The Fee Stack Bites First
You don't need a securities lawyer to spot the problem. You just need a calculator.
A $110 subscription price on a Moonshot AI pre-IPO certificate. The underwriting fee is 5%. You lose 4.76% the moment your capital converts into a claim. Before Moonshot AI's stock moves one tick. Before the IPO bell rings. Before the secondary market even opens, you are down nearly five percent.
Then you sell. Taker fee: 1.5%. Special market fee: another 1%. Stacked on top of the 5% you already paid. And if the thing appreciates, a 20% performance fee takes a fifth of your upside. Add it up: roughly 7.5% in frictional costs. Break-even requires an 8.1% move — in a company you cannot verify, through a vehicle you cannot audit, on a platform that decides its own final terms "based on actual circumstances."
In a consolidation market, new narratives are treated like oxygen. This one wraps the AI trade — Kimi, the Chinese LLM unicorn — inside a real-asset securitization story, then sells it through the distribution machine of an exchange with fifty-eight million registered users. The package is seductive. The plumbing is not. I've spent years reading fee schedules and smart contract bytecode. The fee schedule is the smart contract here. Nobody audited it because nobody has to. Gate builds the product. Gate prices the product. Gate operates the marketplace. Gate decides what happens when the trigger fires. Let me show you what is actually being sold.
What Exactly Is a Mirror Note?
Gate calls this Phase 3 of its Pre-IPO product line. Phase 1 and Phase 2 have already run. The structure on its face is simple: a synthetic instrument that mirrors the market capitalization of Moonshot AI — the company behind the Kimi AI assistant — until that company completes an IPO. The implied valuation: $50 billion. Subscription prices: $105 to $115 per unit. Minimum ticket: 10,000 USDT or GUSD.
The valuation anchor matters. A $50 billion pre-money for a company that has raised significant private capital is a late-stage print, not an early punt. Anyone buying at this level is paying for the 2026 IPO narrative, not for discovery. And the announcement's own range — $105 to $115 — is a ten-dollar spread on a two-digit price, roughly nine percent wide. A pricing band that wide is not a price. It is a menu.
The instrument is a Mirror Note. In traditional finance, that maps to a contingent-payout note: synthetic exposure to an underlying asset without holding the underlying asset. The distinction matters more than most retail participants realize.
This is not equity. It is not a tokenized share. There is no ERC-20 contract to inspect on Etherscan. No on-chain record of who holds what claim. There is a database entry inside Gate's infrastructure. The mirror reflects a valuation Gate assigned, using a fee schedule Gate designed, settling into a market Gate operates. Gate says it serves 58 million users and publishes a 100% reserve proof. That proof is company-wide. It says nothing about whether the underlying Moonshot AI equity exists, who holds it, what legal priority a certificate holder has in a dispute, or whether the "mirror" has anything to reflect.
Moonshot AI itself is a serious entity. Kimi is one of the strongest Chinese LLM products, the team carries Tsinghua DNA, and the company has raised substantial capital. None of that makes the certificate real equity. A reference asset's quality does not transfer to a derivative's legal standing.
Scan the competitive field. Ondo Finance tokenizes Treasuries and money-market funds with compliant issuance and smart-contract enforcement. Backed Finance issues tokenized listed stocks on-chain with audited custody. Forge Global and EquityZen sell actual pre-IPO shares under SEC oversight, usually with fee structures below 5%. FTX sold "tokenized stocks" too — centralized records, no real delivery, and the entire stack collapsed with the platform.
Pre-IPO access in traditional finance is a restricted privilege. Accredited investors, relationship brokers, negotiated allocations, lock-ups, FINRA oversight. The average trader cannot touch a fifty-billion-dollar AI unicorn before its listing. That scarcity is the product. Gate is selling the illusion of access — and charging for access to an illusion. The wedge between what retail thinks it is buying — a chance to own Kimi early — and what the instrument actually is — a claim on Gate's promise to pay if its own conditions are met — is where the fee stack lives.
Gate's product occupies a strange middle. Forge's ambition without Forge's regulation. Backed's wrapper without Backed's verifiability. FTX's structural opacity with a fee schedule that makes FTX's old product look like a discount broker. Now let me price the thing.
Core: The Microstructure of a Synthetic
The fee stack, fully loaded.
The 5% underwriting fee is charged at subscription. On a $110 median price, that is $5.50 per unit. Your real cost basis is $115.50 before any price movement. The fee is certain. The upside is not. The refund language for failed IPOs does not restore the time value of your capital while it sits locked in the subscription window. In options terms, this is an OTC call option where the premium is extracted before you see the strike — and the strike is whatever Gate's database says it is.
The 20% performance fee on "excess returns" is standard for alternatives. But the trigger benchmark is undisclosed. Is it measured from the subscription price? From the IPO price? From Gate's internal mark? The announcement is silent. Silence in a fee schedule is a design choice, not an omission.
The trading tax. Maker fee in the dedicated secondary market: 0.5%. Taker fee: 1.5%. Plus a 1% special market fee. Standard taker fees on major centralized venues run near 0.075%. This product charges twenty times that — inside a venue with no external liquidity and no independent price discovery. The fee is a premium for captivity, not for service.
The all-in math says more than the marketing.
