On June 10, 2026, two days before SpaceX went public, CoinDesk reported that the most active SpaceX pre-IPO perpetual was trading around $157. The IPO had been priced at $135. When the stock opened on June 12, the first print was $150.
So what was the perp referencing for all those weeks? There was no public share price to track, and private shares weren't something a trader could buy and sell alongside the perp. The perp was pricing a company that did not yet have a public price.
That's the central question for two of the newest kinds of perpetual futures: pre-IPO perps, which reference private companies, and data perps, which reference continuous real-world signals like ship traffic. Both have become a visible part of perpification. Both break the assumption that a normal perp is built on.
What a normal perp leans on
A standard perp uses three prices, which we covered in What Is Oracle Pricing in DeFi?:
- Index price: an external reference, usually built from spot markets.
- Mark price: the price the venue uses for PnL and liquidations, typically derived from the index with some smoothing.
- Last price: where the perp itself last traded.
Funding ties them together. When the perp trades above the index, funding generally tilts toward longs paying shorts, which gives traders an incentive to push the perp back toward the reference. Arbitrageurs help: if the gap gets large, they buy spot and short the perp.
All of this assumes the index exists. For a private company or a data signal, there's no spot market to build it from, and no liquid spot asset for an arbitrageur to trade against the perp. The venue has to decide what the "real" price is some other way.
Four ways venues solve it
Public documentation from venues offering pre-IPO perps shows four broad approaches. The venues we reviewed often combine more than one.
1. Let the market price itself
The simplest answer is to build the reference price out of the perp's own order book. trade[XYZ]'s documentation is direct about it: "Before listing, there is no public equity price to use as an external oracle. The market therefore operates with an internal pricing mechanism." Its oracle is a 30-minute exponentially weighted average driven by the order book's own impact prices. Lighter and Kraken describe similar internal indices, smoothed with an exponential average.
The trade-off: the reference is derived from the venue's own trading rather than an independent spot index. It's genuine price discovery, but it reflects that venue's order flow.
2. Blend the order book with private-market data
A second approach mixes in off-chain valuation signals. Ventuals, before it wound down its markets in June, set its oracle as one-third a private-market price from a data provider (drawing on secondary transactions, bids and offers, funding announcements, and mutual fund marks) and two-thirds a moving average of its own mark price. Other designs weight the order book more heavily when it's deep and fall back to private-market marks when it's thin.
The trade-off: private-market data updates slowly and sparsely. It's an anchor, but a loose one.
3. Start from a reference price, then bound the moves
Many venues open a pre-IPO market at a reference price they choose. trade[XYZ] calls its starting level a "discretionary reference price" and says explicitly that it is "not a forecast and not an indication of a public listing price." From there, "discovery bounds" cap how far the mark price can move from a reference in one step. The reference can re-anchor a set number of times before the cap becomes hard.
The trade-off: bounds slow down runaway moves, but they also mean the displayed price can lag a genuine repricing.
4. Turn funding down or off
Without a spot index, funding can't do its usual job, so the venues we reviewed dampen it. trade[XYZ], Lighter, and Hyperliquid's pre-launch markets run funding at 1% of the standard calculation. OKX sets the premium component of funding to zero for these markets. Kraken moved to a fixed hourly rate.
This is the most important point for traders: in these designs, funding is not doing the work of pulling the price toward an external fair value, because there's no external value for it to reference. Whatever disciplines the price has to come from somewhere else.
Why the same company trades at different prices
Put those four design choices together and it's no surprise that venues disagree.
As of early September, CoinGecko tracked Anthropic perps across 19 venues trading at an average implied valuation roughly double its last funding round, with OpenAI perps across 18 venues about 42% above its last round. On a single day in September, one analysis found Anthropic's implied valuation about $150B higher on some venues than on others.
Part of that gap isn't even about opinion. Some venues quote a price per share and must assume a share count; others quote the valuation directly. When the share count is unknown, two venues can agree on the company and still show very different numbers.
Arbitrage can narrow the gap, but it's harder than usual. There's no spot asset to hold. A trader can go long on one venue and short on another, but that means holding margin on both, paying fees on thin order books, and trusting that the two contracts will settle against the same reference later.
What happens at the IPO, or if it never comes
A pre-IPO perp needs clear rules for what happens at listing, and for what happens if a listing never comes. Those rules matter as much as how it traded.
- Conversion at listing. Venues including trade[XYZ] and Kraken convert the contract into a normal equity perp once the company lists, usually from the first regular trading session. From then on it references the public stock.
- Settlement if the listing doesn't happen. Venues publish fallback rules. One common default is a time-weighted average of the market's own price over its whole life. OKX says it would settle at a price it determines in good faith.
