Four waves of RWA, 2017 to 2026

Wave one: security tokens raised capital and built no market

ICO is short for initial coin offering, a sale of tokens with no basis in securities law, and by 2018 those sales were falling under enforcement. STO, security token offering, was the industry's answer: the same fundraising run as a regulated offering of an instrument that is a security under the law, a share or a bond or a certificate. Blockchain Capital's BCAP, sold on 10 April 2017, is the first one on the record.

The money that came in was the first sign that the market had not bought the product.

The best year of the whole STO era, 2019, raised about 8% of what ICOs raised in 2017 alone, and it did so carrying a story about regulatory compliance and institutional money that ICOs never had.

Capital raised is the weaker evidence, though. Secondary trading is the buying and selling that happens between investors after an offering closes, and it was the entire promise of tokenizing an illiquid asset. For security tokens it started in earnest when tZERO opened trading on 24 January 2019, which makes 2019 the first year with a full set of data. Every security token in the world traded 2,410,608 USD that year, an average of 7,156 USD a day.

The largest offering of the wave shows what an investor actually received for that money. tZERO, a subsidiary of Overstock.com, had to report as part of a public company, which is why its deal is the best documented of the era.

ATS stands for alternative trading system, a venue registered with the SEC for trading securities away from an exchange, and for a US security token it was the only lawful route to a secondary market. PRO Securities is the ATS in that last row. The buyer of the wave's biggest offering got a token that lived on a public blockchain and sat in the issuer's wallet for a quarter, and that could then be resold only to another accredited investor. Every restriction of a private placement was preserved, and the cost and technical risk of a blockchain were added on top.

What actually killed wave one

Two explanations circulate, and the wave's own data falsifies both.

The first is the bear market of 2018 and 2019. It did not help, and it does not account for what happened two years later.

The second explanation is regulatory hostility. On 28 January 2026 three divisions of the SEC issued a joint Statement on Tokenized Securities saying that a tokenized security remains the same security under existing federal law, whether the record of ownership sits on a cryptographic network or off it. The statement grants no relief and creates no new regime. Eight years after the wave, the regulator confirmed in writing the rule that applied during it.

The constraint that followed from that rule was arithmetic rather than hostile. An offering sold under Regulation D, the US safe harbour that exempts a private placement from registration, produces restricted securities, and restricted securities cannot be resold to the broad market. So the natural buyer of a wave-one token was another accredited investor, and an accredited investor who wanted that asset could buy it more cheaply and more simply through an ordinary private syndication. The issuer paid a liquidity premium and received no liquidity.

Harbor's deal is the cleanest failure of the wave, and it has nothing to do with trading volumes. The company tokenized a student housing block in Columbia, South Carolina, the Hub at Columbia: 260 units, a deal of about 20 mn USD, cut into 955 tokens of 21,000 USD each. On 11 April 2019 it collapsed, because the property's existing mortgage lender would not agree to the change in ownership structure.

No regulator refused it, no investors failed to show up, and no contract broke. The legal chain of any RWA runs asset, then the legal entity that owns the asset, then equity in that entity, then the token, then the holder. Harbor built the last two links well and the first one vetoed the deal. An asset carrying debt, a covenant or a change-of-control clause has a party in that chain who can block tokenization and no particular reason to allow it. Chapter 1 sets out the three-condition test that this chain sits behind.

Wave two: unsecured credit failed in every case it was tried

Wave two, from 2020 to 2022, ran in the opposite direction from wave one. Wave one tried to bring investors from traditional finance onto a blockchain. Wave two brought assets from traditional finance to investors who were on a blockchain already, and those investors were protocols holding enormous stablecoin treasuries that earned nothing.

The demand was real. The construction was not. Maple Finance lent to trading firms and other institutions on reputation, with nothing pledged against the loan. After FTX collapsed in November 2022 the lenders on that platform took roughly 54 mn USD of bad debt: Orthogonal Trading accounted for about 36 mn USD across eight loans, after concealing its exposure to FTX, and Auros for about 3 mn USD more.

Overcollateralization means a borrower pledges assets worth more than the loan, so that a falling collateral price can be liquidated before the lender is out of pocket. Maple rebuilt on it in May 2024, at 150% and above, and a Spark research note from around August 2026 reports no credit losses since the change and about 2 bn USD of value locked in the protocol. Goldfinch, funded by a16z, ran the same unsecured experiment and did not change the model: more than 100 mn USD of loans originated, about 18 mn USD of defaults, which is 18% of the volume, and a wind-down vote in June 2026 with roughly 56 mn USD of principal still unpaid.

