XRP evidence map 2026: 71 milestones, only 29 measurable

Of the 71 dated milestones for XRP, Ripple, and XRPL that we have classified by level of evidence, 29 reach measurable production; and among the 41 that are corporate agreements or operations, only 6 do so, with none currently being a bank or remittance firm that publishes figures from its own agreement with Ripple. The only one to publish them was MoneyGram, and it did so because it was required to disclose them to the U.S. Securities and Exchange Commission in its accounts: $61.5 million gross received from Ripple in 2019 and 2020 —our sum of two 8-K filings— and the agreement terminated on March 7, 2021. The peak levels of 2026 are of a different nature: court filings, regulator registries, exchange-traded funds reporting to their supervisor, and issuances readable on the ledger. Things that a third party is legally obligated to publish. Meanwhile, the network measures itself, and what it measures points in two directions: approximately 2.70 million transactions estimated by sampling in the 24 hours prior to September 27, 2026, more than the 1.59 million on Ethereum L1, at a cost per transaction 353 times lower; and a TVL (value deposited in DeFi protocols on the network) of $46.1 million that has not moved since April. This map puts both lists on the same table, showing what can be proven, what is merely announced, and what cannot be measured.

Update from October 10, 2026: the two amendments that the table had listed as pending have activated, and within the expected timeframe. PermissionDelegationV1_1 (XLS-75) entered ledger 107,524,865, closed on October 8, 2026, at 21:29:50 UTC, with 28 out of 33 validators above a threshold of 26. BatchV1_1 (XLS-56) entered at 107,540,993, closed on the 9th at 14:47:52 UTC, with 27 out of 32 above a threshold of 25, and fixBatchV1_2 preceded it that same afternoon at 14:15:21. The Amendments object of the ledger now records 97 amendments in force and only one pending majority. Activation is distinguished from a majority within the ledger itself: each of the three is an EnableAmendment transaction without flags, whereas one with tfGotMajority would only announce that it has just gathered the votes.

What can be proven today about XRP and what is only announced?

The scale, declared before use: A is activity in production and measurable (on-chain reading, figures published by the counterparty, accounts presented to a regulator, searchable public registry), B is a signed agreement with a joint document from both parties, C is a pilot or proof of concept declared as such, D is a memorandum of understanding or letter of intent, and E is a single-party announcement without confirmation from the counterparty. An amendment under vote is not A, nor is a devnet, nor a requested license. And a figure does not move up a level simply by being large. Of the 71 milestones, 41 are corporate agreements or operations spanning all five levels; the remaining 30 are regulatory, judicial, or related to exchange-traded products.

The same entity-to-entity classification, ordered by start date from oldest to most recent, including who documents each item, which is the column that determines the level:

The distribution follows a recognizable pattern: in 2026, what produces a level A is a legal obligation to publish. The SEC vs. Ripple Labs case is A because the filings are in the court registry. The European crypto-asset service license is A because Ripple Payments Europe S.A. appears in the ESMA registry with its Legal Entity Identifier. Exchange-traded funds are A because they present accounts to their supervisor, and REX-Osprey's XRPR and the four ETPs in Europe, Canada, and Brazil publish assets as of September 24 and 25. And Deloitte's attestation on RLUSD reserves is A because the firm puts its name to it: 1,589,084,487 RLUSD in circulation on July 6, 2026, and 1,463,694,758 on the 31st, against reserves of $1,688.5 and $1,575.2 million. The August attestation has not been published.

Hence the two qualifiers in the count, "today" and "from its own agreement": two of the six corporate A levels are remittance firms that published figures in the past but have not done so for years: Tranglo, $48 million in 100 days between 2021 and 2023, and SBI Remit, 2.5 billion yen accumulated as of May 29, 2026, without naming Ripple or XRP or breaking down the attributable portion.

At the other extreme are the level E milestones, and it is worth saying which universe is being described. Across the full inventory of 71 milestones they cluster in the most recent and the largest. Within the entity-by-entity census they are the three largest announcements —the March 2026 stablecoin platform, the eleven Brazilian names, and the Rail acquisition— and none of the three is the most recent entry in it: the two newest, Absa CIB on September 21 and the absence of RLUSD from the e-money register on September 24, are B and A. The cleanest evidence works in reverse: on Ripple's customer page, read on September 27, 2026, seven names appear —Tranglo, Nium, Modulr, Sentbe, Instarem, SBI Remit, and Travelex Bank—. That same page, as of September 27, 2026, does not include Santander, MoneyGram, Zand, Mamo, Bitso, Onafriq, Standard Chartered, PNC, American Express, Siam Commercial Bank, National Bank of Egypt, Deutsche Bank, or Western Union. The absence of a partner does not prove the agreement has ended; it is, however, a statement from Ripple regarding whom it presents as a customer today.

