---
title: "What works, what does not, and the next five years"
author: Mariusz Szyma
date: 2026-09-15
lang: en
canonical: https://szyma.co/en/blog/rwa/what-works-and-whats-next/
series: "RWA: Past, Present and Future"
series_part: 10/11
series_url: https://szyma.co/en/blog/rwa/
data_as_of: 4–10 Sep 2026
---

# What works, what does not, and the next five years

Four tests with thresholds, not opinions, applied to six segments sold as real-world assets, leave two that pass. Both of those two predate the narrative that claims them. This chapter scores the six, names the most liquid part of the category as something that fails the definition of an RWA, and then reads the institutional forecasts for what they actually model, which is issuance, not turnover.

## What you'll learn

- Score any RWA segment against turnover, active-address ratio, DeFi use and product continuity.
- Name the two segments that pass all four, and the three that pass some.
- Separate a perpetual on a real asset from an RWA, and say which of the two is liquid.
- Read a governance-token decline as the price of a claim that does not exist.
- State what a 2027 forecast is modelling, and the event that would confirm or falsify each scenario.

## Four criteria, no room for a deck to pass them

Each criterion has a number and a threshold.

**Turnover ratio** is annual transfer volume divided by the value of the assets. Volume cannot be faked by a slide. If the assets do not move, there is no secondary market, only a subscription register. On-chain volume includes mint and redemption, so even 1× a year does not prove a secondary market. The ratio is an upper bound.

| Annual turnover | Reading |
|---|---|
| above 10× | a real secondary market |
| 2–10× | a secondary market exists and is thin |
| 0.5–2× | a subscription and redemption register |
| below 0.5× | dead, or represented in practice |

**Monthly active addresses** are addresses that did something in 30 days; **holders** are addresses that hold. The first measures use, the second measures state. A ratio of active addresses to holders below 5% means the product is held, not used.

**DeFi use** is the share of a segment deposited as collateral and the share actually borrowed against it. It is the only test of composability: whether somebody else's contract will accept the token without asking. If a tokenized asset is not pledged, tokenization gave faster settlement and nothing else.

**Product continuity** asks whether the product has existed without a change in terms. The most common ending in this record is not a default and not a hack. It is the issuer's decision to stop. The test is how many products in the segment have lived three years or more on the original terms.

A fifth, informal reading comes at the end: whether the sector's governance tokens reflect the sector's growth. That is a test of the narrative, not of the product.

## Two segments pass, and both predate the RWA story

Stablecoins: 303.04 to 303.21 bn USD and 285.65 million holders on 3 September 2026, RWA.xyz. USDT dates from 2014, USDC from 2018. Against the rest of the RWA market that is about 8 to 1 in value and 89 to 1 in holders. The only tokenized product with mass adoption is twelve years old, and most reports exclude it from "tokenized RWA" so as not to double-count the Treasury bills in its reserves.

Tokenized gold: 5.14 to 5.55 bn USD of market cap, CoinGecko, March to August 2026, from about 1 bn USD at the start of 2025. Liquidity is the criterion that settles it: 90.7 bn USD of spot volume in the first quarter of 2026 against about 5.1 bn USD of cap, about 5.9× a month, about 71× a year (CoinGecko, data to 31 March 2026). Full-year 2025 volume was 178 bn USD, ahead of every major gold ETF except GLD (Tiger Research, 27 February 2026). Continuity: PAXG since 2019, XAUT since January 2020, both without a change in terms.

Gold works because gold is homogeneous, pays no cash flow and has no credit risk. There is no NAV to strike, no dividend to distribute and no borrower to underwrite. Institutional physical gold in London has a 200 mn USD ticket; an ETF lowered the ticket and gave indirect ownership of fund units. The token keeps a direct claim on metal, in fractions, 24 hours a day. It is the one segment in which tokenization solved a problem that was not already solved.

Three caveats sit on that passing grade. Volume moves with the gold price: October 2025 jumped to 21.38 bn USD a month from 6.73 bn USD the month before. Concentration is extreme, XAUT and PAXG at 71% to 90% of the category. Outside metals the commodity category is a shell: JMWH (energy, 861 mn USD) is listed on no DEX and no CEX, and JSOY_OIL (soy oil, 300 mn USD) shows only mint and burn on-chain. That is 1.16 bn USD of products without a market, looking identical to PAXG in the category statistic.

