RWAs,
Unwrapped

Start with the picture

The map above is the entire market, every holder on a single page. Each cell is one holder, sized by how much it holds and colored by the kind of entity behind it, every balance read straight from the chain. The chain does not come with names, though, and a list of anonymous balances cannot answer much on its own. The work behind this map is putting names to those balances, holder by holder, and it is the foundation the rest of the report stands on.

Why this matters

Tokenization's totals get quoted as proof that it has arrived. But the same billion-dollar balance can be an outside investor putting capital to work, a trading venue, a lending protocol, or an issuer holding unsold inventory of its own token. As numbers they look identical. What separates substance from story is knowing who the holders are and what they do with what they hold. That question runs through this report, and it is why the map comes first.

The shape of the market, by token

A handful of tokens carry most of the value, and the names below recur throughout the report.

What the map already shows

A few cells dominate while a long tail of smaller holders fills the edges. The single largest holder is not an asset manager but Sky, the DeFi stablecoin protocol once known as MakerDAO, holding about $2.9 billion as reserves backing its stablecoin. Other large regions belong to names you might not expect, the issuers themselves holding their own product among them. And a meaningful share sits in the pale region we have not yet attributed. The sections that follow take these one at a time.

Much of the map is wallets the issuers run

One of the map's largest combined regions is not pension funds or asset managers taking on exposure. It is wallets the issuers, their platforms, and their affiliates operate, about $5.35 billion, roughly 38 percent of the total. About $3.27 billion is the issuers holding tokens tied to their own businesses. About $2.08 billion is outside investors pooled in custodial wallets the issuers or their platforms run on their behalf. The first is supply that looks like demand. The second is real demand that never reaches the chain. Throughout this report, issuer is shorthand for the party operating a token's on-chain program, whether the asset manager itself, like Ondo, or the tokenization platform acting as its agent, like Securitize for BlackRock's BUIDL or Libeara for ChinaAMC's money funds.

What the headline looks like once you net the issuers out

Stated as a simple waterfall, the adjustment is large.

The issuers holding their own product

An issuer can hold its own token in several reasonable ways, and each grows the headline without adding a buyer

Treasury parks the issuer's own balance sheet in the assets it issues. Unsold inventory is minted supply no one has bought yet. Cross-backing holds token B in order to issue token A, and Ondo, by its own public account, builds products of this kind. Redemption float sits on the issuer's books until the burn is finalized. All are ordinary practice.

But each raises the market total with no new participant arriving, and when token A is backed by token B and both are counted, the same underlying dollar appears twice. From the headline number alone, none of it can be told apart from a genuine outside purchase.

Backing bought by a third party is a different thing, and we count it as demand

The netting here removes only tokens tied to the issuer's own business. When an outside party buys a tokenized asset to back a product of its own, as Sky, the market's largest holder, does with the $2.9 billion of reserves behind its stablecoin, that is an independent buyer putting capital into the asset. It is crypto-native demand rather than traditional capital arriving, but it is demand, and it stays in the independently held total. We hold that classification, and a reader who draws the line differently can reprice the market with the numbers on this page. Treat the $2.9 billion as demand internal to the crypto system rather than outside capital and it moves across the line: the independently held market becomes about $7.8 billion rather than $10.69 billion, and the value held directly on-chain by independent owners becomes about $5.7 billion rather than $8.61 billion. The classification is a judgment call. The data lets a reader take either side of it.

Seven wallets hold 87 percent of one flagship token's circulating float

The clearest case is Ondo's USDY, a tokenized dollar that pays a Treasury yield. Seven wallets hold 87 percent of its circulating float on Ethereum between them, about $414 million of $476 million, and every one of them was funded by the same hub.

Because they share that funding, on-chain they behave as one holder rather than seven independent investors. The concentration is observed. The common control is our assessment from the on-chain pattern, not a label any wallet carries.

The evidence points to Ondo or its affiliates

Several independent threads tie these wallets to Ondo, the issuer of USDY, or to entities affiliated with it. Some connect to wallets Ondo has itself disclosed. Value flows one way out of them into wallets we attribute to Ondo. The multi-signature custody setup matches Ondo's own.

And the dollars that funded the purchases were withdrawn from the same handful of exchanges into which Ondo's treasury had been depositing its ONDO governance token. That last link cannot be proven on-chain, because exchanges pool everyone's money together, and the chart marks it as a break for that reason. Taken together these threads point, with high confidence, to Ondo or entities closely affiliated with it.