- Settlement if the venue closes the market. When Ventuals wound down its OpenAI and Anthropic markets in June, it froze marks at a 24-hour TWAP and settled there, noting that "these markets do not have a realtime external reference price."
How close did the perps get? Sometimes very close. Coin Metrics reported that the Cerebras pre-IPO perp traded within about 1.3% of the stock's Nasdaq opening price in the hour before its debut. SpaceX's perp traded around $157 to $162 two days before listing, against a $150 open and a close near $161. Those are two data points, not a track record, and the path to them included large swings: the SpaceX perp fell about 27% in the three weeks before the IPO.
Data perps: when the underlying is a signal
Data perps take the same problem a step further. The underlying isn't a company. It's a number that updates over time.
Perp City is the clearest example. Its markets reference indices built from real-world data, such as a count of cargo ships and tankers near the Strait of Hormuz, updated every 15 minutes from AIS ship-tracking data. Its documentation describes a pipeline of data source, sensor, index function, and an on-chain "beacon" that verifies proofs before accepting an update. Compute prices are getting the same treatment: Bitget launched GPU rental-price perps in September, referencing third-party rental indices.
Two risks apply to the category as a whole:
The data can go dark. In May, maritime analytics firm Windward observed 97 vessels near the northern Hormuz corridor on satellite imagery, with only 3 of them transmitting AIS. It also reported location spoofing and GPS jamming affecting hundreds of vessels in a single day. An index that counts AIS signals is measuring reported traffic, which is not always the same as actual traffic. How much that matters depends on the index's validation and fallback rules.
There's no direct hedge. When the underlying is a signal rather than an asset, there's no spot market to hedge against; Perp City's documentation notes that "direct delta hedging is not possible." That makes inventory harder to manage for market makers and can constrain liquidity.
What goes wrong in thin, self-referential markets
Two patterns show up repeatedly when a perp's reference price depends mainly on its own trading.
Thin books can be pushed. In August 2025, a pre-launch token perp on a large on-chain venue moved roughly 3x in minutes, liquidating about $159M of short positions. The venue responded by adding external exchange prices to the mark price for that market type. Pre-launch and pre-IPO designs that rely heavily on their own order books can share this exposure. How severe it is depends on how trades feed into the mark and what safeguards apply.
One bad input can travel. In July 2026, a single thin pre-market print on a Korean trading venue implied a 28.7% drop in SK Hynix. The corresponding on-chain perp fell about 18% and liquidated roughly $57M before its price bounds stopped the move. The deployer covered the losses. The bounds limited the damage; they couldn't prevent it.
Neither case is about one team getting something wrong. Both show the pressure a price comes under when it has few independent sources to check it against. We covered the defensive side, including validator permissions and protection triggers, in Oracle Integrity, and the related problem of stale pricing.
A checklist before you trade one
Before opening a position in a pre-IPO or data perp, check five things in the venue's documentation:
- What is the reference price? The perp's own book, a blend with private-market data, or a venue-set level? If it's mostly the book, the price reflects that venue's order flow rather than an independent index.
- What does funding do? If it's dampened, fixed, or zero, don't expect it to pull the price toward fair value.
- Are there price bounds or open interest caps? Price bounds limit abrupt moves but can delay a legitimate repricing. Open interest caps limit how much exposure the market can build, and can block new positions in one direction.
- How does it settle? At listing, at a TWAP, or at a venue-determined price? Know the rule before the event, not after.
- What leverage makes sense? The venues we reviewed cap many of these markets at 3x to 5x, well below their crypto markets, for a reason. Thin books, weekend gaps, and sparse data all mean larger moves between updates.
How LeverUp approaches reference prices
LeverUp doesn't currently list pre-IPO or data perps. It uses external reference prices delivered through Pyth's low-latency pull oracle, with market-specific trading schedules and risk controls.
LeverUp's MAG7 equity perps trade Monday to Friday, 9:31 AM to 3:59 PM ET. Positions above 10x are force-closed at market close. Outside that session, opening and closing positions, placing new orders, and liquidations are all disabled; traders can add collateral, cancel orders, or edit TP/SL. Each pair has open interest caps, and for Alpha-tagged altcoin and RWA pairs, large positions carry a minimum holding period, because, as the docs note, "sophisticated actors can attempt to influence external order books and oracle prices."
The general principle behind all of this is the same one that runs through this article: a perp can only be as reliable as the price it references. When that price is thin, self-referential, or can go dark, the market's rules need to account for it, and traders need to know which rules they're trading under.
Further reading: