Inside Atlas: scoring the assets the on-chain market has not reached

The market is already built

Two-thirds of the world's stablecoin supply is held by people in emerging markets. Almost none of them can buy a listed company from their own region on-chain.

That sentence is the whole problem, and it is worth sitting with for a moment. The capital is there. The assets are there. They are frequently in the same country. What is missing is the instrument in between.

We have been building toward this argument for a while. In April we made the case that listed assets are the faster entry point for emerging-market tokenization, because they are already priced, audited and market-visible. In May we set out how private assets become investment products, and what a term sheet actually reveals about an asset's readiness. In June we looked at why African yield is real and the access infrastructure is not. Each of those described a gap. This one describes the filter we use to decide what goes through it, and publishes the weights behind it.

Start with where the market actually is, because it has moved faster than most coverage suggests.

Tokenized equity now carries roughly USD 2.5 billion in distributed value across 3,961 instruments, held by 2.04 million holders, with USD 27.01 billion of transfer volume last month and a 1-year sector return of 35.43%. Among tokenized real-world assets it is the fastest-growing category by a wide margin.

The derivatives layer grew faster still. Quarterly volume in real-world-asset perpetuals went from USD 12.37 billion in the fourth quarter of 2025 to USD 202.67 billion in the second quarter of 2026. Over the same window, real-world assets moved from 1.8% of Hyperliquid volume to 32.2%. In the week to 19 July, tokenized equities and commodities generated USD 25 billion on the platform, 52% of weekly volume, overtaking crypto perpetuals for the first time. By late July, HIP-3 markets accounted for roughly half of daily perpetual volume, against about 2% at the start of the year.

The collateral layer matured alongside it. Vault curators, a category that barely existed in 2024, now manage on the order of USD 6 billion across Morpho, Aave and Spark, with institutional participants entering through vehicles such as the Apollo and Morpho credit vault. Licensed retail venues, xStocks and Ondo among them, have processed tens of billions in cumulative tokenized-equity volume. Behind all of it sits roughly USD 302 billion in stablecoins and about USD 38.18 billion of distributed real-world assets across 2.8 million holders, as of 24 August.

So the rails work. The venues work. Demand is measurable rather than theoretical.

What is worth noticing is what all that infrastructure currently points at. The instruments carrying the volume reference a fairly narrow set of listings. On rwa.xyz, of the 3,961 tokenized stocks listed, the largest positions are an S&P 500 tracker, a Nasdaq-100 tracker, Nvidia, Tesla, Alphabet, Micron and a cluster of listed crypto-adjacent companies. On Hyperliquid, TradeXYZ runs an XYZ100 Nasdaq-100 market alongside perpetuals on Nvidia and Tesla, settled in stablecoins rather than shares. Where exposure to other markets exists, it is usually reached through a depositary receipt. SK Hynix appears as its American depositary shares at USD 35.8 million, TSMC and ASML through receipts at USD 14.4 million and USD 12.8 million, and Korea itself through the iShares MSCI South Korea ETF at USD 10.2 million. Each is a receipt^[1] or an exchange-traded fund listed somewhere else. As of 24 August, nothing from Tadawul, B3, the Johannesburg exchange, the Korea Exchange or Nairobi appears in the top fifty tokenized equities by value.

What the next listings bring, and what they get back

The case for widening the asset base is usually made badly. It gets framed as fairness, or as inclusion, or as a vague gesture at global coverage. No allocator has ever revised a mandate because of them, and they are not the reason this matters.

The real case runs in two directions, and both sides are measurable.

What these listings bring to on-chain finance

Return, at a moment when the gap is unusually wide. In 2025 the MSCI Emerging Markets index returned 33.6% against 17% for the S&P 500. Consensus earnings growth forecasts for emerging markets in 2026 sit above 20%, ahead of developed-market projections, and State Street put preliminary 2025 earnings per share growth at 16%. MSCI EAFE returned about 32% over the same year.

Korea is the clearest single illustration. The KOSPI rose 76% in 2025 and was up 108.85% year to date before the rebalancing that began on 3 June, figures given by Korea Exchange chief executive Jeong Eun-bo. Foreign holdings of Korean financial assets crossed USD 3.02 trillion at the end of June, the first time above that mark and the largest quarterly jump since the Bank of Korea began compiling the series in 1994. MSCI still classifies Korea as an emerging market, citing identification requirements and disclosure practices for foreign investors. Two of the three largest tokenized non-US equity positions are Korean, and both are US-listed proxies rather than the Seoul line.

