ChangXin Memory Technologies has not yet officially listed on Shanghai’s STAR Market, but Hyperliquid already has a 24/7 tradable “share price” for it.
On July 14, Trade.xyz launched a pre-IPO perpetual contract for ChangXin through Hyperliquid’s HIP-3 framework under the ticker xyz:CXMT. The contract started from a $5 reference price, supports up to 5x leverage, and settles in USDC.
After launch, the price hit $6, then $7.2, and later $8.64. As of 02:13 UTC on July 16, Hyperliquid on-chain data showed a CXMT mark price of about $7.37, 24-hour trading volume of roughly $50.56 million, and open interest notional value of about $23.07 million.
At the same time, ChangXin has set its A-share IPO price at RMB 8.66 per share, with an issuance valuation of RMB 579.18 billion, and is expected to list on July 27. Using an exchange rate of about 6.77 yuan per dollar, the $7.37 contract price on Hyperliquid works out to about RMB 49.9 per share, or about 5.76 times the IPO price. Based on about 66.88 billion shares after issuance, that implies a valuation of about RMB 3.34 trillion. A Reuters report said some investors expect ChangXin’s post-listing valuation to land in a range of RMB 3 trillion to RMB 5 trillion.
Still, the $7.37 figure is not ChangXin’s real stock price, nor is it a valuation pulled from a private-market database. It is a derivatives price produced through the combined effect of the order book, an internal oracle, mark price logic, funding, and price guardrails.
A contract on future public-market value
Trade.xyz defines a Pre-IPO Perpetual, or IPOP, as a cash-settled linear perpetual contract used to trade the market’s expectation of a company’s public equity value after listing.
The CXMT contract tracks the U.S. dollar value of one ordinary ChangXin A-share, not the company’s total market capitalization. After listing, the oracle is expected to convert the company’s RMB-denominated A-share price into dollars using the exchange rate at that time.
Holding a CXMT contract does not mean holding ChangXin stock. Traders do not receive IPO allocations, dividends, voting rights, or the ability to convert the contract into real shares after listing. What longs and shorts are trading is their view of where ChangXin may price in the public market.
That makes the market on Hyperliquid a venue for trading expectations, not a venue that tells investors what ChangXin is worth today in spot equity terms.
Hyperliquid runs the venue, Trade.xyz sets the market design
CXMT was not listed directly by Hyperliquid’s core team. It was deployed by Trade.xyz through HIP-3.
HIP-3 allows third-party developers to create their own perpetual markets on HyperCore. HyperCore handles the on-chain order book, matching, margin, liquidation, and settlement. The deployer decides the underlying instrument, oracle method, leverage cap, and risk settings, and it can also execute conversion or settlement when needed.
In practice, the setup splits responsibilities across two layers:
| Function | Main party | Role |
|---|---|---|
| Limit orders and trade matching | HyperCore | Determines actual execution prices for traders |
| Initial reference price and contract parameters | Trade.xyz | Defines where the market starts and how it operates |
| Internal oracle and part of mark price inputs | Trade.xyz Relayer | Turns order-book information into funding and liquidation reference inputs |
| Margin and liquidation execution | HyperCore | Manages position risk based on mark price |
So while trading and settlement happen on a fully on-chain order book, the pricing method and oracle updates still depend on Trade.xyz’s market design and relayer.
The first $5 was not discovered by the market
Every price-discovery process needs a starting point.
Under the CXMT contract parameters, Trade.xyz set the initial reference price at $5. The company describes it as a “discretionary reference price,” which means the platform chose it itself. It also says the figure is not a prediction of the IPO issue price, the opening price, or the post-listing stock price.
Trade.xyz did not disclose which financing round, valuation model, or institutional quote, if any, was used to arrive at that number.
That means the contract did not start from a purely free-form market-clearing process. Trade.xyz first placed the origin point at $5, and only after that did the order book begin shaping the market around it.
That distinction matters. The deployer supplied the initial coordinate. Traders then supplied most of the subsequent price movement.
