Who Sets the Price at 3 A.M.? Deepcoin's 7×24 Stock Perpetuals and the Question the Press Release Skipped
On September 10, Deepcoin published a product note announcing global stock perpetual contracts — twenty-four hours a day, seven days a week. The names on the list were NVIDIA and Tesla, which is predictable, and Pop Mart and Yushutech, which is not. There was a temporary 25% trading fee discount and three launch campaigns: a stock trading championship, a sector trading challenge, and a signal-provider leaderboard. The framing sentence was "completed upgrade of multi-asset trading infrastructure."
I read it twice, then did what I always do with a derivatives announcement: I opened a clock. New York closes at 16:00 ET. Between Friday's close and Monday's open, roughly sixty-four hours pass in which no regulated exchange on earth prints a real price for NVIDIA. Deepcoin is offering to trade NVIDIA through all of them.
So at 03:14 Hong Kong time on a Sunday, who prints the price?
The announcement never says. No index methodology. No quote source. No market maker structure. No funding rate formula. No liquidation engine description. No jurisdiction list. No audit. No reserves language. No team disclosure. Behind every hash, a heartbeat — and this time the heartbeat is a quote tick that nobody has been asked to verify.
Context: What Is Actually Being Sold
Deepcoin is a mid-tier centralized exchange. That is not an insult; it is the single most important fact for reading the announcement correctly. When Kraken launches tokenized equities through Backed, or Robinhood ships stock tokens under a European framework, or Bybit expands its derivatives book, the market already knows what those firms are and what they can absorb if something breaks. When a smaller venue announces that it has "upgraded its multi-asset trading infrastructure," the sentence is doing a lot of work that no paragraph beneath it supports.
There is a real trend underneath this, and it deserves respect. Tokenized and synthetic equity exposure became a genuine growth category across 2024 and 2025. European retail got access to stock tokens through licensed wrappers. On-chain structures like xStocks gave a legal claim to a custodied share, wrapped in a token, settleable on a public chain. Traditional contracts-for-difference brokers — IG, Plus500 — have sold leveraged equity exposure for two decades under mature supervision. The category is not a fad. The category is a migration.
But a perpetual contract is not a tokenized stock, and conflating them is where retail gets hurt. A perpetual has no expiry. It stays open indefinitely and is tethered to reality by a funding rate: periodically, longs pay shorts or shorts pay longs, based on the gap between the perpetual's traded price and an index that represents the asset's "true" price. For Bitcoin, that index is trivial — dozens of deep, continuous spot venues. For NVIDIA, the index is a philosophical problem. The real market has hours. The perpetual does not. Something has to fill the gap.
What fills the gap is the entire product. Everything else — the leverage tiers, the fee discount, the leaderboards — is packaging.
Core: The Three Clocks Problem
Start with index construction, because every liquidation on the platform ultimately resolves against it. In a closed session, a synthetic equity index can be built in only a handful of ways. You can use the last traded price, which is stale and therefore trivially gameable. You can use a single market maker's quote, which is a private opinion wearing a public number. You can build a fair-value model off correlated instruments that do trade overnight — index futures, ADRs, sector ETFs — which is the most defensible approach, and the most opaque. Or you can do some blend of the three and rebalance it with parameters nobody outside the firm can see.
Every one of those choices produces the same structural fact: during the hours when the underlying market is closed, the price you are liquidated against is a proprietary number.
This is not a hypothetical concern. In crypto, we have spent a decade watching thin books get pushed through liquidation clusters. The mechanics are well documented, and nobody disputes them anymore. The difference here is that crypto liquidations at least occur against a price that exists in many places simultaneously. A closed-session equity mark exists in exactly one place: the platform that publishes it. If you are levered into that number, your counterparty and your oracle are the same entity.
Now add the funding rate. Funding exists to arbitrage the perpetual back toward the index. It is a corrective mechanism that assumes the index is external and trustworthy. If the index is itself a model, then funding is a self-referential loop: the contract pulls toward a number that was derived from the contract's own trading environment. There is no arbitrageur standing outside the system with a better price, because the better price does not exist until Monday. Any premium that forms on a Saturday night is unfalsifiable, and unfalsifiable premiums do not correct. They compound.
Then there is the mark price used for liquidation. In most perpetual designs, liquidation triggers on the mark, not the last trade, precisely to prevent manipulation through wick-thin prints. That safeguard assumes the mark is anchored to something real. Anchor it to a model, and the safeguard becomes the attack surface. Someone who understands the model's inputs can move the mark by moving those inputs.
Which brings us to the counterparty question. Most mid-tier exchanges do not purely match orders; they internalize a share of flow and act as the counterparty to their users. In that architecture, user losses are platform revenue. I want to be careful here, because internalization is legitimate, disclosed, and standard in regulated markets — every retail broker does a version of it. But it requires supervision, reporting, and a best-execution obligation. When internalized flow meets an index that the platform itself constructs during the only hours when the underlying cannot contradict it, the conflict stops being theoretical. It becomes a design property.