![Four reality criteria scored against seven things sold as RWA: only stablecoins and tokenized gold pass all four, and perpetuals fail the definition before they reach the grid, on readings from May to Sep 2026.](https://szyma.co/blog/img/reality-criteria-matrix.53e467e3.svg)

## Treasuries, private credit and real estate: three different partial verdicts

Tokenized Treasuries are the largest serious segment: 15.20 bn USD distributed, 76 products, 58,658 holders (RWA.xyz via CoinLaw, data to 4 May 2026). The verdict is positive and narrower than the marketing.

Yield is real and worse than the underlying. Tokenized Treasuries paid 3.36% seven-day APY on 4 May 2026 and 3.28% in July 2026 against a federal funds rate held at 4.25 to 4.50% on 29 July 2026. The roughly 100 bps spread is the cost of the layer: management fee, transfer agent, custody, the short end of the curve. Protocol treasuries, market makers and firms with a stablecoin balance sheet pay it because on-chain settlement is worth more to them than the last basis point. A retail investor buying bond exposure buys a Treasury-bill ETF and receives more.

There is no secondary market. There is a subscription register, and it was designed that way. BENJI, live since 2021 on nine chains, has cumulative peer-to-peer transfer volume above 211 mn USD against 2.44 bn USD of assets (MetaMask citing Franklin Templeton, 3 August 2026). That is 8.6% of the fund moved between holders over five years, not per year. Daily volume for BENJI is not public and is not guessed here.

Securitize as a proxy for the segment: 4.99 bn USD distributed, 565.71 mn USD of monthly volume, 1,872 holders, 61 active addresses a month (RWA.xyz, about August 2026). That is 0.11× a month, about 1.4× a year, below the 2× threshold, so a register, and 3.3% of holders active, below the 5% threshold. Sixty-one addresses on the platform that holds BUIDL, ACRED, and funds from Hamilton Lane, KKR and VanEck.

Aave Horizon, the largest RWA-collateral market, held 539.8 mn USD of deposits and 163.5 mn USD of loans in July 2026, 30.3% utilization, 0.43% of the distributed market. Under 10% of tokenized value reaches DeFi at all. [Chapter 8](/en/blog/rwa/rwa-in-defi/) is the measurement; this chapter only needs the composability score.

Private credit ran a controlled experiment. Maple started in May 2021 without requiring collateral, took about 54 mn USD of bad debt in 2022, Orthogonal Trading 36 mn USD across eight loans after hiding FTX exposure, Auros about 3 mn USD, pivoted in May 2024 to 150% or more overcollateralization, and reached 4.65 bn USD of assets, up 81% year on year, with zero losses since, while overall DeFi TVL fell about 38% in the same window. Goldfinch ran the same experiment without changing the base model: over 100 mn USD originated, about 18 mn USD in defaults (18% of volume), TVL from 53.5 mn USD in February 2022 to 1.63 mn USD, wind-down voted June 2026, 56.15 mn USD unpaid, recovery estimated at two years or more. Yields in the segment compressed to 4.3 to 5.2% from 8 to 15% a year earlier, a 1 to 2 percentage-point premium over tokenized Treasuries at about 3.3%. Maple's own net interest margin is 0.79% on 22.09 mn USD of trailing-twelve-month revenue. Centrifuge has 1.6 bn USD of TVL and 792 active addresses a month, 2.02 mn USD per address.

One model works. The other went into wind-down. Figure's HELOC token, 20.10 bn USD on 7 July 2026, is represented: there is no token to buy. It finances home-equity loans in the real economy and dwarfs every crypto-facing credit protocol in the same category statistic.

Real estate has a completed case. RealT raised about 140 mn USD from 14,000 to 22,000 investors across about 700 properties, mostly Detroit, from 2019. Detroit sued 165 related LLCs from summer 2024. Courts banned rent collection and eviction. Liquidation was announced on 2 July 2026. Escrow was about 640,000 USD, 29 to 46 USD per investor, and burns at the trustee's own rate, 178,000 USD in two months, in about seven months, before any holder payout. News travelled mainly on Telegram and YouTube, with no official notice on the company site. On-chain, for years, a holder saw a separate LLC per property, rent arriving weekly in stablecoins, and a transferable token. The warning sat in Detroit's building-code register.