Worth separating what those two classifications measure. An index classification describes the operational friction an index tracker meets. Atlas measures whether a foreign holder can obtain the asset, receive its income and repatriate capital with enforceable rights attached. A market can score badly on the first and well on the second, and several do, which is why Samsung carries a maximum accessibility score further down this article while Korea remains an MSCI emerging market.

Valuation, at a level not seen in two decades. The MSCI EM index carries forward price-to-earnings ratios somewhere between 11.6 and 13.5 times, while the S&P 500 trades north of 20 times. That is a discount of roughly 40 to 50%, against a historical average nearer 25 to 28%. Emerging-market equities have fallen below half the valuation of US equities for the first time in at least twenty years.

Diversification that actually diversifies. This is the one that matters most to anyone running a collateral book. On-chain collateral today is overwhelmingly US duration and US credit, plus a technology-heavy equity sleeve. Those things move together. When the cycle turns, they turn as one, and a base assembled that way does not spread risk so much as concentrate it under a different label. Al Rajhi Bank does not trade on the same drivers as Nvidia. Neither does Petrobras, nor Vale, nor Safaricom, nor SK Hynix on its Seoul line rather than its New York receipt. That is not a marketing claim, it is the reason curators ask for the exposure.

Yield in the band the market is short of. Tokenized US Treasuries anchor the low end of the on-chain yield curve near 4%. Tokenized private credit sits at the high end, nearer 8 to 12%. In between, the four to eight percent band with audited collateral and predictable cash flow is exactly what published curator frameworks ask for, and exactly what the current inventory has least of. Emerging-market dividend equity lands in it. Dubai Islamic Bank, Al Rajhi, Banco do Brasil and Itaú all sit inside that band on trailing dividends.

What on-chain finance brings back

Here is the part that rarely gets said out loud, and it is the stronger half of the argument.

The buyers are already there. Goldman Sachs estimates that roughly 66% of global stablecoin supply is held by people in emerging markets. Seven of the top ten countries in the Chainalysis Global Crypto Adoption Index are developing economies. Growth in value received is fastest in exactly those regions: Asia-Pacific up 69% year on year to USD 2.36 trillion in value received, Latin America up 63%, Sub-Saharan Africa up 52% to USD 205 billion, against 49% in North America and 42% in Europe. On Tron, USDT alone accounts for roughly 59% of stablecoin supply and about 74% of on-chain trading volume concentrated in Asia, Latin America and Africa.

Where that goes next is not speculative either. Standard Chartered estimated in October 2025 that up to USD 1 trillion could shift out of emerging-market bank deposits into stablecoins within three years, with Egypt, Pakistan, Bangladesh and Sri Lanka most exposed. S&P Global projected in January that dollar stablecoin holdings across 45 emerging markets could reach USD 730 billion.

Read those two sets of numbers together and the shape of the problem is obvious. Enormous dollar balances sitting on-chain in emerging markets. Enormous listed asset bases in the same markets. And essentially no instrument that connects one to the other.

For the issuers and exchanges in those markets, on-chain distribution is not a technology upgrade. It is reach. It means a buyer base that does not require a local brokerage relationship, a sub-account, a currency approval or a minimum ticket that rules out most of the people who would want in. It means secondary liquidity outside the 09:00 to 15:30 window that Seoul, São Paulo or Dubai actually trade in, which matters when the investor base spans fifteen time zones. It means a Tadawul or B3 listing can be collateral in a Morpho vault, not only a line in a custody account.

Both sides of that trade are currently unserved. To an allocator that is the definition of an inefficiency worth underwriting, and to a portfolio it is uncorrelated supply the existing collateral base cannot manufacture. That is what makes it worth building for, and it is also why the selection has to be strict, which brings us to the actual subject.

What a score has to answer

A curation score has to answer two questions that are easy to blur together.