Executed prices come from the order book
Once the market went live, traders and market makers could place bids and offers in HyperCore’s central limit order book. Matching follows price-time priority.
If buyers are willing to pay more and sellers are not willing to sell at the current level, the execution price rises. The reverse is also true. Before ChangXin lists, there is no external spot market continuously producing a tradable reference price for the stock, so there is no outside oracle forcing CXMT to stay near the IPO price of RMB 8.66.
That is very different from a typical crypto perpetual. If a BTC perpetual strays from BTC spot prices across multiple exchanges, arbitrageurs can buy one side and sell the other to pull prices back together. Before listing, CXMT has neither freely tradable spot shares nor deliverable stock inventory, so traders cannot run the same kind of riskless arbitrage.
As a result, the contract can stay above or below the IPO price for extended periods. It reflects the expectations of participants in the order book rather than a price that can be verified through spot arbitrage.
Why the price paused at $6, $7.2, and $8.64
CXMT uses a 20% discovery bound.
With the initial reference price set at $5, the mark price could move only between $4 and $6. When persistent buying pushed the internal oracle close to the upper limit, the system could re-anchor the reference to $6, making the next band $4.8 to $7.2.
If price strength continued, the reference could move again to $7.2, which made the next upper bound $8.64.
The mechanics are straightforward:
- First upper band: 5 × 1.2 = $6
- Second upper band: 6 × 1.2 = $7.2
- Third upper band: 7.2 × 1.2 = $8.64
CXMT allows up to seven upward and seven downward re-anchors. So the 20% figure is only the instantaneous band around the current reference price. It does not mean the contract is permanently trapped within 20% of the original $5 level.
That also explains the step-like structure seen early in CXMT trading. The levels at $6, $7.2, and $8.64 were not organic technical resistance points. They were contract-defined rails derived directly from the market parameters.
The design reduces the risk that a newly launched market gets ripped higher or crushed lower by one trade. But it also changes how the chart should be read: when the market touches a bound, the visible pause may reflect a system limit rather than a natural equilibrium between buyers and sellers.
The oracle is not the last traded price
CXMT has no external stock price feed to reference before listing, so it uses Trade.xyz’s internal oracle.
According to Trade.xyz’s oracle documentation, the system first calculates the order book’s impact bid price and impact ask price, meaning the average executable price for a specified notional amount on each side of the book.
It does not simply take the best bid, best ask, or last trade. Instead, it checks whether the executable order-book prices as a whole sit above or below the current oracle:
- If executable bid-side prices are broadly above the oracle, the internal oracle moves up gradually.
- If executable ask-side prices are broadly below the oracle, the internal oracle moves down gradually.
- If the oracle still sits between the bid and ask sides, the adjustment can be zero.
The update is then smoothed with a continuously weighted exponential moving average using a 30-minute time constant.
So trades can jump quickly while the oracle moves more slowly. Only if stronger bids or offers persist does the oracle catch up. Compared with using the latest trade directly, this approach is less sensitive to one small anomalous print.
But the oracle is not an independent fair-value estimate. It is a lagged and smoothed representation of the order book itself.
Mark price drives liquidation, but it is not the same as trade price
Beyond the trade price and the oracle, the system also uses a mark price to calculate unrealized profit and loss, margin, and liquidation.
Trade.xyz says the mark price is the median of three inputs:
- the internal oracle price;
- the oracle plus a 150-second EMA of the perpetual midpoint’s deviation from the oracle;
- the median of the best bid, best ask, and last trade.
Taking the median of three inputs lowers the odds that one bad print directly triggers liquidations. Each relayer update is also limited to within ±0.5% of the current price to slow abrupt jumps.
That leaves several different prices coexisting in the CXMT market:
| Price | Main use |
|---|---|
| Trade price | The actual execution level where traders buy or sell |
| Oracle price | Reference for funding and an input into mark price |
| Mark price | Used for unrealized PnL, margin, and liquidation |
| IPO price | The real stock’s issue price, which does not directly constrain the on-chain contract right now |
When people say “CXMT reached $8,” they are usually referring either to the execution price on the order book or the mark price shown on the interface. In volatile conditions, the two do not have to match exactly.