What it means, and what we are not saying

There are ordinary reasons an issuer might hold its own token at scale. Some of this balance may be unsold inventory, custodial or bridge holdings, or liquidity the issuer is seeding in its own market rather than an absence of outside demand. We are not alleging wrongdoing, and the point here is the market, not the conduct.

What the data shows is plain: 87 percent of the circulating float of one of tokenization's flagship assets sits in wallets we assess to be controlled by the issuer or its affiliates, not with outside buyers. We call that share the self-held float: the portion of circulating float sitting in wallets linked to the issuer or its affiliates rather than in independent hands. It is the most extreme version of the pattern this section describes, and none of it is visible in a headline total.

Outside money, pooled off the chain

The second region is a different thing entirely

About $2.08 billion of it is not the issuers' own money. It is outside investors pooled in custodial, or omnibus, wallets the issuer or its tokenization platform operates on their behalf. Standard Chartered's Libeara platform pools outside fund investors this way, and WisdomTree pools retail buyers of its tokenized funds the same way. This is genuine outside demand. The largest of these pooled wallets, Libeara’s $706 million, is worked in full in Section 6.

But it is demand for the asset, not for the tokenization

These buyers never hold the token themselves, so they cannot move it, self-custody it, or use it anywhere on-chain. WisdomTree, by its own description, keeps the blockchain entirely invisible to its retail users.

The benefits tokenization is meant to deliver are exactly the ones a pooled holder does not receive. The clearest exception is reach. Platforms like Libeara use tokenization to offer assets across borders to investors who would otherwise struggle to access them,² a real advantage even when the holding stays pooled. For most of this segment, though, the tokenization is back-office plumbing the buyer never touches.

So the honest gauge is smaller than the headline

That is the $8.61 billion line: independently held, on-chain, by the owner. A meaningful part of even that is crypto-native demand, tokenized cash parked by on-chain businesses, rather than traditional capital moving in.³ The headline counts the industry's own inventory and its off-chain custody alongside genuine adoption. Separating them is the only way to see how much of each there is.

Composability is the cleanest test of whether tokenization delivers something new

A holder can post a token as collateral, lend it, trade it, or plug it into an automated strategy, without asking permission and without taking the asset off the chain. Every other advantage this report tests has an off-chain approximation, liquidity in a secondary market, broader access in a feeder fund. Composability has none.

A Treasury bill in a brokerage account cannot be pledged as collateral in an unrelated venue in the middle of the night. A tokenized Treasury bill can. If these assets are widely used this way, that is real adoption. If they mostly sit still, the tokenization is closer to form than function.

We measure composability by DeFi flow, not by holdings

A trading pool can hold almost nothing while turning over enormous volume, and a reserve can hold billions while doing nothing but back something else, so we measure the volume of RWA tokens actually moving through DeFi venues and exclude assets held as passive backing. Backing is a holding, not a use.

Measuring it this way has a steep prerequisite. Telling a first hop from an interior one means classifying every address these tokens touch and naming the venues, so the measure rests on the same attribution layer as the holder map, extended across the 130 DeFi protocols these tokens interact with. Without that map, the same activity would be counted several times over.

Touching DeFi is common, using it in earnest is not

The 208 tokens (of 418) that have interacted with a DeFi protocol at least once carry about 68 percent of the market's value, and the experimentation spans every asset type. But for most the contact is a handful of transfers, often one-off tests: 186 of those 208 have under $100,000 of lifetime DeFi volume apiece. Genuine, repeated use is the work of a few names. Five tokens are 91 percent of all DeFi volume, and two Maple credit tokens alone are about 60 percent of it.

The part that is used is growing fast, up 26-fold in fifteen months

Volume through DeFi has scaled from about $13 million a month at the start of 2025 to $344 million in March 2026. Over the same window the outstanding balance of RWAs grew about 3.8-fold, so each dollar of RWA is being put to work in DeFi far more than it used to be.

And the testing is not confined to one corner. The tokens carrying real dollar volume are mostly crypto credit and a few short-term treasury funds, but the long tail of small and first-time DeFi transfers runs across treasuries, credit, funds, and tokenized equities alike. The honest read is an early market testing widely and using narrowly, with the trend pointing up on both.