The first is whether the asset can be wrapped at institutional size. That is feasibility, and it resolves into operational facts rather than opinions:

• Is there enough free float and daily volume that creation and redemption can happen without moving the underlying market
• Is there a compliant channel through which the asset can be held on behalf of investors, with enforceable rights attached
• Is there a working custody chain in that corridor
• Can dividends actually leave the country and reach the investor
• Is the local currency convertible at market rate, without quotas or queuing

Any one of those failing makes the rest academic, however good the asset looks on your screen. There is more than one legitimate way to solve them, and different issuers will solve them differently. What matters for scoring is whether they are solved, and whether the solution holds up when something goes wrong.

The second question is whether the wrapper, once it exists, will find demand. That resolves somewhere else entirely: in the published due-diligence frameworks of vault curators such as Steakhouse, Gauntlet and Sentora, in whether a Chainlink, Pyth or RedStone feed exists on the chains the instrument will live on, in whether comparable collateral has already cleared a Morpho, Aave or Spark market, and in how regular the cash-flow pattern is, because on-chain accounting breaks on lumpy distributions.

Five of the six Atlas principles answer the first question. One answers the second. Weights are fixed for the current cycle and reviewed annually. The methodology is published. The scoring functions inside each principle are not.

Swipe or scroll for all six

Weights are a calibration of the current demand cycle, not a permanent statement. The full universe is re-scored quarterly and the weights are reviewed annually.

Why DeFi Fit carries the highest weight

Giving the largest weight to on-chain absorbability looks, at first glance, like a crypto-native bias. It is the opposite. It reflects where the binding constraint actually sits.

Every other principle establishes whether the asset can be wrapped. DeFi Fit establishes whether the wrapper will be wanted. Institutional allocator demand is real and durable, but it assembles across quarters of due diligence, committee cycles and mandate revisions. On-chain demand behaves differently. Once an instrument clears a curator framework and has a working price feed, allocation can arrive in days. When a methodology has to choose which signal to weight most heavily, it should weight the one that moves fastest and fails most often. For a listed asset outside the corridors that are already served, the question is rarely whether anyone eventually wants the exposure. It is whether anyone can take it at the moment it exists.

DeFi Fit is also not one thing. It carries more sub-signals than any other principle because on-chain demand arrives through five distinct channels, and they want different assets.

Channel	What it absorbs	What it needs from the asset
Decentralised derivatives	Reference exposure for perpetual contracts. The largest volume channel by a wide margin: Hyperliquid alone carried USD 202.67 billion of RWA perpetual volume in the second quarter, and tokenized assets occupy 23 of its top 30 pairs by open interest.	A reliable oracle, recognisable liquidity, and a price that behaves outside home-market hours. Perpetual demand pulls spot demand behind it, because basis and delta-neutral strategies need the underlying.
Vault curators	Yield-bearing collateral inside permissionless credit vaults on Morpho, Aave and Spark, a category managing roughly USD 6 billion.	Predictable, forecastable cash flow, audited financials, and clean adverse-event handling. Lumpy or surprise distributions break vault accounting.
Licensed retail venues	Spot tokenized equity on venues like xStocks and Ondo, reaching a verified user base in the hundreds of millions.	Brand recognition and trading velocity. This channel rewards names people already know and want to hold.
Yield-protocol composability	Collateral backing stablecoin issuance and structured on-chain yield.	Stability and a defensible haircut more than headline return. Absorptive capacity here is large and largely unfilled for equity-class assets.
Institutional and family-office allocation	Direct holdings through permissioned rails.	Net delivered return after tax and currency friction, plus a regulatory posture that survives an investment committee and fits an existing mandate.

An asset can be excellent for one of these and wrong for another. A steady dividend payer with modest trading interest may be ideal curator collateral and unremarkable on a retail venue. A high-recognition cyclical name may trade heavily and still fail the cash-flow predictability test that vault inclusion depends on. Score DeFi Fit as a single undifferentiated number and you lose the one thing you needed: where the instrument should actually go.

So Atlas produces a routing alongside the score. Every asset is assessed against three access pathways, defined by how much identity verification the buyer has completed, because that determines both the addressable base and the regulatory posture the wrapper has to carry.

Routing follows the principle profile, not the headline number. Strong DeFi Fit points toward permissionless channels. Strong scale with workable accessibility points toward licensed retail. Strong net-return efficiency points toward institutional wholesale, because that buyer cares most about what lands on the books after withholding and currency cost. Most assets route to two channels, with one primary.