How is XRPL today, measured on-chain?

Everything that follows was read against a fixed ledger via public RPC —s1.ripple.com:51234, clio 2.8.0 on rippled 3.4.1— and not from a dashboard: fixing the ledger allows the reading to be repeated in six months to obtain the same figure to the decimal.

Two of these figures cross-reference what Ripple declares on its XRP page as of June 30, 2026. Ripple reported 32.6 billion XRP in escrow, and our measurement on September 27 gives 31.7 billion: 900 million less in about 2.9 months, matching the net reduction rate of about 320 million per month. Of the XRP outside of escrow, Ripple declares 5,056,053,914, and the 84 accounts in the label census total 3,610,289,387.77, so our figure is a floor and we know by how much. The labels are from xrpscan, a third party; the figures are on-chain. Hence the phrase "accounts labeled Ripple," with 1,445,764,526 XRP outside the census.

Why do micropayments with x402 memo dominate XRPL payments?

In the contiguous window of 200 ledgers on September 27, 2026 (12.98 minutes), there were 35,669 transactions, and 5,998 of them carried an x402 memo: the tag for the scheme that repurposes the HTTP 402 code —"Payment Required"— so that a program pays for each call to a service instead of signing a subscription. Of the 5,110 payments delivered in RLUSD in that window, 5,088 are of this class, or 99.57%. Together they moved 5.09 RLUSD. Five dollars and nine cents, spread across more than five thousand payments.

This detail reorders the reading of the rest. In the 111 sampled ledgers over 24 hours, there were more payments delivered in RLUSD (2,701) than in XRP (1,803), which by count would make Ripple's stablecoin the dominant payment asset of the ledger. By amounts, the conclusion is reversed: the median of an RLUSD payment is 0.001 RLUSD, while the 3,959 XRP payments in the window delivered 10,983,904.62 XRP, and a single payment of 9,508,988.89 XRP accounts for 86.57% of that total. With the count set by automated agents and the volume by an isolated transaction, any "average amount per payment" without the median alongside it distorts both sides.

There is a limit that affects everything else: on-chain, a customer payment cannot be distinguished from a treasury movement. One can describe the form of the payments —size, median, memos, 1,174 sending accounts, and 900 receiving accounts in those 12.98 minutes— but not the intent. Nothing measured here confirms that a volume is "real use," nor the opposite.

How does XRPL stand against Ethereum, Solana, and Tron?

The comparison uses the same calls and the same time for all four networks. The last column indicates if the row allows for comparison: four of the six carry a caveat, and those caveats are part of the result.

The two most clashing readings are true at the same time. XRPL moves more transactions per day than Ethereum L1 —2.70 million gross and 2.05 million successful versus 1.59 million— at a cost per transaction 353 times lower. The Ethereum cost is a floor, so the real multiple is higher; in the other direction, an Ethereum transaction can bundle a rollup batch, so the count is not the same unit. At the same time, XRPL's TVL is 0.086% of Ethereum's and 0.69% of Solana's, remaining where the April 2026 piece left it: about $46 million then, $46,112,232 today. And there is a derived metric that better ranks the four: TVL per dollar of stablecoin on the network, 0.039 on XRPL, 0.060 on Tron, 0.367 on Ethereum, and 0.408 on Solana. XRPL and Tron resemble each other —transfer networks with little protocol layer on top— and there, among transfer networks, XRPL's stablecoin size is 1.25% of Tron's, which we already wrote about in our analysis of its infrastructure in emerging markets.

The cost row combines two measurements and two approximations, and the difference matters when citing it. XRPL and Ethereum were measured on September 27: fees actually paid divided by transactions in the first case, a floor modelled on the base fee in the second. Solana and Tron were approximated on September 28, each for a different reason.

On Solana, the September 27 obstacle turned out not to be where we had noted it. We recorded it as no public RPC allowing block reading, but the actual error was different: the server rejected the request because it was not asked to support version 1 transactions, which is a call parameter and not a provider limit. Once corrected, 16 finalized blocks yielded 71 simple SOL transfers, and the median fee paid was $0.000595. Nearly half of them —49.3%— paid exactly the base fee per signature, without any priority tip. Measured the way XRPL was measured —fees paid divided by transactions, consensus votes excluded— the figure rises to $0.002709, because that brings in the swaps and mints that do bid for priority.