What is driving the growth

The activity is what any liquid asset is used for, now done on-chain without a gatekeeper. RWA tokens are traded on decentralized exchanges (Uniswap), pledged as collateral (Aave), managed by curated vaults that act as on-chain funds (Morpho), and split into principal and yield components (Pendle). None of these uses is large yet, but each is an RWA doing something its traditional form cannot. A meaningful share of the lending is recursive leverage looping, a position pledged, borrowed against, and redeposited to multiply exposure, which lifts these volumes and concentrates risk in a way we return to at the end of the report.⁴

Centralized venues, too

The same property is at work when these tokens move onto centralized rails. Exchanges, lending desks, and trading firms hold about $124 million of them, standing balances rather than flow, small next to the asset base but a second place these assets do something their traditional form cannot.

This finding is only about private markets, the $4.71 billion where the underlying asset is genuinely hard to sell. Treasuries and public equities are already liquid, so tokenizing them adds around-the-clock access but no real breakthrough. In traditional form these private assets lock investors in for years: secondary stakes turn over about 2 percent of value a year, and the vehicles built to offer an exit cap redemptions near 2 percent a month. If tokenization delivers liquidity anywhere, it should be here, and at the headline level it looks delivered, with close to 30 percent of the base redeemed or sold on-chain every month. That average does not survive a look at which tokens are doing it.

The turnover is cash-like tokens, not the asset class

Five tokens, Maple's syrupUSDC and syrupUSDT, Janus Henderson's JAAA (tokenized on Centrifuge), Superstate's USCC, and Resolv's RLP, are two-thirds of the private-markets base and account for essentially all of the turnover. They redeem about 39 percent of their value a month, two of them more than 100 percent, which is not an illiquid asset being unlocked but a cash-like instrument being parked and pulled.

Strip the five out and tokenized private markets are no more liquid than traditional ones

The remaining 41 tokens, $1.56 billion of more traditional private assets, redeem about 2.5 percent of their value a month, the range of a traditional interval fund or non-traded REIT. Their secondary trading is $0.16 million a month across all 41, with only three trading at all. The largest does not move: BCAP, $757 million of tokenized venture capital and the sixth-largest token in this report, turns over zero.

What is new is the exit rail, not a market

We count the ways out, redemption to the issuer and sale to another buyer, because for an illiquid asset liquidity is the ease of getting out. The on-chain redemption rail genuinely works, around the clock and on demand, faster than any traditional structure allows. But the distinctive thing tokenization is supposed to add, a secondary market where illiquid assets change hands between buyers, has not appeared outside those few tokens.⁵

Where a secondary market exists at all, it is thin and pricing can stray far from NAV

One token in this set, sAID, a tokenized specialty-finance credit fund, is among the very few illiquid RWAs with any secondary trading, and isolated trades printed as low as 30 percent of NAV. Most trades clear much closer, and prints that low come from a book of a handful of trades rather than a sustained level. They are still a reminder of what a functioning exit is worth when the market for it is this thin.

Liquidity only for the cash-like

For the assets the promise is about, tokenized venture capital, real estate, and lock-up private credit, the redemption on offer is no better than a traditional fund's and a secondary market has not formed. One caution survives: the redemption door that lets cash-like tokens be pulled on demand in calm conditions can carry a rush for the exit in stressed ones, faster than any traditional structure would.⁶

Compliance is where programmability meets a hard legal line

The pitch is that a tokenized security can carry its own rulebook, eligibility and transfer restrictions written into the contract, so an ineligible transfer simply fails. Regulators have been emphatic that the wrapper changes nothing. "Tokenized securities are still securities," SEC Commissioner Hester Peirce has written, because blockchain "does not have magical abilities to transform the nature of the underlying asset."⁷ SEC Chairman Paul Atkins put it more bluntly: "Economic reality trumps labels."⁸ But a traditional security never enforced its own restrictions. Registered intermediaries did. A token can pass from wallet to wallet with no intermediary in between, which raises the question this section tests: when it moves on its own, what keeps the restriction in force?

We identified each token's enforcement model empirically, by testing transfers on-chain

Using Tenderly, we simulated multiple transfers per token, to and from arbitrary addresses and addresses on the OFAC sanctions list, and watched which clear and which revert.⁹ That sorts the market into three models.

A gated token (a whitelist) blocks every transfer unless the receiving address has been approved. This is the model built for regulated securities and standardized in ERC-3643,¹⁰ the permissioned-token standard the SEC chair has pointed to as a template for compliant tokenization.

A blocklist token is weaker. It allows transfers freely but blocks any address the issuer has listed, sanctioned wallets among them. This is the sanctions-style screening US law now requires of stablecoin issuers under the GENIUS Act.¹¹ It is a sanctions screen, not the control over who may hold the asset that a security calls for.

