The Price Has to Come From Somewhere: How a Tokenized Stock Gets Marked Onchain

The problem in one sentence

A tokenized share has two prices at all times. One is the price of the underlying share, discovered on a regulated exchange during its trading session by a matching engine with a queue of orders. The other is the price of the token, discovered onchain by whatever a pool or an order book says at that moment. Nothing in the code forces the two to be equal. Everything that pulls them together is external machinery, and most of that machinery has a dependency on a price feed that goes quiet at four in the afternoon.

What an oracle actually does here

An oracle is not a price. It is a process for putting a number that originated off the chain into a place where a contract can read it, plus a set of rules about when that number is allowed to change and who is allowed to change it. In practice a feed for an equity has four moving parts: sources, which are the venues or vendors whose quotes get sampled; an aggregation rule, usually a median across sources to make one bad quote harmless; a publication trigger, which says the number gets written when it moves more than some threshold or when some time has passed, whichever comes first; and a staleness rule, which tells any contract reading the feed how old the number can be before it should be treated as unusable.

That last part is the one that does the work. A contract cannot tell the difference between a price that is correct and unchanged and a price that stopped updating because the publisher fell over. It can only see the timestamp. So the discipline is entirely in the reader: a lending market or a perpetual venue that consumes an equity feed has to decide what it does when the stamp is older than its own tolerance, and the honest answer is usually to stop doing anything that needs a mark.

Three places the price gets used, and they are not the same

The first is trading. In an automated pool nobody reads the oracle at all. The price is a consequence of the ratio of the two assets in the pool, and it moves because someone traded against it. The oracle is irrelevant to the swap itself. What the oracle does is tell arbitrageurs, or the bots working for them, that the pool has drifted from the reference, at which point they trade the pool back and take the difference. That means the pool price tracks the share price only as well as the arbitrage path works, and the arbitrage path is a chain of steps that each need to be open.

The second is collateral. If a stock token is posted against a loan or margins a position, some contract has to mark it, and there is no way to do that without a feed. Here the oracle is not advisory, it is the thing that decides whether a position lives or dies. A feed that is thirty seconds late during a fast move liquidates people who were solvent, or fails to liquidate people who were not, and the protocol eats the difference.

The third is redemption. Where a token is redeemable for the underlying, the redemption is usually struck at a price the issuer determines from its own books rather than from anything onchain. This is the strongest anchor of the three, because it does not depend on a public feed at all. It is also the slowest, and it is typically available to a short list of authorised parties rather than to whoever is holding the token at three in the morning.

Where it breaks: the session gap

Equity feeds inherit the trading calendar of the thing they track. Between the close and the next open, and across weekends and holidays, there is no continuous stream of real trades to sample. Publishers handle this in different ways, none of them free. Some hold the last close and mark the feed stale, which is honest and makes anything that depends on a live mark unusable overnight. Some keep publishing from extended-hours or futures-implied sources, which keeps the number alive but changes what the number means, because the liquidity behind an after-hours quote is a fraction of the session's. Some synthesise from a related instrument, which imports that instrument's basis risk.

The consequence is that a stock token has a different risk profile at different hours of the day, even though the contract is identical. During the session the pool is tethered by arbitrage against a deep venue. Outside it, the tether is a quote from a thin book or nothing at all, and the pool price is whatever the last person to trade decided it was. A halt does the same thing in the middle of the day: the reference stops, the token does not.

Why the numbers on the chain matter to this

Depth is the shock absorber. A pool that can absorb a large order without moving much makes the gap between the token and the share expensive to open and cheap to close, which is exactly the condition arbitrage needs. Robinhood Chain carried $895.2m of total value locked and turned over $1.89bn of decentralised exchange volume in a day, with $903.9m of that on Uniswap V3 and $595.8m on Uniswap V4, on a stablecoin base of $1.01bn, according to DefiLlama. Chain fees ran $13.0m in a day and $246.4m over thirty days on the same source. Aggregate depth is not the same as depth in the specific token you hold at the specific hour you need it, and the aggregate figure will look reassuring on a night when the individual pool is empty.

What to watch

Four things tell you whether a stock token is properly marked. Whether the venue publishes which sources feed the price and how they are combined, rather than just naming a provider. What the stated staleness threshold is, and what protocols reading the feed do when it is breached, because pausing is a design choice and not a failure. Whether the feed continues outside session hours and, if it does, what it is sampling instead of trades. And whether redemption is open to a wide set of participants or a narrow one, because that determines whether the hard anchor exists for you or only for someone else. A token that trades continuously against a feed that does not is not broken, but it is a different instrument after the closing bell than before it, and the disclosure that says so is worth more than the one that says the price is accurate.