Two protocols, one model, and no counterexample in the data behind this series. Everything that works in on-chain private credit in 2026 has collateral, a whitelist of borrowers and valuation by an outside administrator, which is not a technological advance but a return to standards the credit market settled on long before blockchains. Chapter 8 follows that collateral rule into the lending loops built on RWA tokens today.

Wave three: Treasuries worked because the buyer was already on-chain and desperate

Wave three, in 2023 and 2024, is the one that worked. The usual explanation is that Treasury bills are the easiest asset to tokenize. True, and not enough: simplicity does not explain why it happened in 2023 rather than in 2019.

Four conditions had to hold at once. Every bill of a given maturity is identical to every other, so the token is fungible because the asset is. The price is public and nobody disputes it, which solves the hardest operational problem in RWA by choosing the asset rather than by building technology. A bill has no tenant, no roof and no covenants. And the fourth condition decided the rest: the buyer was already on the chain and had no alternative. In 2023 the federal funds rate passed 5% while tens of billions of dollars of stablecoins sat on public chains earning zero, including the billions of USDC that MakerDAO held in its Peg Stability Module at no yield at all.

That buyer proved itself before the products arrived. In January 2023 MakerDAO deployed 500 mn USD into short-term Treasuries through the Monetalis Clydesdale structure, which established that a protocol could buy half a billion dollars of government paper and have the operation hold.

The loudest date in that table is not the turning point. The quieter one is 6 April 2021, when Franklin Templeton's FOBXX went live as the first SEC-registered fund to use a public blockchain as part of its share registry, three years before BlackRock arrived. Its own prospectus describes a hybrid of book-entry records and public chains under the transfer agent's control, so the first-mover claim and the filing do not quite agree, and chapter 1 keeps both readings.

BUIDL became the institutional reference product: a British Virgin Islands fund company registered on 18 September 2023, offered under Rule 506(c) and Section 3(c)(7), a 5 mn USD minimum subscription, and more than 100 mn USD of dividends paid through December 2025.

The contrast that defines this wave sits in the same data. Tokenized real estate was about 457 mn USD in July 2026 and the category was shrinking. Eight years after Harbor, the asset wave one started with still does not work, and the reason is the one Harbor found in 2019 rather than anything about the state of the technology.

Wave four: the biggest numbers and the smallest market

Wave four, running through 2025 and 2026, has a different product from the three before it. It does not sell a market to investors; it sells operational savings to institutions, by moving their existing books onto a distributed ledger.

That shows up immediately in the two ways this market is measured. Distributed assets can be moved to a wallet outside the issuing platform and sent between wallets. Represented assets cannot, and exist to record a position on a chain rather than to put it in an investor's hands. Chapter 1 sets out both.

Between March and June 2026 the tradable half of that table was flat and the untradable half grew. Headlines described a growing tokenization market during those months. What grew was the register.

The construction of wave-four products matches their purpose. Feeder fund is the name for a vehicle that raises capital from investors and invests it into a single master fund, which does the actual investing; the arrangement is older than any of this and chapter 2 covers it. Apollo's ACRED and Hamilton Lane's HLSCOPE tokenize that feeder rather than the master fund, which bolts a distribution channel onto an existing structure and leaves the structure alone.

The honest second half of this picture is usually missing from critical accounts. Distributed value grew roughly six- to sevenfold between the start of 2025 and August 2026, from about 5 to 6 bn USD to about 38.3 bn USD. Reclassification does not explain that, because the taxonomy change of November 2025 cut the reported numbers rather than inflating them. The tradable market is growing fast, and it is about ten times smaller than the figure above 400 bn USD that reaches a headline.

The rule that explains all four waves at once

Four waves, four asset classes, one variable. Wave one tokenized real estate, venture fund interests and private company equity: heterogeneous assets with no continuous price, bought by investors who were off-chain and had better tools for buying them. Wave two found a buyer with money already on the chain and burned a lending model that had no collateral behind it. Wave three put the validated buyer and the easiest possible asset together and worked. Wave four moves institutional ledgers onto distributed infrastructure, which produces the largest numbers in the category and the smallest market an investor can participate in.

Tokenization never created demand for an asset. It connected to demand that already existed, and where the demand had to be manufactured first, it lost.

So the first question about any new RWA product is not the chain, the token standard or the jurisdiction. It is who the buyer is and where that buyer sits today. An answer along the lines of "an institutional investor who currently buys the same thing more cheaply off-chain" describes wave one in new packaging. The second question is the one Harbor's mortgage lender answered on 11 April 2019: which party in the legal chain above the token can veto this, and what reason does it have to agree. Chapter 10 scores every wave discussed here against a harder four-criteria test.