Hook
On October 15, Bitget announced perpetual contracts on FASTRETAILING. The disclosure listed four parameters: 20x leverage, 24/7 trading, USDT settlement, and a product line that now spans 356 stock contracts. It did not list a fifth. Where does the price come from when the underlying market is closed?
That omission is not a footnote. It is the entire product.
A perpetual contract on a stock is not a stock. It is a synthetic price exposure, cash-settled, with no delivery, no dividend, no voting right, and no claim on the issuer. Everything the holder receives is a function of one number: the index price the exchange publishes. Everything the holder risks is a function of how that number behaves when the reference market goes dark โ Friday 16:00 in New York, a Japanese holiday, a typhoon closure in Tokyo. The product is marketed on the hours it trades. The risk lives in the hours it cannot verify.
Context
Stock perpetuals are a mature, unremarkable category. Bybit runs them. Gate.io runs them. Bitget now runs 356 of them. The mechanism is copied wholesale from crypto: a funding rate tethers the contract to a reference index, a mark price drives liquidations, and the exchange sits as central counterparty. No new cryptography. No protocol-layer innovation. This is product engineering, not protocol engineering, and it should be read that way.
The precedent that matters is Binance's tokenized stock program. It launched in 2021 and was pulled within months after BaFin and the FCA signaled that offering German and British retail exposure to equity derivatives without local authorization was not something a Seychelles entity could accomplish by relabeling the wrapper. Bitget's structure is the same: offshore incorporation, global user base, restricted-jurisdiction geoblocking, USDT settlement.
That history tells you where the regulatory ceiling sits. It does not tell you what happens between now and enforcement. Enforcement is slow, jurisdiction-by-jurisdiction, and mostly retrospective. The mechanism, by contrast, is live on the next tick.
Core Analysis
Start with what a stock perpetual actually is. It is a contract that pays the difference between entry price and exit price in USDT, with no underlying asset moving anywhere. The exchange needs no shares. It needs no borrow. It needs a number.
The entire risk surface of a cash-settled equity perpetual is the integrity of that number.
In crypto perpetuals, the index is built from multiple spot venues that trade continuously โ deep order books, no closing bell, arbitrageurs who can hit every constituent feed simultaneously. When I spent four months inside 0x v2's order-matching logic in 2017, the three race conditions I found were all front-runnable, and all of them existed for the same reason: the engine trusted a state snapshot that could be invalidated between validation and execution. A stock perpetual carries the same class of defect at a larger scale. The index is a snapshot of a market that stops producing snapshots.
Here is the structural problem. FASTRETAILING, if it trades on a Japanese venue, is priced roughly 09:00โ15:00 JST, five days a week, minus holidays. Bitget's contract is priced 24/7. Those two statements cannot both be fully true. Between Friday's close and Monday's open, the perpetual has no reference market, no arbitrage path, and no convergence mechanism. It has a funding rate โ slow, capped, absorbable โ and it has whatever the exchange decides the index should be.
Three consequences follow. None of them appear in the announcement.
Gap risk is transferred, not eliminated. A trader holding 20x leverage into a weekend carries an exposure that cannot be closed at a fair price. If the underlying opens 8% lower, the position liquidates at the open โ in a market where, at the instant of liquidation, no bid existed on the reference venue. The liquidation engine will price off an index that is stale or extrapolated. Either way, the loss is determined by a number the trader could not observe, could not contest, and could not have exited.
The index composition is the attack surface. With 356 contracts, maintaining independent multi-source feeds for every symbol is expensive, and the rational engineering choice is consolidation: fewer vendors, fewer sources per symbol, more interpolation. That is precisely where manipulation gets cheap. A thin Japanese small-cap carrying a 20x perpetual can be pushed at the open, and the perpetual liquidates the crowd before the cash market corrects. In my 2021 review of ERC-721A metadata storage, the finding that received the least attention was centralization: five collections, one Merkle root each, single points of failure nobody had audited because they were "just storage." The oracle here will be the least scrutinized and most consequential component of the entire product line.