An open token carries no on-chain restriction at all. Whatever eligibility the issuer checks happens off-chain at the point of sale, and once the token is issued nothing in it stops the buyer from passing it to anyone.

Just over half the value is gated, but five in six tokens are not

Gated tokens hold 56 percent of the value, the large institutional funds whose compliance teams require that only vetted wallets hold the asset. By count, 85 percent of tokens are not gated at all, the long tail of tokenized equities and DeFi-native credit almost anyone can hold. So whether programmable compliance is real depends on what you count. For most of the money, it is real and enforced in code. For most of the tokens, the rulebook, if there is one, sits off-chain.

Most of these tokens never check who is eligible to hold them

In the simulations, the gated tokens refused any recipient that was not pre-approved. The blocklist tokens blocked only the sanctioned addresses. The open tokens, about $3.3 billion worth, let everything through, screening nothing at all, not even an address sanctioned for terrorism financing.¹² The open label describes a token's on-chain transfer behavior only. It is a structural property of the contract, not evidence that any issuer has failed a legal obligation, since for these tokens eligibility is screened off-chain at the point of sale rather than enforced in code.

$3.3 billion of tokenized assets carry no on-chain screen against a wallet sanctioned for terrorism financing. We tested it in simulation: no tokens moved. Whatever eligibility checks exist happen off-chain at the point of sale, not in the token.

Restricted securities are not supposed to move this freely

Securities sold under US exemptions can be held only by eligible investors and cannot be freely resold without another exemption.¹³ On-chain, only a gate written into the token or a registered intermediary standing in every trade can hold that line. The gated tokens have the first. The blocklist and open tokens have neither, and a sanctions screen is not an eligibility check. We are not lawyers, and the SEC itself has not said compliance must be coded in. It has set that question aside.¹⁴

Many of the open tokens are also tokenized non-US equities or DeFi-native credit whose securities status is genuinely contested, and we are flagging a structural gap, not asserting that any token has broken the law. Still, by the regulators' own standard, and only if these tokens are US securities, the large majority by token count would be moving more freely than the rules for restricted securities contemplate. The market is growing on a compliance question no one has answered. The last time digital assets scaled this fast on a question this open was the 2017 initial coin offering boom. The SEC found the tokens at issue were securities subject to the same laws,¹⁵ and a wave of enforcement followed.

One variable keeps reappearing: the compliance model

Sorted by the transfer model from the previous finding, the earlier findings line up, though not into the two camps you might expect. The market sorts into three. Gated tokens, 56 percent of the value, are institutional money in relatively few hands. Blocklist tokens, 20 percent, are the most broadly held of the three. Open tokens, 24 percent, are broadly held as well, and they carry almost all of the on-chain use.

Broad ownership and real use fall on different lines

Adoption usually bundles two things, a wide base of holders and active use, and the three models pull them apart. By ownership, the line is the gate. Small wallets are about 91 percent of blocklist holders and 84 percent of open holders. Among gated holders they are only about 28 percent, once five legacy 2018-19 security-token offerings and sub-$100 dust are set aside, and the value sits with institutions.

By use, the line falls elsewhere. Open tokens turn over about 21 percent of their value through DeFi each quarter, gated and blocklist tokens a fraction of one percent. The blocklist tokens are the clearest case, the most small holders of any group and almost none of the use, tokenized stocks and dollar tokens that many people hold but few do anything with. Wide ownership reaches two of the three models. On-chain use reaches only the open one.

Where access actually lands

The open and blocklist tokens together carry on the order of eighteen thousand holders, around nine in ten of them under $10,000, ticket sizes far below what the traditional versions allow. The gated tokens hold most of the money among roughly two thousand holders. Tokenization's promise of open access is real, but it lands on the non-gated side.

United States securities law is part of why: it still restricts these assets to accredited or qualified investors, whatever form the token takes.¹⁴ Tokenization can shrink the ticket. It cannot lower the eligibility bar.

The gated segment lacks openness, not demand

It is young, it is where the regulatory ground is firmest, and it draws genuine demand from institutions, funds, treasuries, and the reserves that back other on-chain instruments. What a gate built for compliance forecloses is the open-market properties, a broad base of small holders and on-chain use.

The central challenge this market faces

The institutions the bull case counts on can enter only through the gated door, and that door opens onto the segment with the least on-chain use. The on-chain substance is in the open tokens, where serious regulated capital will not yet go.