The one product that works in the category does not fractionalize. Propy handles whole-property closings conventionally and mints an NFT only as an on-chain receipt, not a substitute for state title law, with over 5 bn USD of lifetime volume.

![Turnover-to-AUM on a log scale: gold 71×, tokenized stocks 51×, PAXG 31×, Securitize institutional funds 1.4×, each an upper bound because transfer volume includes mint and redemption, on knowledge base readings to Aug 2026.](https://szyma.co/blog/img/turnover-to-aum.fe8ec3c4.svg)

## The most liquid part of the RWA market fails the definition of RWA

Perpetuals on real assets did 524.8 bn USD of volume in the first quarter of 2026 against about 19.3 bn USD of spot value, 27 times. On tokenized stocks specifically, 376.3 bn USD in perpetuals against 7.5 bn USD of spot, 50 to 1 (CoinGecko, September 2026). A perpetual on a tokenized stock fails the three-condition test in [chapter 1](/en/blog/rwa/what-is-an-rwa/): no off-chain asset generating the token's value, no identified counterparty, no redemption path. Trading is better served by the perpetual. Ownership and pledging are what tokenization is for. Mixing the two in one statistic produces most of the confusion about how big this market is.

Registration is not adoption. RWA.xyz on 2 September 2026 counted 2.54 million tokenized-stock holders against 152,006 active addresses across 3,918 assets, holders up 159.10% and active addresses down 89.65% in the same 30 days, a 16.7 to 1 ratio. Ten assets are 76.6% of volume. The MSFT reference price was halted at its first fixing because both of its venues failed liquidity gates (AltStreet, about 2 August 2026). In June 2026 exchanges had to cancel tokenized SpaceX offerings and return more than 1 bn USD of orders, because xStocks, the shared intermediary, did not secure an allocation; Ondo delivered a tokenized SpaceX on IPO day.

The "idle assets" figure is disputed even inside one source: a 56% headline against a computed 70.6% (only 379 of 1,289 assets over 100,000 USD had any transfer in a typical week). The two figures describe different universes, about 60 bn USD against a greater-than-100,000 USD subsample, and are not reconciled. Part of the idleness is specification: represented, first-level assets were never meant to trade.

![Holders of tokenized stocks rose 159.10% and monthly active addresses fell 89.65% in the same 30-day window to 2 Sep 2026: state and use, moving in opposite directions.](https://szyma.co/blog/img/retail-signups-vs-use.e765090a.svg)

## Governance-token declines are correct pricing, not an anomaly

The sector grew about sixfold from early 2025, from about 5 to 6 bn USD distributed to about 38 bn USD. Six of the seven largest RWA-project tokens returned between −44.7% and −98.8% from January 2025 to March 2026, and only SYRUP gained, +28.6%. By August 2026 MPL was −98.96% at a 680,000 USD cap, GFI −99% under 1 mn USD, CFG −94%, PENDLE −78%, ONDO −79 to −85%.

ONDO's circulating supply rose from 31.6% in June 2026 to 48.7% on 22 August 2026 of a 10 bn-token maximum, with about 6.5 bn more due 2026 to 2028. Market cap rose 35.8% to 1.81 bn USD while the price fell from 0.4202 to 0.3709 USD. The "growth" was circulating supply coming out of vesting, not price appreciation. SYRUP has the cleanest revenue link in the category, 10 to 30% of monthly revenue to buybacks, yet four buybacks since September 2025 totalled only about 1.4 mn USD against a 244 mn USD market cap.

The underlying business kept moving. Ondo received FINRA authorization on 23 July 2026 for tokenized stocks and funds in the US. Its tokenized iShares Core S&P 500 reached about 74.6 mn USD, three times its March 2026 size. It was the only issuer to deliver a tokenized SpaceX on IPO day.

A **compound annual growth rate**, CAGR, is the constant yearly rate that takes a starting value to an ending value over a stated number of years. It is the language forecasts use, and it is also the language that makes a "conservative" base case look less conservative once the multiple on today's market is written down.

