A market built on attention is forming quickly where meme coins and tokenized stocks meet. Bitquery Research says that over a 31-day sample, five launchpads generated 43.3 million trades tied to 148,000 tokens, and nearly three-quarters of buyers lost money. Aggregate net losses came to about $287 million. By the report’s measure, the only entry window that produced consistently positive returns was the first 10 seconds after a token’s first trade.
The study, written by Bitquery Research and translated by TechFlow, tracks trades across Solana, Robinhood Chain, and BNB Chain where one side of the pair was a meme coin and the other was a stock token. It also makes a broader claim: meme-coin settlement now accounts for two-fifths of onchain trading volume in tokenized stocks.
Buying the joke coin now can mean holding a tokenized stock first
On Sept. 9, at almost 8 p.m. UTC, a trader on Solana bought a meme coin using a token tracking the S&P 500 index. The payment asset was not SOL or USDC. It was an index-linked token. Bitquery says this was the first such trade on Pump.fun.
Pump.fun launched Custom Pairs on Sept. 9. The feature lets token creators choose what their coin is priced in, including Tesla, NVIDIA, the Nasdaq, or gold. Within 12 hours, the report says, hundreds of tokens had already used the feature.
The report separates a few terms. A stock token is a token that tracks a real stock and is issued by a company that says it holds the underlying shares. Backed calls them xStocks, Binance uses bStocks, and Robinhood and Ondo use their own naming. A launchpad is a site that mints a new coin and gives it an instant market. The quote asset is the asset used to price that market, and it is what a buyer must hold before entering. Once SOL is replaced by tokenized NVIDIA, the buyer is taking on two bets at once.
Bitquery indexed every trade it could see on Solana, Robinhood Chain, and BNB Chain, then pulled a 31-day window and isolated every trade where one side was a meme coin and the other was a stock token. That produced 43.3 million trades. From there, it tracked what each wallet put in and what it took out, and then followed the money to the final recipients.
The report also sets out one limitation up front. Its window starts on Aug. 11 because that is as far back as this level of trade history could be reconstructed. It is not an all-time dataset, and the Pump.fun numbers cover only about a day and a half.
Five launchpads entered the race over four weeks, and Pump.fun was neither first nor biggest
Pump.fun captured attention, but Bitquery says it was not first and not the largest venue in this category.
On Robinhood Chain, Pons had been using Robinhood’s own stock tokens as quote assets since Aug. 12. The report says the number of comparable tokens there is now about 28 times that of Pump.fun. Raydium’s LaunchLab followed on Solana on Aug. 24. Flap went live on BNB Chain in September. Four.meme, also on BNB Chain, had already been doing this quietly and labeled the category “Stock Memes.”
Share matters more than raw count in the report’s framing because it reflects what creators are actually choosing. On Raydium LaunchLab, four out of every five new coins are now stock-priced. On Flap, that share moved from zero to about half in three days.
The chart caption says daily issuance rose from 53 to 4,270 between Aug. 12 and Sept. 9. The data stops on Sept. 9 because only partial data was captured for Sept. 10.
Bitquery says stock-priced coins also show higher first-hour survival and are more likely to reach real exchanges. On Pons, a standard coin lasts 14 minutes rather than five. Across all five launchpads, a higher share of these tokens graduated from the bonding curve. The report then warns that this does not mean buyers had better outcomes.
Nearly 500,000 wallets bought and never sold, and net losses reached about $287 million
All figures in this section are measured on a cash basis: actual dollars paid versus actual dollars taken out. There is no mark-to-market accounting, and the report later explains why.
Nearly 500,000 wallets bought one of these tokens and never sold even a single unit. That group alone lost about $302 million. Wallets that did manage to exit posted a combined net gain of $251 million. In Bitquery’s reading, that is why the market can keep going: enough people escape with gains to make the whole thing look playable.
Losses looked similar across issuers. Whether the quote asset came from Robinhood tokens, Binance bStock, Backed xStock, or Ondo wrappers, the share of profitable wallets still landed between one-quarter and one-third. The report’s conclusion is that the problem appears structural rather than specific to any one issuer.
One reason is depth. Tokenized stocks themselves are thin, according to the report. A $10,000 order can already move the market before a meme coin is layered on top, and that is visible in standard DEX trade data.
Bitquery also compares like-for-like launches on the only platform that ran both categories at scale for a full month and could be paired by issuance date. The result: stock-priced meme coins lived longer and reached real exchanges more often, but their buyers still lost more. A longer lifespan, the report says, simply meant more time for distribution to later buyers.
Seller gains were concentrated, with 418 wallets taking two-fifths of the money
If buyers lost $287 million, someone else had to be on the other side. Bitquery identified wallets that sold a token without ever buying it. In the report’s interpretation, those counterparties are easy to map to creator inventory, team allocations, or airdrops.
There were about 264,000 such wallets. They paid nothing and extracted about $268 million, equal to 93% of buyer losses, just in the opposite direction. The money was highly concentrated. A small set of wallets appeared across thousands of different coins. Bitquery describes the top of that distribution as a scaled business, with a real but much smaller long tail underneath.
