The Institutional Rail Map

What the numbers say once you measure them yourself

Every figure below was pulled or measured by hand on 5 August 2026.

Institutional money is moving onchain, and the figures circulating about it are softer than they look. Some come from aggregators that do not cover the same universe. Some come from articles quoting other articles. And the single most discussed deployment of the year turns out to be measurable in one request, by anyone, at a size almost nobody states out loud.

This report is the sequel to our July study of which listed companies genuinely operate onchain. That one asked who operates. This one asks on which rails they land, and how much is actually there.

This is edition 1 of a quarterly series. We will re-measure the same axes with the same method every quarter, next in November 2026. That is the point of the exercise: a single snapshot tells you where capital sits today, a series tells you where it is going. The trajectory data in section 6 comes from public history; from this edition onward, the series is our own.

1. What we measured, and how

We pulled the full protocol and chain universe from DefiLlama on 5 August 2026: 7,990 protocols across 461 chains, of which 153 are classified as real-world assets. We aggregated four axes per chain: decentralized finance value locked, real-world asset value, stablecoin supply, and fees over 30 days, plus the 90 and 180 day history for trajectory.

Where an aggregator was silent or ambiguous, we went to the contract or to the issuer rather than to a secondary article. That distinction matters more than it sounds, and section 2 is the reason why.

Three methodological notes.

First, chain names are not stable across endpoints of the same source. The same chain appears as Binance in one place and BSC in another, as Ripple in one and XRPL in another, as Optimism in one and OP Mainnet in another. An analyst who aggregates without normalizing these aliases silently produces zeros for two of the largest real-world asset chains. We normalized them.

Second, we distinguish throughout between a measured fact, a labelled proxy, and a hypothesis. Stablecoin supply is used as a proxy for settlement volume, never as settlement volume itself, because no free source publishes the latter per chain.

Third, and this is a rule we apply to ourselves: we do not state that something is absent from the market on the strength of a single source. We state that it is absent from a named source, on a named date. The two are not the same claim, and the first one is usually wrong.

2. JPMorgan on Base: one million dollars

The tokenized deposit that JPMorgan launched on Base, JPMD, is the most discussed institutional deployment of the year. It is also, unusually, one that anyone can measure exactly.

The contract address is published by JPMorgan's own Kinexys content hub. Calling totalSupply() on it returns a raw value of 0x5f767a0, with two decimals and the symbol JPMD. One deposit token equals one dollar.

Total outstanding: 1,001,000 dollars.

One million. Not a rounding of a larger number, not an estimate, not a range. That is the entire supply of the instrument on 5 August 2026, and a public block explorer displays the same figure independently.

We state this plainly because the gap between the volume of commentary and the size of the position is itself the finding. A proof of concept from a systemically important bank on a public chain is a genuine event. It is not yet a flow.

What this changes for your allocation. Treat announced institutional deployments and deployed institutional capital as two separate datasets. The first is public and abundant. The second usually requires one request to a contract, and almost nobody makes it.

3. The aggregators do not cover the same universe

Franklin Templeton's BENJI is one of the two reference tokenized money market funds. It carries 720.8 million dollars on 5 August 2026, distributed across eight chains.

That distribution is the interesting part. Two thirds of one of the largest tokenized funds in the world sits on Stellar, not on Ethereum, and this is rarely reflected in how the chain hierarchy gets discussed.

The distribution is also invisible on the dashboard most people open. BENJI carries no entry in DefiLlama's real-world asset category as of 5 August 2026, while rwa.xyz tracks it in full. Neither source is wrong; they cover different universes. But a chain ranking built on one of them and quoted as though it described the market will be off by hundreds of millions.

We ran into the same problem from the other direction while preparing this report. A widely syndicated article put BENJI at 2.45 billion dollars in July. The issuer-facing tracker says 720.8 million today. We use the tracker.

Then there is the case that inverts the exercise entirely. Canton, the network built for institutions, is the largest chain in the world by fees collected: 55.2 million dollars over 30 days, and 540 million over the trailing year. On the decentralized finance axis it shows 198 thousand dollars of value locked, and it carries no real-world asset line and no stablecoin line at all.

We wrote about Canton in June, when Apollo, BNP Paribas, Citadel Securities, HSBC, S&P Global and the CME funded the network. Our reading then was that these institutions were no longer testing the technology but financing the settlement of their own transactions. The fee data is consistent with that reading. The reserve we published at the same time still stands: what is settled in volume, as opposed to what is paid in network fees, is not public.

A necessary caution on the fee axis. These figures are measured but not comparable to each other. Canton's fee methodology aggregates traffic purchases, pre-approval burns, setup burns and holding fees, which is not gas. Ethereum's figure includes blob fees, so part of its layer 2 activity is already counted in the layer 1 number. And Robinhood Chain shows a 30 day change above 18,000 percent for a simple reason: essentially its entire lifetime fee history, 3.79 million dollars out of 3.81 million, falls inside the current window. That is a launch artefact, not a trend. Any league table scoring these numbers against each other without saying so is comparing incompatible units.

What this changes for your allocation. Before comparing two institutional figures, establish which source produced each one and what it covers. In this dataset, they routinely disagree by an order of magnitude.