Which is why the weight stops at 30%. An instrument built only for on-chain absorption has no second buyer when curators rotate, and curator capital rotates by design, monthly, on risk-adjusted spread. Holding DeFi Fit below a third keeps the other seventy percent, the part that describes the asset rather than its distribution, in control of the outcome.

A high score is not an approval

Weighted composites have a known failure mode. They hide fragility. A high score built from six balanced results looks identical, on paper, to an equally high score built from one outstanding principle carrying five weak ones. The second asset is materially riskier, and the headline number cannot tell an allocator which one is in the portfolio.

Atlas handles this with floors. Every principle clears a minimum independently, whatever the composite says. An asset scoring top of the range on DeFi Fit that falls below the floor on accessibility does not proceed, because the operational failure that principle describes does not get less real when averaged against something else. Above the floors, a threshold applies to the composite itself, set deliberately so the framework admits fewer assets than pass the initial screens.

This is usually where the highest-yielding candidates fall out. The patterns recur:

• Mortgage-paper vehicles yielding eleven to twelve percent with mandated payouts have close to ideal on-chain mechanics, and concentration in a category with a cyclical default history. That surfaces in return profile and net-return efficiency, not in DeFi Fit.
• Distressed property vehicles above twelve percent have quarterly distributions, an established price feed, and net asset value erosion that the yield is compensating for.
• Listed banks in corridors under active enforcement have brand recognition, a working oracle and a reliable distribution history, alongside trading-halt and dividend-suspension risk that no on-chain metric surfaces.
• Recently listed consumer brands with strong retail interest can fail on free float alone, because the redemption mechanic does not work at wrapper size when the float is locked up.

The generalisable point: headline yield well above the band curator frameworks ask for is usually compensation for a risk the framework exists to find. Occasionally it is not, and those are the interesting cases. The discipline is that the burden of proof sits with the high number.

Assets clearing both the floors and the threshold route to one of three sleeves. Anchor carries institutional credibility and applies a volatility ceiling on top of scale and return, so a name cannot qualify on size alone. High-Yield is for income-led names where dividend reliability rather than capital appreciation defines the role. Diversifier is for assets that clear the threshold without meeting either set of gates and provide sector or geographic balance. Where an asset satisfies both Anchor and High-Yield gates, the conflict resolves at committee on positioning rather than algorithmically, and that override is tracked separately so it cannot quietly become a back door. The same selection discipline governs private assets, scored against a different set of dimensions suited to unlisted supply.

Three completed scorecards show what the output looks like.

Swipe or scroll for all three

Three eligible assets, three different reasons, and none of them tells you what you would guess from the total. Samsung is held back by DeFi Fit alone, which says nothing about the company and everything about missing on-chain infrastructure. Petrobras is deep and liquid, with tax and currency friction as the binding constraint. Dubai Islamic Bank scores maximum on accessibility and gives back ground on scale. Similar headline numbers, three different instruments, three different routings. That is the information a composite alone would have buried, and the reason the scorecard shows every principle rather than the total.

What comes next

Publishing a methodology means saying where it is still weak.

Atlas scores on three to five years of history, so regime changes and abrupt regulatory shifts land at the next re-scoring rather than in real time. The universe is re-scored quarterly, with event-triggered overrides for corridor-wide shocks. Average daily volume is a ninety-day proxy, and in a real stress window historical liquidity overstates what is available exactly when redemption is needed, so wrapper sizing assumes a substantial contraction rather than the observed figure. Some DeFi Fit inputs are reflexive: oracle availability and collateral precedent are influenced by activity T-Blocks itself undertakes, so precedent scoring excludes T-Blocks instruments and a price feed counts only where another issuer relies on it too.

The next iteration separates derivatives demand from spot demand, which profile differently on volatility and corporate-action frequency and are currently aggregated. And the weights are a read on this demand cycle, not a permanent judgement. If institutional wholesale becomes the dominant channel, the thirty percent moves.

The market at the top of this article is not waiting for permission. It has venues, liquidity, collateral infrastructure and hundreds of billions in capital already on-chain, most of it held by people in exactly the markets whose assets it cannot reach. What decides which assets reach it next is not enthusiasm about tokenization. It is whether anyone is willing to be specific, in public, about what qualifies and why most things do not.

These are our six. The weights are above. If you think one of them is wrong, say so in the open, because that is the only way a framework like this gets better.