Closing that distance, through compliance that can be enforced without forbidding participation, through enough regulatory clarity for institutions to hold freely transferable tokens, or through technology that lets the compliant and open sides interoperate, is the work this market has yet to do.

Two things have to be true before a balance means anything

You have to know the wallets move together, and you have to know who moves them. Clustering uses on-chain behavior, shared funding and shared signer keys, to group addresses controlled together, without yet saying by whom. Attribution then links that cluster to a named party, using independently known addresses and public disclosures. An attribution is only as good as the cluster beneath it. The two cases below run the method in opposite directions, one showing apparent demand to be wallets the issuer or its affiliates control, the other turning an apparent single whale into a pool of many owners. In both, the aim is the same: to get the number right.

Case one · Ondo and USDY

The concentration reported in the issuer-holdings finding is an observed fact. The identity of the holders is an inference we draw from the on-chain analysis, held with high confidence. This section shows how that attribution is made and lays out each finding so a reader can weigh it independently. The numbered findings are keyed to the badges on the figure.

How to read the on-chain analysis

Every wallet in this analysis is a Gnosis Safe, a smart-contract wallet controlled by signer keys that are recorded on-chain, so two Safes that share a signer are demonstrably linked by a common keyholder. These particular Safes are also generated in bespoke vanity-address families, sharing a first and last hex pattern and a common signer set. Where a Safe shares signer keys with a wallet whose identity is independently established, the link is cryptographic rather than a matter of labeling. Each finding below is tagged as clustering, attribution, or both, ordered to show first that these wallets are one network, then that the network traces back to Ondo or entities affiliated with it.

What the figure shows, read left to right

On the sell side, the ONDO treasury distributes its ONDO governance token through a fan of ONDO routers to six exchanges, about 2.1 billion ONDO over the period.

On the buy side, a single USDC hub, roughly $709 million across the period, provisions a fleet of Safes (the spokes) that subscribe for USDY and OUSG. At the snapshot seven of those spokes held about $414 million of USDY between them, 87 percent of the circulating float, and the fleet held about $88 million of OUSG.

The findings

The numbers match the badges on the figure. Findings 1 to 4 point to these wallets being a single network. Findings 5 to 8 point to that network being Ondo or its affiliates. The seven spoke wallets that held the USDY are listed in full in the appendix. Where a finding cites a smaller number, it applies to that subset of the fleet.

Case two · the largest unattributed wallet

The largest address our attribution could not place at first pass, 0x3700bab28d29f5abc70ecf70a8efd96eed0f9a8f, holds about $706 million, most of it in the three ChinaAMC fund classes shown in the figure. That is the largest balance carried by any single address in this study (larger totals, like Sky’s, span several addresses). Counted as one investor it would be among the market’s largest positions. We assess it instead to be a pooled, or omnibus, custodial wallet: one on-chain address standing in for many underlying owners who hold through shared custody. It is one of at least two such wallets the same operator runs for these Libeara-tokenized ChinaAMC funds, and the largest component of the pooled-custody subtraction in the Finding 01 waterfall, so naming it correctly is what moves this $706 million from apparent concentration to genuine outside demand. It is the mirror of the Ondo case, the same method turning an apparent single holder into many pooled owners, to get the number right rather than to find fault. The determination rests on three independent lines that converge, numbered below the figure.

The findings

We attribute the wallets descriptively to Libeara, the SC Ventures tokenization platform that is the on-chain agent for exactly these tokens, rather than to Standard Chartered, the parent bank and custodian of the funds’ underlying assets, because we have no on-chain proof of which brand operates the keys. We hold the pooled, omnibus structure with high confidence and the precise operator descriptively. We do not know, and do not claim, the number of underlying owners, and we do not describe the $706 million as retail: its largest slice, the institutional Class I fund CUMIU at about $546 million, reads as a small number of large allocations, while the retail subscriber signature is specific to the CAMMF class. On the opening map the wallet appears as the Libeara omnibus cell, and it is counted inside the 83 percent of value attributed to a named holder.

Full address lists for both clusters, together with the limitations of each attribution, are set out in the appendix.

The growth is real, and so is some of what sits behind it

In particular places tokenization does something its traditional form cannot. The on-chain redemption rail works, and where it is used the exit is on demand.

Compliance is genuinely written into the tokens that carry most of the value, and it is enforced. The use of these assets as building blocks inside other on-chain systems is small but growing faster than the market itself. Where this report finds substance, it is concrete and it is checkable.