The listing choices say the rest. NVIDIA and Tesla are the global retail aspiration trade. Pop Mart is a Hong Kong-listed toymaker that delivered one of the most violent momentum rallies of the past two years. Yushutech is a recent US listing that drew exactly the kind of attention that makes a derivatives desk salivate. These are not allocation instruments. They are churn instruments. Picking them is the most honest disclosure in the entire announcement: it tells you the platform is optimizing for turnover, not for portfolios.
And the incentive architecture confirms it. A temporary 25% fee discount plus three separate campaigns — a trading championship, a sector challenge, and a signal-provider leaderboard — is a volume-acquisition program. None of those mechanisms rewards holding, hedging, or learning. All of them reward frequency. The signal-provider leaderboard deserves particular attention because it formalizes the copy-trading funnel: rank the loudest narrators, let followers mirror them, and split the fees. I have watched that funnel from the inside. In 2017, my team at Ethos Ledger interviewed 120 first-time investors who had lost savings. Almost none of them were destroyed by a chart. They were destroyed by a person narrating a chart. The instrument was secondary.
There is also a quieter accounting point. The 25% discount is described as temporary. Volume acquired under a subsidy is not volume; it is rented liquidity, and it leaves when the discount does. Temporary pricing tells you the platform expects retention to be earned elsewhere. Nothing in the announcement explains where.
Now the part I keep returning to. This product is presented as a step toward multi-asset infrastructure, which places it adjacent to the real-world-asset narrative. I have been skeptical of that narrative for three years, and this announcement illustrates why. The institutions that can actually hold equities do not need a public chain to sell equity exposure. They need a license, a clearing member, and a prime broker. Tokenization solves a settlement and composability problem for people who already have legal claims; it does not solve a distribution problem for people who lack them.
A genuine tokenized equity requires three things: an issuer, an identified custody chain, and a legally enforceable claim on the underlying share. Deepcoin's note contains none of the three. That leaves two possibilities, and both are disclosures. Either the underlying is not held — in which case this is a synthetic contract-for-difference with a crypto interface, which is fine but should be named that way — or it is held, in which case someone holds it, and that someone has never been named. Trust no one, verify everyone, feel everyone; that last clause is the one the industry keeps skipping.
On reserves, the caution is familiar by now. Merkle-tree proof of reserves is a snapshot: it proves a subset of liabilities at a single instant, it does not prove liabilities are complete, and it is not continuous. A snapshot is not a solvency proof; it is a photograph of a moving object. Deepcoin's announcement contains no reserves language at all. Absence of a claim is not evidence of wrongdoing. It is, however, the cheapest disclosure on earth to add, which makes its absence a choice.
Finally, the technical signal that is hardest to argue with: no audit, no repository, no testnet transition, no latency figures, no depth figures, no fee tier table. For a centralized exchange product, secrecy about architecture is normal. But if a firm describes its work as an infrastructure upgrade, it has implicitly accepted the metrics that word implies. None were given. What was given was a discount.
Contrarian: The Gray Zone Is a Market, Not an Accident
The comfortable reading of this event is that a small exchange chased a hot narrative. That reading is correct and uninteresting. The uncomfortable reading is that Deepcoin is serving a segment the licensed world structurally cannot serve, and it is doing so because the demand is real.
Think about what Robinhood, Kraken, and the European brokers cannot offer. They can offer equity exposure. They generally cannot offer high retail leverage on single names, and they cannot offer weekend pricing, and they cannot list the momentum names that Asian retail actually wants to trade. The regulated perimeter is a perimeter precisely because it excludes those things. So the gray zone is not an accident of enforcement lag. It is a market that exists because the rules forbade its legal version.
That reframes the fix. The instinctive response is more regulation, but regulation at the boundary just pushes the boundary. The response that actually reduces harm is disclosure at the product level: publish the index methodology, publish the closed-session quote logic, publish the mark and funding formulas, and let people choose with their eyes open. Code is law, but empathy is truth — and the truthful version of this announcement would have been three paragraphs longer and infinitely more useful.
Takeaway
The event to watch is not the launch. It is the first weekend when a closed-session mark price gaps several percent away from reality and a wave of positions liquidates against a number no external market ever confirmed. That weekend will define this category, in the same way a single depeg defined algorithmic stablecoins. Until then, the useful question for anyone holding this product is not whether Deepcoin is credible. It is whether they can name the entity that prints the price while New York sleeps. If the answer is "the exchange," they already know their counterparty. Surviving the winter to plant the spring has always required knowing which season you are standing in.