## Forecasts model issuance, not turnover

Institutional forecasts range from 2 to 30 trillion USD and disagree mainly in what they count, not in optimism. BCG with ADDX, 2022: 16.1 tn USD by 2030, assuming 10% of global GDP is tokenized, later revised to 9.4 tn, a 42% cut in three years. McKinsey, June 2024: about 1.9 tn base, 1 to 4 tn range, excluding stablecoins, deposits and CBDC, not updated since. Citi, June 2026: 5.5 tn base, 2.7 to 8.2 tn, off a stated "about 17 bn" starting base. Standard Chartered, May 2026: 4 tn by end-2028, half of it stablecoins.

McKinsey's "conservative" base case is a 50× multiple on today's distributed market, a 146% CAGR held for four years, only a third below the market's own trailing twenty-month pace of about 7.0×, roughly 221% CAGR, from about 5.5 bn to 38.32 bn USD. No forecast in this set models turnover. All of them model issuance. The bottlenecks this chapter scored, idle assets, a 100 bps yield spread, product wind-down, the absence of a diversified index, do not appear in those models.

Scenarios make sense as observable conditions, not as target numbers. A bear case for 2027, about 30 to 35 bn USD distributed, wants three checkable events: the DTC Tokenization Service, planned for October 2026, delays or launches narrow; the CLARITY Act is not passed by end-2027; Federal Reserve cuts push tokenized-Treasury yield materially below 3%. A base case, about 65 to 75 bn USD, wants DTC launching as announced with measurable volume, and RWA-collateral deposits crossing 1 bn USD.

> **My take:** Tokenization has not created a market. It has plugged into markets that already existed. Wave one tokenized assets that had no on-chain buyer. Wave three tokenized Treasury bills that already had a desperate one. Real estate has now failed that test in two waves. The next decade will not be decided on the token layer, where the largest products do not even use the standard written for them, or on the chain layer, where 0.45% and 431% turnover sit on the same network. It will be decided on servicing and redemption: who keeps the register, who values an asset with no market, and what happens when the issuer decides the product no longer pays.

![Three 2027 scenarios for the distributed market, each with the checkable event that would confirm or falsify it; none of the rows models turnover, and institutional 2–30 tn forecasts sit outside the table because they count a different universe.](https://szyma.co/blog/img/scenarios-2027-2030.65d53840.svg)

## Key takeaways

1. Four criteria with thresholds, not opinions: turnover above 10× a year for a real secondary market, active addresses at or above 5% of holders, DeFi use at scale, and product continuity of three years without a change in terms.
2. Only stablecoins (303 bn USD, 285.65 million holders, since 2014) and tokenized gold (5.14 to 5.55 bn USD, about 71× a year, since 2019–2020) pass all four, and both predate the RWA narrative.
3. Tokenized Treasuries at 15.20 bn USD distributed are a subscription register, about 1.4× a year on the Securitize proxy, with a real yield about 100 bps below the federal funds rate; BENJI moved 8.6% of its value peer-to-peer over five years.
4. Maple's overcollateralized model reached 4.65 bn USD of assets with zero losses since May 2024; Goldfinch, on the unsecured model, voted to wind down in June 2026; RealT liquidated on 2 July 2026 with 29 to 46 USD of escrow per investor.
5. Perpetuals on real assets did 524.8 bn USD of volume in Q1 2026 and fail the three-condition RWA test; tokenized-stock holders rose 159.10% while active addresses fell 89.65% in the same month to 2 September 2026.
6. Six of seven large RWA-project tokens returned −44.7% to −98.8% from January 2025 to March 2026 while the sector grew about sixfold: the tokens have no claim on protocol revenue.
7. Institutional forecasts of 2 to 30 tn USD model issuance, not turnover; McKinsey's 1.9 tn "conservative" base is a 50× multiple on today's distributed market.
8. A 2027 bear of about 30 to 35 bn USD distributed and a base of about 65 to 75 bn USD are the author's scenarios, each tied to a named, dated event, not a bank target.

## Glossary

- **Turnover ratio**: annual transfer volume divided by asset value. An upper bound, because volume includes mint and redemption.
- **Active address**: an address that did something in the window, here 30 days, as against a holder, which is an address that holds.
- **CAGR**: compound annual growth rate, the constant yearly rate that takes a start value to an end value over a stated number of years.
- **Fee switch**: reminder from [chapter 9](/en/blog/rwa/whos-who/): a governance vote that would pay protocol fees to token holders.
- **Wind-down**: the issuer's decision to stop the product. The most common ending in this record.
- **Idle assets**: value that recorded no transfer in a typical week. Two published figures, 56% and 70.6%, describe different universes.