But the headline runs ahead of the breadth

A large share sits in wallets run by the issuers, their platforms, or entities closely affiliated with them rather than in independent buyers' own hands. Netting that out, the independently held market is about $10.7 billion and the part held directly on-chain by independent owners is about $8.6 billion, well below the $14 billion total.

Most of the value that remains is gated to approved institutions. Most tokenized assets are still bought and held exactly as their off-chain versions would be. And the programmability most clearly in use restricts access rather than widening it. Neither the dismissal that it is all hype nor the verdict that the revolution has arrived survives contact with the holder-level data. The market is real where it is real, and overstated where it is summed up.

What would move the verdict

None of this is visible from a headline number. A total cannot tell an outside buyer from an issuer holding its own product, or a working market from a dormant one, which is why this report works at the level of who holds what and what they do with it.

The signals worth watching are the ones that would widen the substance: whether the open segment's base of small holders keeps growing, whether composability's steep trajectory continues, whether redemption liquidity holds up under stress, and whether compliance and open participation stop being mutually exclusive. If those trends hold, the substance catches up to the story. For now, the distance between the two is the finding.

Financial stability is the largest question we have not priced

Several threads in the data point the same way: collateral looped through lending to multiply exposure, gated tokens reaching open DeFi through freely transferable wrappers, reserves of one on-chain instrument backing another, a few assets turning over more than their entire base each month, and a handful of very large holders who could leave at any moment. In a tightly connected on-chain system, these are the ingredients of systemic fragility. Mapping that connectedness is the next study we intend to publish.

The rest of the distributed market sits outside this report's frame

The same kind of assets live on roughly thirty other public chains: billions of dollars more in holdings, issuers we have not counted, and market dynamics that may look nothing like Ethereum's. All of it is measurable with the same approach, and mapping it is the natural next expansion of this work.

And the holder-level view opens more than we chased

USDY is the only issuer-linked case we presented in full. The market-level netting in Finding 01 already counts other tokens whose supply sits partly in wallets linked to their issuers, and some of that will have entirely ordinary reasons. We worked one case at full evidentiary depth, and separating the ordinary from the notable, issuer by issuer, is work the same method can do.

So is drawing the line between genuine redemption and mere bridging or conversion, and naming the market-makers behind on-chain liquidity. To follow the next installment, leave your email at anagram.digital.

1. Scope and definitions

This report covers tokenized real world assets of the securities type on Ethereum: tokenized treasuries, private credit, funds, equities, and similar instruments. In the distributed market the token itself is the transferable instrument, held in the owner's own wallet. In the represented market the token is only a record on a permissioned ledger while the asset stays in the conventional system. Only the distributed market can be studied on-chain at the holder level, so it is the only market we measure.

We exclude what is not a tokenized security: stablecoins, including those backed by real world assets, passive crypto trusts and exchange-traded products, and tokenized commodities such as gold and silver. The universe is fixed at 418 token contracts as of the snapshot date, 31 March 2026 (a few products deploy the same token across more than one contract and are counted per contract), assembled into a master token list that records each token's platform, asset class, asset profile, supply, and transfer behavior.

Throughout the findings we group tokens by asset profile, a simpler division than asset class. Private markets covers private credit, funds, real estate, venture, and other higher-yield holdings. Money markets covers treasuries and short-term sovereign instruments. Public equity covers tokenized stocks and exchange-traded funds.

2. Data and reproducibility

Two layers sit behind every figure. Balances are observed, read directly from on-chain state at the snapshot block rather than estimated or self-reported. Token balances were read through the Etherscan API, and any reader can reproduce them from the same public state. The USDY token contract at 0x96f6ef951840721adbf46ac996b59e0235cb985c, for example, shows its concentrated holder structure on any block explorer.

Dollar values use each token's net asset value in force on the snapshot date, read by RPC call to the token's price oracle where a usable one exists, otherwise from the issuer's own published NAV and other public sources.

The second layer, the identities behind the balances, is inferred rather than observed, and is described next.

3. Attribution methods and our technology

Matching addresses to the entities behind them is what turns a public ledger into a readable market. Using Anagram Digital's entity-resolution technology and expert judgment, we attribute addresses to issuers, protocols, exchanges, dealers, and a wide range of other on-chain services. Of the value studied here, 83 percent is attributed to a named holder.

Attribution is a judgment, not a direct reading, and we treat it as such. It draws on known-address sets, the behavior and transaction graph of each address, and corroboration across independent signals, and each attribution carries a level of confidence. Where a single attribution carries a finding, as with the Ondo holdings, we set out the specific on-chain evidence and state how confident we are.