In Tron, there are two cost regimes rather than one, which is why the cell contains two figures. The protocol charges in resources, not in currency: bandwidth for a simple transfer, energy for a smart contract. Those who have staked TRX receive these resources and their transaction burns nothing; those who do not must pay the equivalent in TRX. At the resource prices the node itself publishes, the median consumption of 1,492 simple transfers works out to 0.265 TRX: $0.0883 without staked resources, zero with them. Which of the two is the norm is answered by the sample: out of 4,977 receipts, 4,353 were issued with zero fee, or 87.46%. A USDT transfer is another order of magnitude because it consumes energy rather than bandwidth: $2.26, roughly 26 times the simple transfer. The cell carries the simple transfer because it is the one comparable to the other three networks.

The context is missing, without which the TVL row is misleading. Real-world assets tokenized on XRPL are not included in those 46.1 million dollars, because the aggregator does not count a tokenized issuance as value deposited in a protocol. According to rwa.xyz, there are 501.15 million dollars distributed on XRPL across 274 real-world asset holders: Ondo at 189.9 million —190.4 according to our on-chain reading of 1,627,043.03 OUSG—, VERT Capital at 116.11; Archax at 55.4; Franklin Templeton's sgBENJI at 48.3 across ten holders and Guggenheim's paper via Zeconomy at 40.36 across four holders, with zero monthly volume. The XRPL asset layer is ten times its TVL and measures something else: liquidity in protocols versus tokenized issuance. Providing only one of the two distorts the picture in either direction. Concentration is extreme —in the XRPL OUSG, three accounts hold more than 99.99% and a single one holds 72.6%— and the Ondo series on XRPL is trending downward, from 323 million in May 2026 to 212.6 in August and 189.9 in September.

How long does it take for a transaction to become irreversible on XRPL, Ethereum, Solana, and Tron?

Chain speed is almost always cited with a single number, and that number is usually the inclusion time: how often a block closes. This is only half the story. Inclusion means the network has already placed the transaction into a block, slot, or ledger and treats it as accepted. Irreversibility means it can no longer be undone by a chain reorganization. Between these two, there are networks where no difference exists and others where the factor is 64, so they go in two columns: a single number ends up comparing the inclusion of one chain with the irreversibility of another.

The order of the four changes depending on the column being read, and it is this result that justifies separating them. Solana includes 14.5 times faster than XRPL —0.268 seconds measured against 3.89— and takes 2.1 times longer to become irreversible: 8.3 seconds against 3.89. Ethereum is the slowest in both, but the gap widens in the second: 3.1 times slower than XRPL in inclusion and 197 times slower in irreversibility. Anyone comparing "0.4 seconds for Solana against 3 to 5 seconds for XRPL" is pitting the inclusion of one against the approximate irreversibility of the other, and the result is reversed when both columns are measured equally.

The fourth column is the ratio between the previous two and carries a specific meaning: how many confirmations each protocol requires before treating something as final. It is derived by dividing two measured times and aligns with what each network documents in writing, which serves as verification that the measurement is clean. In Ethereum, it is 64 slots, the two epochs of Casper finality. In Solana, 31 slots measured against the 32 vote confirmations it documents to root a block. In Tron, 19 blocks, which represent two-thirds plus one of its 27 super representatives. In XRPL, it is 1, because its consensus does not accumulate confirmations: the validated ledger is final at the moment of closing, and what we measure as the closing is already finality.

Two warnings that the figures do not inherently show. The first is that these columns measure what the protocol guarantees, not what an intermediary requires: an exchange may request twenty confirmations where the chain considers one sufficient, and thus the time the user experiences is set by the custody policy. The second comes from Solana's measurement: its slot step, 0.268 seconds, is below the 400 milliseconds it publishes as a target, 33% less time per slot than its own parameter. That explains why its measured irreversibility stands at 8.3 seconds rather than the 12.8 that 32 slots at 400 milliseconds would yield.

Where do the $511,088 in fees attributed to XRPL come from?

Two fee figures circulate for XRPL, and between them is a factor of 1,830. Neither is wrong: they measure different things and only one is called network fees. DeFiLlama's fees-by-chain endpoint returns $511,088 in 24 hours, which are the application fees attributed to the chain, and $501,022 of those dollars —98%— are from THORChain, a cross-chain protocol. XRPL's own line in the same endpoint is $429, in the order of the $279.32 we measured on-chain by counting burned XRP: 1.643044 XRP in the window, 182.23 per day. Anyone citing "$511,000 in fees on XRPL" is citing the revenue of a multi-chain protocol; the $279 are what the network burns to process 2.7 million transactions.