## Go deeper

- [The four waves](/en/blog/rwa/four-waves/): the criteria in this chapter retest every wave against a harder bar, including the unsecured-credit losses Maple and Goldfinch produced in wave two.
- [RWA in DeFi](/en/blog/rwa/rwa-in-defi/): the 0.43%-of-market composability score used here is measured there.
- [How to analyze any RWA in 30 minutes](/en/blog/rwa/thirty-minute-teardown/): the four criteria become checklist items with a time budget.

## Sources

- 🟢 RWA.xyz, `/platforms` and `/networks`, 3 Sep 2026: stablecoins 303.04 to 303.21 bn USD, 285.65 million holders.
- 🟢 RWA.xyz, `/platforms/securitize`, about Aug 2026: 4.99 bn USD distributed, 1,872 holders, 61 active addresses, 565.71 mn USD monthly volume.
- 🟢 RWA.xyz, `/stocks`, 2 Sep 2026: 3,918 assets, 2.54 million holders (+159.10% / 30d), 152,006 active addresses (−89.65% / 30d).
- 🟢 CoinGecko, *RWA Report 2026*, data to 31 Mar 2026: gold spot 90.7 bn USD in Q1 2026; perpetuals on RWAs 524.8 bn USD.
- 🟢 CoinGecko, *Tokenized Equities Report*, Sep 2026: spot 7.5 bn USD, perpetuals 376.3 bn USD, ten assets 76.6% of volume; SpaceX IPO refunds above 1 bn USD.
- 🟢 Aave, *Aave Horizon Launches*.
- 🟡 Tiger Research, *2026 Commodity Tokenization Market Analysis*, 27 Feb 2026: gold 178 bn USD volume in 2025; JMWH 861 mn USD unlisted; JSOY_OIL 300 mn USD mint-and-burn.
- 🟡 Stobox, *The State of RWA Tokenization — 2026 Mid-Year Report*, 10 Jul 2026: 56% idle; under 10% of tokenized value in DeFi.
- 🟡 Spark, *Tokenized Private Credit* and *Tokenized Real Estate*, about Aug 2026: Maple 2022 losses, Goldfinch wind-down, RealT, Propy.
- 🟡 CryptoBriefing, about 28 Jul 2026, and Cointribune, about 6 Jul 2026: RealT liquidation, escrow, trustee costs, communication channel.
- 🔴 CoinLaw citing RWA.xyz, 27 May 2026, data to 4 May 2026: Treasuries 15.20 bn USD, 76 products, 58,658 holders, 3.36% seven-day APY.
- 🔴 MetaMask citing Franklin Templeton, 3 Aug 2026: BENJI cumulative peer-to-peer volume above 211 mn USD, 2.44 bn USD AUM.
- 🔴 CryptoDaily, 10 Jul 2026: Aave Horizon 539.8 / 163.5 mn USD.
- 🔴 CoinPaprika, Aug 2026: governance-token prices and ONDO supply; Maple AUM 4.65 bn USD.
- 🔴 CryptoRank / BlockchainReporter, 15 Jul 2026: six of seven large RWA tokens −44.7% to −98.8%, Jan 2025 to Mar 2026.

## Caveats

Treasury-segment turnover is computed on a Securitize proxy, because that is the one platform with value and volume from a single read. Securitize does not include BENJI, USYC or USDY. How much of the 565.71 mn USD monthly volume is mint and redemption, rather than holder-to-holder, was not established. The register-not-market verdict leans on BENJI's five-year cumulative figure for that reason.

The 56% and 70.6% idle figures are not averaged. No breakdown of idle assets by class was found, so the dispute between "represented by design" and "failed liquidity promise" stays open.

Many operational numbers in this chapter, active addresses per platform, governance-token caps, Maple's net interest margin, exist only on tertiary aggregators. Directions and orders of magnitude are strong; any 🔴 figure needs a fresh read before further citation.

Canton Network's about 371.3 bn USD of represented value, cited elsewhere in the series, remains unverified after a dedicated September 2026 search pass and is not restated here as settled. No confirmed, SEC-sanctioned tokenized-stock trading venue was established as live as of September 2026.

The 2027 scenarios are the author's construction from cited inputs, not a bank forecast. Time-to-recovery for Goldfinch is an estimate from secondary sources.