4. Holdings, issuer-affiliated classification, and the supply-to-adoption waterfall

Headline supply is not the same as independent demand, so we separate the two. We write issuer for the party operating a token's on-chain program, the asset manager itself or the tokenization platform acting as its agent. Issuer-affiliated wallets are the issuer or its affiliate's own, covering treasury, operational and deployer addresses, unsold inventory, cross-backing between an issuer's own products, and redemption float. They are reported apart from outside holders. Omnibus custody, where one address pools many underlying owners, is treated as genuine demand but flagged, because it hides the number of beneficial holders behind a single cell. Pending-issuance supply, such as USDY still inside its restricted period and shown separately as cUSDY, is counted in gross supply but set aside from free-float findings like the USDY concentration figure.

Netting these out gives the waterfall the findings rely on. A headline of $13.96 billion, less issuer-held supply of about $3.27 billion, leaves roughly $10.69 billion independently held. Removing about $2.08 billion pooled in off-chain custody leaves roughly $8.61 billion held directly on-chain by independent owners. Concentration measures are computed on this netted base, with each pooled account counted as a single holder. The issuer-affiliated figures aggregate attributions across the full issuer and platform universe. This report publishes one issuer-level case in full, the Ondo attribution in Section 6, as the worked example of the evidentiary standard applied to the rest.

5. Transaction restrictions (gated, blocklist, open)

Each token's transfer behavior was tested empirically rather than read from its code. Using Tenderly, we simulated multiple transfers for every token, sending to and from arbitrary addresses and recording which transfers reverted. Empirical testing is more reliable than reading a contract, because what a token does can depend on configuration and state that source code alone does not reveal.

The result sorts each token into one of three models. Gated tokens block any transfer whose recipient is not on an approved list. Blocklist tokens allow transfers to ordinary wallets and block only specific addresses the issuer has listed. Open tokens carry no on-chain restriction at all. Most gated tokens follow the permissioned-token standard ERC-3643.

The open classification was confirmed against real addresses on the OFAC Specially Designated Nationals list rather than synthetic ones. Among the recipients used was 0xa7e5d5a720f06526557c513402f2e6b5fa20b008, an address designated in connection with terrorism financing. For the open tokens these transfers completed in simulation. For blocklist tokens they reverted. Gated tokens rejected any recipient that was not pre-approved.

6. DeFi volume and composability

Composability is an activity, so we measure it as a flow, the value of RWA tokens moving into DeFi, rather than as a balance sitting inside it. A single DeFi interaction usually involves a chain of transfers hopping between several protocols, so simply adding up every transfer into a DeFi address would count the same activity many times over. To avoid that, we count only the first hop, the transfer that moves a token from a non-DeFi wallet into a DeFi protocol.

Transfers between DeFi protocols are not counted, and neither are transfers carrying tokens back out to a non-DeFi wallet, because counting both the way in and the way out would book one interaction twice. For this purpose a non-DeFi wallet is any address not assigned to a DeFi protocol, namely ordinary wallets and multisig wallets. All other smart contracts are treated as DeFi venues. The venue side of this classification is an attribution product of its own. The map behind it resolves 130 named DeFi protocols across 859 contract addresses, built with the same entity-resolution technology as the holder attributions. The default can overcount where an unidentified contract is custodial rather than financial, and we flag that residual rather than assume it away.

Two further exclusions follow the same logic. Assets held as passive backing are not composability and are left out. Transfers between wallets belonging to one entity, along with bridging, are removed as internal movement rather than economic use. A naive count of all on-chain transfers would run several times larger than the figures we report.

The relative measure used in the scatter is a turnover ratio, a token's trailing-three-month DeFi volume divided by its asset base, which shows how intensively each token is used rather than its raw size. Collateral held at centralized venues is reported separately, as a standing balance rather than a flow.

7. Redemptions

On-chain a redemption appears as a burn, a transfer to the zero address, but not every burn is a redemption. Burns also include cross-chain bridge-outs, where a token is destroyed on Ethereum and reissued on another chain, and token conversions, where one version of an asset is swapped for a sibling in the same transaction.

To separate them we inspect each burn transaction. Bridge-outs are identified from a fixed set of cross-chain bridge events in the transaction logs, the events emitted by bridges such as Wormhole, Chainlink CCIP, and LayerZero. Conversions are identified by a sibling token minted in the same transaction. The remaining burns, those with no bridge event and no offsetting mint, are treated as redemptions.