4. The matrix: which institutional product sits on which chain

The most structural fact in this table: BlackRock's BUIDL now holds only 34 percent of its assets on Ethereum. Two thirds sit on Aptos, Solana, Avalanche and four other chains. The flagship product of institutional tokenization has been multichain for some time, while most infrastructure commentary still reasons as though Ethereum were the only rail that counts.

One reserve we published in July applies directly here, and we restate it rather than quietly drop it. Real-world asset market share per chain is a real positioning metric, but a volatile one: a single ticket is enough to flip it. A chain that gains four hundred million dollars of share in a week has not gained an ecosystem, it has gained one allocator's decision. Read this table as a photograph of where capital sits, not as a measure of durable adoption.

A line we flag rather than hide. Circle's USYC shows 2.92 billion dollars on BSC. That single line is 11 percent of all measured real-world asset value and by itself gives BSC its second place. We could not confirm it against a primary source in the time available. We publish it because a figure of that weight cannot be silently removed, and we label it unverified because it is. Treat the BSC ranking as provisional.

5. The gold trap: 18 percent of "real-world assets" is metal

Real-world asset league tables get quoted constantly. Almost none state what is inside the category.

Of the 26.78 billion dollars DefiLlama classifies as real-world assets, 4.81 billion is tokenized gold: Tether Gold at 2.90 billion, Paxos Gold at 1.80 billion, and roughly a dozen smaller issuers. Gold has nothing to do with the tokenization of funds, treasuries, private credit or bonds. And it is concentrated on one chain: 4.68 billion of Ethereum's 13.60 billion total is metal.

Removing it changes the ranking materially.

Ethereum loses ten points of share. Every other chain gains. Neither figure is wrong; they answer different questions. What is wrong is quoting one without saying which category definition produced it.

6. Five moves this quarter

The market fell broadly over the window, so relative share carries more information than absolute levels. Figures are 90 day changes to 5 August 2026.

Stellar and XRPL roughly doubled their stablecoin supply, up 127.9 percent and 153.0 percent respectively, from a small base. On Stellar the real-world asset data corroborates the move from three independent products: Spiko holds 1.38 billion dollars there, Ondo 532 million, and BENJI 475 million. Both networks have long been positioned as institutional payment rails, and this is the first window in which that positioning shows up in measurable supply rather than in announcements. We hold a published reserve on both from July, when the Swift ledger moved onto ground these rails had claimed; this data sits alongside that reserve rather than cancelling it.

BUIDL went multichain, with Aptos as its second rail at 822 million dollars.

Aptos lost 77 percent of its decentralized finance value locked while retaining 857 million dollars of real-world assets. Speculative capital left, institutional capital stayed. It is the clearest available demonstration that these two axes measure different things and should never be collapsed into one score.

Base held on both axes, down 1.2 percent on value locked and 2.0 percent on stablecoin supply, while comparable chains fell between 10 and 35 percent. Its fees fell 21.1 percent over the same window.

Canton established itself at the top of the fee table, on the caveat set out in section 3.

7. What this changes for an allocator

Exposure. The rail creates value; whether the token captures it is a separate question, and for most chains above it is unresolved. This report deliberately stops short of answering it. A defensible answer requires a value capture framework comparing burn, staking, fee share and sequencer revenue on a like for like basis, and we do not consider such a comparison ready until it does.

Infrastructure. If your settlement, custody or distribution roadmap assumes Ethereum as the default institutional rail, two figures challenge that assumption directly: BUIDL at 34 percent on Ethereum, and BENJI at 66 percent on Stellar.

Framework. Permissioned rails such as Canton, and bank-operated instruments such as JPMD, are measured very differently from permissionless ones, when they are measured at all. Assurance about them comes from the operator rather than from an independent observer. Whether that is acceptable is an allocation policy question, not a technical one.

Competition. Ten institutions in the table above already distribute a tokenized product on at least three chains. Multichain distribution is the current standard among the largest issuers, not an edge case.

8. What this report does not say

It does not rank chains. A ranking would require the fee axis to be comparable, and section 3 shows it is not.

It does not measure developer activity. No free interface publishes it, and we would rather leave the axis empty than fill it with a stale proxy.

It does not measure settlement volume. Stablecoin supply is used as a labelled proxy throughout.

It does not measure the history of real-world assets per chain. DefiLlama keeps history for value locked and stablecoins, not for the real-world asset breakdown. Our own series starts with this snapshot and will build with each edition.

It contains no buy signal, no price target and no time horizon, and it takes no position on any token.

Talk to us

This is the public layer of a longer piece of work: the full institutional rail map, including the primary sourcing of the instruments the aggregators do not carry, and the value capture framework referenced in section 7.

Edition 2 lands in November 2026, with the same axes, the same method, and the first quarter-on-quarter deltas measured by us rather than inherited from public history.

If you allocate capital and this is the level of reading you need, contact us directly. We will send you the complete report and put you on the list for the next edition.

These analyses are drawn from the Crypto Deep Research desk. Data measured on 5 August 2026: DefiLlama for chain-level aggregates, rwa.xyz for BENJI, and a direct contract call on Base for JPMD. Facts, labelled proxies and hypotheses are distinguished throughout.

Crypto Deep Research, a fundamental research house. Nothing in these lines constitutes personalized investment advice or a recommendation to buy or sell.