The same happens with the word TVL —$46,112,232 via the chain endpoint, $42,134,244 summing DeFi protocols that declare XRPL, and another $138,650,592 from Doppler Finance excluded for being classified as CeDeFi—: all three are correct, and what is missing when citing them is the perimeter, the same problem we broke down in how to read DeFiLlama open interest. The stablecoin census fails in both directions: DeFiLlama includes $72,771,000 of USDV, whose issuer we could not locate, and does not list the 109,064,033.71 BBRL or the 33,365,506.33 USDB from Braza that we did measure. SG-FORGE's EURCV comes out to 10,000,000.00 in our reading and 11,377,398.44 in the aggregator, unresolved. And the aggregator contradicts itself with Ondo and OpenEden, which it lists with activity on XRPL while their tokens declare only Ethereum: their figures are taken on-chain.

Where is the XRP and why does its native credit stay at $59,203?

In exchanges, and by a margin of 135 to 1. In the key that DeFiLlama uses for XRPL, most of what is attached are balances from centralized exchanges: $5.702 billion, compared to $42.1 million in native DeFi. That is 135 times more value under the custody of a third party than in protocols where the user keeps the key, and the demand for price exposure that feeds that custody was covered in the flow analysis of XRP and Solana ETFs. Within those $42.1 million, 97.8% is the ledger's own DEX, the order book that lives inside the protocol: $41.2 million. Lending: $59,203; yield: $8,861; the rest, about $832,000, is distributed among protocols the aggregator does not break down.

The $59,203 loan comes from Templar Protocol, a third-party protocol: the ledger's own protocol is not enabled. As of September 27, 2026, the mainnet node shows LendingProtocol with enabled=false; xrpscan scrutiny adds 15 votes out of 35 validators, while the activation threshold is 28 (80%). SingleAssetVault, the single-asset vault that serves as the building block for credit, stands at 19 out of 35, and LendingProtocolV1_1, added in rippled 3.4.0 on September 17, is at zero; its specification was still in draft form on September 15. All three function on devnet and none on testnet, so native lending on mainnet is zero by design. Furthermore, the $138.65 million from Doppler Finance, the largest figure attributed to XRPL, is the one that least fits the credit label: yield with custody in Fireblocks and Ceffu, with zero audits counted by the aggregator and with its documentation stating that on-chain strategies are "not currently deployed." We have been unable to read their transparency reports, so the figure is included as reported by the aggregator and without a credit label.

Which XRPL amendments remain unactivated on mainnet?

Lending is not the only missing building block, and this can be verified without believing anyone: the identifier for each amendment is the SHA512Half of its name, so "not enabled" is verified locally. Votes are from xrpscan; the enabled status is from the node.

The last two rows establish caution: an amendment with a majority is not yet active, and two that were active were disabled after a bug and restarted the clock.

What do SEC filings, the ESMA registry, and the OCC say about Ripple?

The fine was paid in full. The final judgment of August 7, 2024, imposed $125,035,150 plus an injunction against repeating the Section 5 conduct, compared to the nearly $2 billion the Securities and Exchange Commission had requested. On May 8, 2025, both parties agreed to lower it to $50 million, and Judge Torres denied that agreement twice: on May 15 as "procedurally improper" and on June 26 for lack of "exceptional circumstances." On August 18, 2025, in a letter to the court, Ripple's lawyers communicated the full payment of the penalty plus legal interest and the return of the balance of the escrow deposit the company had established at 111% —about $138.8 million, according to our calculation on that percentage—. On August 22, the Second Circuit issued the mandate that closed the appeal and the case. The version that Ripple "settled for 50 million" still circulates and is false.

The joint interpretation published by the Securities and Exchange Commission and the Commodity Futures Trading Commission on March 17, 2026, effective the 23rd (91 FR 13714), cites XRP as an example of a "digital commodity" in its footnote 51. It is a real and verifiable document, but it is an interpretation: the text itself states it does not substitute or replace the Howey test. The law that would do so, the CLARITY Act, fell one vote short of opening debate in the Senate on September 15, 2026 (cloture 49-50), as we analyzed in the roll call count of that vote.