One case resists this test. Where an issuer runs its own custodial burn-and-mint bridge, the Ethereum side of a bridge-out is a bare burn that looks identical to an off-chain redemption, and the two cannot be told apart from Ethereum data alone. We flag this residual rather than assume it away.

8. The Ondo USDY attribution

The full walkthrough of this attribution, with the figure and the numbered findings, is in Section 6. What follows here is the reference material: the complete list of addresses and the limitations of the analysis.

Scope and regulatory context

What we report is the on-chain structure we observe and the inferences we draw from it, under the limits set out above. Questions of motive or intent lie outside what on-chain data can establish, and we do not address them. A separate U.S. Securities and Exchange Commission review of Ondo, opened in 2023 and closed in late 2025 with no enforcement action, concerned the securities-law treatment of Ondo’s tokenized Treasury products and of the ONDO token, as publicly reported. So far as the public record shows, it did not address the on-chain holder concentration described here, and we draw no connection between the two.

Address reference

Full addresses for every labeled node.

Limitations of this analysis

Ondo does not publicly disclose the operator wallets behind its products, so the apparatus is knowable only by on-chain forensics, and the identity attributions are inferences from on-chain artifacts rather than disclosed facts. The BitGo deposit address is a user-specific forwarding address tied to a single BitGo account. Wallets we independently attribute to Ondo deposit into it, which supports the inference that the account is Ondo’s and that the other wallets depositing there belong to the same account holder. The bespoke vanity-Safe construction is likewise specific to this cluster, not a generic custody-vendor artifact. The sell-side path is collapsed for legibility, and the raw network spans about 189 addresses across the distribution fan and exchange deposits. Any proceeds-recycling interpretation cannot be proven through commingled exchange balances, and we do not assert it. We regard the common-control finding as very strong, hold the Ondo-affiliated-identity attribution with high confidence, and keep the economic-ownership question open. The headline figures behind the finding, the roughly 87 percent share of about $476 million in USDY supply, are set out in the finding itself and verifiable from the 0x96f6ef951840721adbf46ac996b59e0235cb985c token's holder list as of 31 March 2026. Of the $537 million gross USDY on Ethereum at the snapshot, about $61 million sat bridge-locked as backing for other chains, leaving the $476 million float.

9. Reliance and limitations

What this report rests on is public Ethereum state, which anyone can read and reproduce, named price and NAV sources, and Anagram Digital's own attribution layer over the raw data.

The limitations follow from that. Attribution is an inference, not a fact written on the chain, and about 17 percent of the value studied is unattributed. Dollar totals depend on NAV, which is oracle-fed or issuer-published, so they value holdings at stated marks rather than at what a secondary market would pay, a distinction Finding 03 shows can be large for illiquid tokens. Pooled and omnibus accounts hide the number of beneficial owners behind them. Some attributions are provisional, and the line between an issuer and an affiliated entity is a matter of judgment. The view is a single point in time, the snapshot date, and it covers Ethereum only, so assets and activity on other chains sit outside the frame. The securities status of some tokens is genuinely contested rather than settled, which is why the compliance finding is put conditionally. Simulated transfers can have edge cases that a single test would miss, which is why several were run for each token. Where any of these could change a conclusion, we say so in the finding itself. This is independent research. Neither the author, the contributors, nor Anagram Digital holds a position in any in-scope token or in ONDO, nor any direct position in any company named in this report. No named party commissioned or paid for it, and nothing in it is investment advice.

10. Glossary

Distributed versus represented

A distributed token is the transferable instrument itself, held in the owner's wallet. A represented token is a record on a permissioned ledger while the asset stays in the conventional system. This report measures the distributed market.

Gated, blocklist, open

The three transfer models. Gated allows only approved holders, blocklist blocks only listed addresses, open applies no on-chain restriction.

Issuer-affiliated versus omnibus

Issuer-affiliated wallets are the issuer or issuer affiliate's own holdings. Issuer covers the token program's operator, whether the asset manager or its tokenization platform. An omnibus wallet is one address that pools many underlying owners.

Composability

The use of a token as a building block inside other on-chain systems, such as collateral, lending, or automated strategies.

Turnover ratio

A token's DeFi volume over a period divided by its asset base, a measure of how intensively it is used.

NAV. Net asset value, the per-token value of the underlying asset, used to convert token balances to dollars.

Observed versus inferred

Balances are observed directly on-chain. Identities are inferred through attribution.

Asset profile

The grouping used across the findings: private markets, money markets, and public equity.