestamps; two single-block probes prove it is the boundary:

The exclusions, enumerated. First, 2,383,561 swap events (1.2% of the 200.6M observed) come from 1,252 contracts that emit Uniswap-style Swap logs without any recorded factory creation event; the largest have four or more tokens at once, which no two-token pool does, and none of the top holders include a registry stock. They are excluded rather than guessed at. Second, 2,486,475 swaps (1.2%) are in known pools pairing two tokens this method cannot price; 37,698 of them involve a stock token with no usable USDG price that day: 35,054 had a pool below the $10,000 floor and 2,644 had no USDG pool at all, so the stock series is a floor too. Third, the wash guard: eight pool-days on August 19, all Uniswap v4 pools paired against four non-registry tokens, claimed $17.19B across just 514 swaps, an average of $33M per swap against a chain-wide average of $171 per priced swap, and against $438 in the single busiest pool on the chain. That is flash-loop accounting, not trading, and it is excluded by a stated rule (any pool-day averaging over $5M per swap), which fires on those eight pool-days and nowhere else in 1.6 million pool-days. Before that rule fires, this report's own August 19 total is $17.78B; after it, $593.7M. DefiLlama publishes $510.9M for the same day, so both measures cut the day down by orders of magnitude, though what their filter starts from is not something this page can observe.

Pricing. USDG has 6 decimals and USDe 18, both at the $1 peg. The ETH price is derived inside the chain: total USDG through the ETH/USDG and WETH/USDG anchor pools divided by total ETH through them, per UTC day, which put ETH between $2,444 and $2,506 over the last week of August. A stock token's price, where needed, is that day's USDG through its pools divided by shares through them, the same volume-weighted construction the August 7 guide used, with its $10,000-per-day liquidity floor. The ERC-8056 multiplier cancels out of every volume number on this page because pools trade raw units; it enters only the float valuation of section 7, where raw supply is scaled by each token's current multiplier.

The weekend finding from section 4 reproduces end to end in a short Pipes SDK script. It streams every mint event in the chain's history, the scaled-UI transfer signature with topic1 pinned to the zero address, and tallies by UTC weekday:

Run against the pinned range it covers 50.4 million blocks in about twenty minutes on a single stream, averaging roughly 40 thousand blocks per second, because the Portal sends only the blocks that match the filter. Moving the zero address from topic1 to topic2 counts burns instead, which is how the 2,078 burns behind the same finding were counted: likewise none on a Saturday or Sunday.

The volume headline is a longer job and worth being honest about the cost. Reproducing $2.21B means two more sweeps of the pinned range and a good deal of arithmetic on top: every pool-creation event from the Uniswap v2, v3 and v4 factories plus Algebra, to learn which two tokens each of the 1,204,799 pools pairs; every Swap event, 200,632,972 of them, aggregated to one row per pool per UTC day; then the two derived price series above; then the cascade and the three exclusions. Both sweeps are the same Portal stream and the same filter shape as the script above, and the swap pass is much the slower of the two, since it matches 200 million events where the mint filter matched 28 thousand. Nothing in it needs an archive node or an external price feed, which is the property worth copying whether or not you reproduce this particular number.

10. Robinhood Chain data with SQD

Every measured number here comes from four filtered log sweeps and one transaction sweep over SQD Portal's robinhood-mainnet dataset, the exceptions being named where they appear (DefiLlama's published volumes, the two X posts, and the verified contract source): swap events by three topic filters, pool creations by four, stock issuance and its multiplier updates by two signatures, and dollar mints and burns by token address. The block-header probes behind the pin and the per-day block counts are transaction and header reads over the same dataset, and the token names come from a block explorer's token API, not from the dataset. The dataset serves the chain from block 0 with logs, transactions, and traces, keeps the ArbOS transaction types in the type field, and streams filtered history at tens of thousands of blocks per second, which is what turns a full-history market study into an afternoon rather than an infrastructure project. The Portal serves 210+ networks through the same query shape, and the Pipes SDK wraps it in typed streams like the script above.

Frequently asked questions