Ten seconds was the only profitable entry window
Bitquery reordered each trade by how long after a token’s first trade the buyer arrived. Its answer was blunt: the first 10 seconds were the only window with positive returns.
Returns in that interval reached +18.4%. After that, price performance decayed monotonically. Buyers who entered a day later lost more than one-fifth of what they put in. About one in every 10 trades happened in that earliest window.
The report argues that winning here depends on execution that very few participants have. It split each active wallet’s history into first-half and second-half samples and found win-rate correlation at only about 0.5, so leaders usually stayed ahead. Dispersion in win rates was more than double what luck alone would be expected to produce. Even then, only about 1% of active wallets posted win rates above three-quarters.
Bitquery also pushes back on the idea that the activity is mostly wash trading. Cases where the same wallet both bought and sold the same token in a single trade represented less than 1% of trade count and about 2% of volume, with a similar pattern on Pumpswap and elsewhere. The report’s conclusion is simple: the volume is largely real, and so are the losses.
The stock leg barely mattered in most short-held trades
Buying a meme coin priced in tokenized NVIDIA means making two bets at once: the coin against NVIDIA, and NVIDIA against the dollar. Bitquery separated those exposures and says that in 7.1 million round-trip trades that had already been closed, the second leg barely moved.
Most of the story is just six minutes long. A blue-chip stock usually does very little in six minutes, so the stock in the pair worked more like a label than a meaningful position. Over longer holding periods of one day to one week, the stock side did start to cost traders money, to the tune of about $23 million, and it almost never picked the winner.
The only exception the report highlights is Four.meme’s tokenized Cea Industries. It describes that stock as small and highly volatile, with moves over comparable holding periods around 60 times those of a Nasdaq token. In that case, the quote asset started to matter.
This market largely ignored stock-market hours
The assets used to price these coins are tradable elsewhere for only about 32 hours a week. The meme coins tied to them trade around the clock.
The report breaks activity into hourly heatmap cells. Outlined cells mark New York trading hours, from 09:30 to 16:00. That window makes up 19.3% of the week and 20.8% of total traffic, which means activity was broadly flat across the week. Weekends alone accounted for about one-quarter of total flow.
In other words, this market was not waiting for the opening bell. Bitquery says it had seen a similar pattern earlier when looking at Binance bStocks trading at 3 a.m., and adding a meme coin made that behavior sharper.
Meme settlement now makes up two-fifths of tokenized-stock onchain turnover
Bitquery says its most important finding is not really about meme coins alone. It looked at every transaction in the window that touched any of the 403 stock tokens in the dataset, no matter which side of the pair the token appeared on.
The result: two-fifths of onchain turnover in tokenized stocks now comes from meme-coin settlement. For Backed, Ondo, Binance, Backpack, and Robinhood, this has become one of the biggest real usage channels for their products, despite the absence of any dedicated launch announcement or press release. The report adds that when it mapped the xStocks market on Solana earlier in the summer, this channel was still a footnote. On BNB Chain, Four.meme now mints stock tokens itself when a ticker starts trending, and that has become a focus.
How the sample was built and what was excluded
Bitquery counted a quote asset as a stock token only if it had been minted by a recognized stock-token issuer. That list included Robinhood Chain tokens, Binance bStocks, Backed xStocks, PreStocks, Backpack Securities, Ondo Global Markets, Four.meme stock tokens, and Robinhood’s pre-IPO and Korea-listed names. The total came to 403.
The report says name-based search fails here. It would pull in a meme coin called “Stock Cat” while completely missing Binance bStocks, because those tokens use plain company names such as NVIDIA Corp with no issuer tag. That unlabeled segment makes up 40% of the market.
Pairs were collected from both directions. Most venues record the security as the quote asset, but about 600,000 trades were reversed and treated the security as the base asset. Those trades sat in separate pools and were not duplicates.
Four classes of data were removed because each one would distort the results:
- Robinhood Chain’s Uniswap v4 hook writes fees from each swap as two extra swap rows, so more than one-quarter of the venue’s rows were not real trades. That also exposed fee economics: 1.14% per swap and about $35 million over a month.
- Trades between one stock token and another form a real market, but not the one analyzed here.
- One liquidity pool contained 201 rows with amounts large enough to hit the maximum of a 64-bit counter. Bitquery treated that as a decoding artifact.
- Marking unsold holdings to the last trade would have created $2.39 billion of paper “profit.” The report gives one example where a position cost $383 but would have been valued at $363 million under that method, which is why there is no mark-to-market accounting anywhere in the study.
The report also notes two things it does not want to overclaim. First, the buyer comparison in section three came from only one launchpad, because it was the only one running both categories for a full month in a way that could be matched by issue date. Second, Pump.fun’s Custom Pairs was not limited to stock tokens. It also launched tokens priced in PUMP, USDC, and wrapped bitcoin.
All data in the report was pulled through Bitquery’s trading API.