Corporate actions break the contract's continuity. A dividend is a price drop with no economic loss. A split is a discontinuity with no economic event. A perpetual tracking a raw price series registers both as real moves, triggers funding, and in leveraged accounts triggers liquidations. Handling it correctly requires adjusting either the index or the position. The announcement specifies neither. This is not hypothetical โ it is the most common source of disputes in synthetic equity products, and it is a solved problem only where someone has bothered to write down the solution.
Then there is leverage. Regulation T sets initial margin on US equities at 50%, roughly 2x. Bitget offers 20x on instruments whose daily volatility is a fraction of Bitcoin's โ but whose Monday-open gap volatility is not. Leverage on a continuous market and leverage on a discontinuous market are different products wearing the same number. A 20x crypto perpetual can be exited in seconds. A 20x equity perpetual held through a weekend is a position with no exit and a binary outcome. The geometry is inverted: less volatility, more tail.
Now the counterparty question. Someone is on the other side of 356 equity perpetual books. If Bitget hedges externally, it needs prime-brokerage access to Japanese equities at scale โ unlikely. If it does not hedge, it runs net exposure and nets internally, which makes it the house in every trade, holding pricing discretion over the index that determines who wins. That is not an allegation of misconduct. It is a description of the trust model, and it belongs in the disclosure.
What a correct specification would contain is not exotic. Multi-source index construction with per-source weight caps. Corporate-action adjustment tables published before listing. Weekend and holiday margin multipliers, raised mechanically rather than at the exchange's discretion. Funding-rate caps that tighten when the reference venue is closed. Mark-price bands that refuse to liquidate against an index more than N minutes stale. Every one of these is standard in regulated futures markets. None can be verified from the outside unless the exchange publishes them, and none appear in a product announcement that leads with leverage and uptime.
The 356-symbol scale carries its own second-order effect. Order books fragment across hundreds of symbols, most of which will never attract meaningful depth. A perpetual with thin liquidity and a leveraged position on the wrong side is a self-liquidation machine: the position's own exit moves the mark price, which triggers the next liquidation. In 2020, modeling Uniswap V2's constant-product curve taught me that liquidity depth is not a footnote to price โ it is the price. Here, the depth is undisclosed for every one of the 356 symbols.
I have argued elsewhere that most data-availability engineering optimizes for rollups publishing kilobytes per day while the feeds that actually determine outcomes go unexamined. This product is that argument in miniature. No one is worried about DA here. Everyone should be worried about the price feed. The system's unintended consequences will not be a data-availability failure; they will be a pricing failure, executed on schedule, every Friday at 16:00 New York time.
Contrarian Angle
The consensus risk ranking for this product puts regulation first. I would invert it.
Regulatory action is discrete, dated, and priced the moment it lands โ Binance's delisting was a headline, not a slow bleed. The oracle problem is continuous and unpriced. It generates no announcement, no enforcement action, no identifiable villain. It generates a Monday morning on which a few thousand accounts were liquidated at a price that never existed in any cash market. Regulation closes a product. Bad pricing closes accounts, quietly, one weekend at a time.
The second inversion concerns the 24/7 feature itself. It is sold as access โ you can trade while Tokyo sleeps. In practice it converts market hours into product hours, manufacturing exposure during precisely the window when price discovery is impossible. A feature that lets you trade when nobody can price is not access. It is an unhedgeable position sold as a convenience.
The growth logic behind 356 contracts deserves the same skepticism. Expansion at this scale is not evidence of user demand; it is revenue diversification under fee compression โ the same reflex that produced subsidized liquidity programs whose TVL evaporated the week emissions stopped. Volume on a 356-symbol line will look impressive until you separate maker rebates from organic flow.
Takeaway
The first systemic event in tokenized equity perpetuals will not be a regulator's letter. It will be a weekend.
Watch two signals: whether Bitget publishes index composition and corporate-action methodology for these 356 symbols, and whether funding-rate caps tighten mechanically around exchange holidays. If neither appears before the next global market closure, the product's real specification has already been written โ just not by anyone trading it.