In Europe, there are two facts that should not be conflated. The service is authorized: Ripple Payments Europe S.A. appears in the ESMA crypto-asset service provider registry with a full license under MiCA for 29 states and four service entries covering exchange for money, exchange for other crypto-assets, and transfers, and does not include custody. The token is not: the ESMA e-money token issuer registry, in its September 24, 2026 version, has 49 entries from 23 issuers, and RLUSD does not appear in any of them. A negative data point, verifiable by downloading the file: Ripple's stablecoin is not authorized as an e-money token in the European Union as of that date. That same registry, read on July 31, supported our count of legal stablecoins in the European Union. The full chronology of RLUSD licenses by jurisdiction, which we do not repeat here, is in our tracking of Ripple's multi-jurisdictional stablecoin.

And in the United States, the bank is not operational. The Office of the Comptroller of the Currency's approval for Ripple National Trust Bank, dated December 12, 2025 (letter CA 1359), is preliminary and conditional: it sets a minimum Tier 1 capital of $11.7 million, requires it to be established before December 12, 2026, and to open before June 12, 2027. Until that happens, "Ripple has a national bank" describes a filing, not a functioning entity.

What has changed in XRP since April 2026, and what remains unmeasurable?

This piece does not re-discuss the thesis we published on April 28, 2026, in XRP's promise vs. its reality —that the vision was correct and others are executing it—. What it does is measure where each figure stands five months later, including the three we published incorrectly and have since corrected there. Neither Deutsche Bank nor BlackRock have ever documented what we attributed to them: for the former, there is only one German media report, without a note from the bank or Ripple; for the latter, what exists is a swap of BUIDL for RLUSD announced by Ripple and Securitize in September 2025, not a use of RLUSD as collateral by BlackRock. And the tokenized assets we assigned to Archax were the XRPL total from all issuers.

Alongside these corrections is an upward nuance: the Palau digital currency pilot, which in April we grouped among projects "none in production," has live and measurable issuance on the ledger, at a scale of 14,588 units. The RLUSD market cap figures we have published since February range from $1.387 to $2.491 billion, each with its date and chain perimeter; the April piece already carries the 2,491 and keeps its 1,560 as a historical anchor. And the 2,491 total sums two chains when the Wormhole integration reaches more than forty, so it is also a floor.

What remains is what cannot be measured, which is part of the portrait. Ripple Payments volume —$70 billion annualized in May 2025, $95 billion in December, and $100 billion in March 2026, with markets dropping from "over 90" to "over 60"— is self-reported and cannot be distinguished on-chain from any other transfer: it does not move up a level by being repeated or by growing. The cost per transaction for Solana and Tron could not be measured with public RPCs and is not compared. The total number of XRPL accounts would require traversing the entire ledger, so active accounts are provided. And regarding network performance, what was observed on September 27, 2026, remains: 45.79 gross transactions per second and 29.04 successful. Faced with the next "a bank adopts XRP," the test that organizes this map and the Solana institutional payment adoption map consists of four questions in this order:

1. Is there a document from the counterparty, and not just from the technology vendor?
2. Does it provide a figure with a cutoff date, or only an accumulated volume without breakdown?
3. Is it in a searchable public registry or in accounts presented to a supervisor?
4. Can it be read on the ledger, account by account, without believing either party?

If the answer to all four is no, the announcement is level E, no matter how large the figure.

Sources and links: Final Judgment of Aug-7-2024, ECF 974 (CourtListener) · Letter of Aug-18-2025 with full payment of fine, ECF 992 · MoneyGram, Q1 2021 10-Q (Note 16, termination of Ripple agreement) · MoneyGram, 2020 8-K ($50.2 and $11.3 million from Ripple) · OCC, Preliminary and Conditional Approval CA 1359 · ESMA, E-Money Token Issuer Registry (Sep-24-2026 version) · Ripple, Deloitte Attestation on RLUSD Reserves · Ripple, Customer Page (read Sep-27-2026) · Ripple, Declared XRP and Escrow Figures as of Jun-30-2026 · rippled, Pull Request 8255 to withdraw XChainBridge · RippleX, Recommendation to withdraw XLS-38 · xrpl.org, xrpld version 3.4.1 release notes (Sep-25-2026) · xrpl.org, How amendment voting works · DeFiLlama, TVL by chain (endpoint consulted) · DeFiLlama, Application fees attributed to XRPL · rwa.xyz, Real-world assets on XRP Ledger · SG-FORGE, EURCV on the XRP Ledger · XRPL EVM, Ethereum-compatible sidechain