TVL is Misleading: The Hidden Risks Inside Tokenized Assets

Total Value Locked (TVL) has become the default scoreboard for tokenized Real-World Assets (RWAs). This category now sits anywhere between $17 billion and $35 billion, depending on which tracker the information is sourced from. The issue is that a dollar of TVL in a tokenized treasury fund does not carry the same risk as a dollar of TVL in a thinly traded private credit pool. With the risk not being the same, generalized metrics fail to make any distinction. For developers building on top of RWAs, that gap between headline value and actual liquidity is where the real exposure sits. 

RWA tokenization has risen to become one of Web3’s biggest sectors of late. Tokenized treasuries, private credit, and commodities pushed past decentralized exchanges (DEXs) to become one of the larger categories within DeFi by TVL in late 2025. The number has only kept growing. BlackRock’s BUIDL, Franklin Templeton’s BENJI, Ondo’s OUSG and USDY, and Apollo’s ACRED are no longer pilot programs, and builders are constantly shipping lending markets, vaults, and structured products on top of them.

All of that is established and here to stay. What’s worth questioning is the metric used to describe how healthy this category actually is. TVL has always been a blunt instrument, letting onlookers know how much value sits within a contract, but doing nothing to describe how that value behaves, how easily it moves, or who controls it. With RWAs, it is worse: since the value backing these tokenized assets lives off-chain, the price comes from outside the chain’s view, and the assets’ behavior under stress depends on factors that TVL cannot capture, such as legal and operational machinery.

The Same TVL, Two Completely Different Risk Profiles

A recent academic paper using data from RWA.xyz made this idea concrete by comparing BUIDL and BENJI, two tokenized treasury products that are always lumped into the same buckets on every dashboard. BUIDL carries roughly $2.5 billion in asset value across 108 holders and exhibits monthly transfer volumes worth about $1.1 billion—a turnover ratio near zero. BENJI, by contrast, holds a lighter $823 million in assets, but spread across 1,100 holders, with monthly transfer volumes closer to $10 million.

The TVL line alone here makes BUIDL look like the more liquid product because it boasts three times the value of “its competitor”. But, looking at the moving parts flips the picture: its float barely turns over relative to its size, while BENJI, having a lower asset value, boasts a wider, more active base. Knowing which fund is safer because of that is a debate of its own, but what matters here is much narrower—the fact that TVL alone could not have painted the complete picture. Builders need the turnover and holder data underneath it, and dashboards built for DeFi tokens rarely show either by default.

On-Chain Price Does Not Reveal the Full Picture

Despite TVL being a flawed metric, there is a second layer that introduces a structural problem in this context—the TVL is only as good as the price used to calculate it. For instance, a tokenized treasury fund is not a price discovered on an order book; rather, it is a net asset value (NAV), which is the per-share value of the fund’s underlying holdings.

This figure is calculated off-chain by a fund administrator, usually once a day, on the same schedule a traditional money market fund would publish its own NAV, and that figure then has to be pushed on-chain somehow. Although every TVL dashboard assumes the handoff to be current and clean, that is not the scenario, leading different issuers to land on different solutions to bridge the gap.

Some issuers are pushing pricing frequency to its limits, like Superstate, which issues tokenized short-term treasury funds like USTB and USCC, running a continuous oracle updating yield accrual every second. It does so even though the underlying NAV is still calculated daily by a third-party administrator, with the per-second feed being an interpolation over a once-a-day number, not a genuinely continuous valuation. Centrifuge, which tokenizes private credit pools rather than Treasuries, offers another solution, albeit in the opposite manner. It batches deposits and redemptions into epochs and settles them against a verified NAV only when each epoch closes.

Both expose the same tension, with the DeFi protocols wanting the price for every block, while regulated funds generate one on their own schedule based on their legal setup. A lending market that treats a tokenized Treasury token as continuously price-discoverable is quietly borrowing precision the oracle never actually had.

There is also no shared way to read these prices across the category. Ondo’s OUSG accrues yield by raising the NAV-per-token, so its dollar price climbs slowly, and a contract has to read that climbing price correctly. Its rebasing sibling, rOUSG, instead holds price near a flat $1 and pays yield by minting more tokens to holders, so a contract reading price alone would miss the yield. It has to read the balance change too.

Maple, a private credit platform, prices tokens more like a vault share under the ERC-4626 standard that DeFi vaults already use. Here, the share price itself rises as underlying loans accrue interest.

A developer integrating across BUIDL, Ondo’s tokens, and Maple has to write separate logic for each, since none of them expose prices in the same way, and that same lack of standardization makes these feeds targets for ordinary oracle manipulation. Here, a price built from thin or infrequently updated sources gives an attacker, or just a bad market, room to push a wedge between what a contract believes an asset is worth and what it can actually be sold for.

TVL Does Not Know About Redemption Gates

The other issue with TVL is that it cannot indicate what happens when someone tries to leave. Tokenized funds inherit the redemption mechanics of whatever they wrap, and those mechanics were not designed with on-chain composability in mind. A token backed by a daily-dealing money market fund can usually be redeemed close to instantly. A token backed by private credit or real estate often cannot, because redemption can take days, run in batches, or sit gated behind the issuer’s discretion during stress.

None of this shows up in the contract that relays balances; the token looks exactly as liquid as any other ERC-20 until someone tries to exit and discovers the underlying asset has its own queue. This matters for anyone building collateral logic. If a protocol prices a tokenized RWA off a continuous oracle while the underlying redemption is gated or epoch-based, a liquidation event can try to act on liquidity that does not exist on the timeline the contract assumes.

Morpho’s isolated market design is a deliberate response to exactly this. Its RWA collateral grew past $620 million by March 2026, but is kept walled off so asset-specific liquidity risk cannot bleed into pools holding different collateral.

The Concentration Problem Hiding Inside Every Headline Number

Holder concentration is another blind spot, and it compounds the first two: TVL hides how much of that value actually trades, and it hides how reliable the price behind it even is. It also hides who is actually holding the float. A large share of RWA value sits with a small number of issuers, so a single issuer hitting operational trouble or a compliance freeze can move a category’s effective liquidity at once, even if the aggregate TVL number barely flinches.

A token with $2 billion in TVL spread across 100 wallets and a token with $800 million spread across 1,000 wallets are not exposed to the same tail risk, even though the first number looks more reassuring on a dashboard. Few RWA analytics products surface this by default, leaving the burden on the integrating developer to pull holder distribution data separately rather than trust the TVL widget that shipped with the issuer’s docs.

What This Actually Means for Builders

None of this is an argument against building with tokenized assets. The category is real, and the on-chain rails for issuing, transferring, and settling these tokens work. The argument is narrower—TVL is the wrong primary signal for determining whether an RWA integration is sound, and treating it as sufficient due diligence is how risk quietly gets inherited.

A few things are worth pulling into any integration before TVL gets near a go/no-go decision:

• Turnover, not just asset value
• The NAV update cadence, and whether the feed is a genuine on-chain price or an interpolation sitting on a slower off-chain number
• What redemption actually looks like for the underlying asset class, sized into collateral parameters rather than the timeline a continuous oracle suggests
• Holder concentration, pulled separately, since a clean TVL number can sit on an ownership base that’s one issuer problem away from a liquidity event

Most of this still has to be assembled by hand, asset by asset, and issuer by issuer. TVL will keep climbing as a headline number regardless. But whether that growth represents durable, well-understood risk or a stack of untested assumptions is a question the metric itself was never built to answer.