Funding dampens deviations, but does not anchor the contract to fair equity value
A standard perpetual uses funding to push the contract back toward spot. CXMT has no spot market before listing, so funding can only compare the contract price with the internal oracle.
Trade.xyz set the IPOP funding multiplier at 0.005, while regular XYZ stock perpetuals use 0.5. In other words, the funding intensity for the pre-IPO contract is only about 1% of that used in standard XYZ stock perps.
Funding still exchanges between longs and shorts every hour, but its effect is much weaker. One reason is that the internal oracle itself comes from the same order book. If funding were too high, a one-sided position could become expensive to hold over a long wait for the IPO.
Lower funding lets traders hold views on the eventual listing price for longer periods. The trade-off is that the market lacks a strong outside anchor. Funding can damp short-term deviations between the contract and the internal oracle, but it cannot determine whether ChangXin should be worth RMB 500 billion or RMB 3 trillion.
External stock pricing takes over only after listing
Once ChangXin starts trading normally on the STAR Market and enough external market data exists, CXMT is expected to convert into a standard stock perpetual.
At that point, the oracle would use ChangXin’s A-share price as the base and convert it into U.S. dollars using the prevailing RMB exchange rate. The funding multiplier would also revert from 0.005 to 0.5, the level used for regular XYZ stock perpetuals.
That conversion may be the market’s real stress test.
If the post-listing stock price is close to the on-chain contract price, the switch from the internal oracle to the external oracle could be relatively smooth. If there is a meaningful gap, the mark price may jump during the transition, changing PnL and potentially triggering liquidations.
Trade.xyz set July 27 as CXMT’s expected listing date, with a grace period extending to Sept. 25 at the latest. If the IPO is delayed for a long time or canceled, the contract can by default settle based on a full-period TWAP from launch to settlement. But in the event of an acquisition, a material adverse event, or other special circumstances, Trade.xyz also reserves room to use other settlement methods.
How useful is this kind of price discovery?
A previous Trade.xyz launch, the Cerebras pre-IPO contract, offers one data point. According to Talos, the Hyperliquid contract’s VWAP in the final hour before Cerebras opened on Nasdaq was about $354.54, versus an actual opening price of $350, a gap of about 1.3%. Its IPO issue price, however, was only $185. That suggests the contract was closer to the eventual public-market opening level than to the underwriters’ issue price.
One case is not enough to prove the mechanism can price every unlisted company in a stable and accurate way.
The CXMT market has several limits in particular:
- Before listing, traders cannot directly arbitrage spot shares against the contract, so mispricing may persist.
- On-chain participants are not the entire capital market, and prices can be affected by a small number of large accounts, crypto-native risk appetite, and market-liquidity structure.
- The discovery band shapes short-term market action, so stair-step breakouts are not purely the product of natural supply and demand.
- Trade.xyz sets the initial price and contract parameters while also calculating and submitting oracle inputs, so deployer risk cannot be ignored.
For that reason, the CXMT price on Hyperliquid is better understood as a public, continuous expectation indicator backed by real money and real carrying cost, not as a definitive statement of what ChangXin will trade at after listing.
Bottom line
Hyperliquid did not conjure up the stock price of an unlisted company out of nowhere.
More precisely, Trade.xyz set a $5 starting reference and a market rule set. Traders then expressed real bids and offers through the on-chain order book. The internal oracle smoothed order-book impact prices into a funding reference. The mark price handled margin and liquidation. The discovery bound limited how fast price could move in a short time. Once the actual stock begins trading, the external A-share price is expected to take over the oracle.
In that sense, Hyperliquid’s pre-listing “share price” is the product of a manual starting point, order-book supply and demand, internal-oracle smoothing, funding-based damping, and a moving price band.
It is not ChangXin’s real stock price, and it is not a simple translation of the IPO issue price. It is a market-based wager, expressed in actual capital, on what one share may be worth once public trading begins.