5% underwriting plus 2.5% trading costs equals 7.5% round-trip friction. Break-even requires an 8.1% appreciation in a pre-IPO synthetic with zero independent price discovery. Work the numbers at the $10,000 minimum. Subscription at $110: 90 units. Underwriting fee: $500. Cost basis: $10,500. If the dedicated market opens at $115 and you exit, gross proceeds: $10,350. Taker fee at 1.5%: $155. Venue fee at 1%: $103. Net received: $10,092. You just lost $408 on a trade where the reference asset gained 4.5%. The seller does not even need the stock to fall. The seller only needs the venue to open.
I know what this kind of friction does to a trade. In 2021, I ran a custom Python script arbitraging price discrepancies between Uniswap V3 and SushiSwap. Four hundred and fifty micro-trades in a day, netting around $28,000. The edge on each trade was a few basis points. A round trip with 7.5% costs would have made the entire exercise a donation. The KIMI certificate's fee stack is not an inefficiency you can exploit. It is the exploit.
The subsidy is a demand-generation expense, not a yield.
Unallocated subscription funds earn 3.8% APR. GUSD gets daily distribution. USDT gets hourly snapshots. Sounds like a hedge against a failed allocation. It is. But 3.8% is below the 4% to 5% available from mainstream Treasury-backed RWA products in the same period. Gate is not paying you a yield premium. Gate is paying you a small discount on the opportunity cost of parking 10,000 USDT in a subscription window with no guarantee of allocation. That is customer-acquisition spend, not investor alignment. The difference appears the day the IPO timeline slips by a year. Meanwhile, the capital sits inside Gate's balance sheet. If Gate is running a Treasury position beneath that 3.8%, the spread between the actual yield and the paid yield is Gate's revenue. The subsidy is not a gift. It is a narrower bid.
Why GUSD? The choice is deliberate. USDT is the dominant stablecoin, but Tether's reserve opacity has been a nagging narrative for years. GUSD carries a New York trust-company charter. Selecting GUSD gives the product an aura of regulatory hygiene without obligating Gate to submit the certificate itself to any regulator. It is a signaling move, and the signal is aimed at the retail compliance officer inside the buyer's head. The 3.8% APR on GUSD also implies Gate is holding actual short-term Treasuries behind the product's idle funds, or has a partnership that does. That is fine. But none of it addresses the core question: what happens to the claim itself if Gate's discretionary decisions go wrong?
How the dedicated market actually works — and doesn't.
The secondary market opens about a month after distribution. Participation is limited to Gate's KYC user base. There is no cross-listing. No market-maker commitment is disclosed. No historical depth data from Phase 1 or Phase 2 is published. From a microstructure view, this venue is a captive pool with a toll booth at both ends. The announcement never answers the only question that matters for price discovery: what happens when a large holder wants to exit? In any real pre-IPO market, exits happen through negotiated transfers, rights of first refusal, and structured secondary rounds. Here, they happen inside a database, at prices Gate's matching engine produces, taxed at 2.5% per crossing.
I studied this kind of settlement lag after the Bitcoin ETF approvals in January 2024. I spent weeks correlating BlackRock's IBIT and Fidelity's FBTC creation and redemption windows against on-chain BTC movement. The pattern was clear: large OTC desk sales preceded ETF spot purchases by roughly fifteen minutes. That lag is a settlement artifact — a structural interval created by how institutional mechanics interact with crypto markets. It is predictable. It is exploitable. The KIMI certificate's dedicated market has no external settlement event. No creation window. No redemption mechanism. No arbitrage that can force the mirror price back toward reality. The only anchor is Gate's book. That is not a market. It is a price.
Counterparty forensics: where is the other side of the trade?
If these certificates mirror Moonshot AI equity, someone must hold actual shares. Gate does not say who. No independent custodian is named. No third-party audit of share ownership is referenced. No disclosure of whether a pre-IPO fund, a GP, or a secondary seller provided the offsetting position.
In a properly constructed synthetic, the issuer holds the underlying asset or a hedge. If Gate purchased Moonshot AI shares to back these notes, the terms of that purchase — rights of first refusal, lock-ups, transfer restrictions — determine whether the mirror ever reflects anything. A single ROFR trigger could force a repurchase of the underlying shares. Your certificate is cancelled "in accordance with applicable rules." Translation: Gate decides. The product's own language hands all interpretive power to the party that collects the fees.
I have audited this exact failure pattern. In May 2022, during the Terra collapse, I spent 72 hours tracing Anchor's oracle interactions on Etherscan. The kill mechanism was not complexity. It was a trust assumption hidden inside a price feed: the system assumed the oracle would keep reporting accurately under stress. When that assumption broke, the death spiral executed itself. The KIMI certificate runs on the same architecture: it assumes Gate's internal records, Gate's custody arrangement, and Gate's discretionary judgment all function perfectly for the full life of the instrument. That is not a protocol. That is a counterparty. And counterparties have balance sheets, moods, and lawyers.
The legal architecture: an IOU wearing a suit.
Run the Howey test. Money invested: yes — 10,000 USDT minimum. Common enterprise: yes — funds pooled toward one AI company's pre-IPO exposure. Expectation of profits: yes — the entire pitch is early access to appreciation. Efforts of others: yes — Moonshot AI's engineers and the IPO bankers do all the work. Four for four.
The use of GUSD — a New York-regulated Gemini stablecoin — creates compliance theater. GUSD's regulatory status does not transfer to the certificate. The product is an unregistered claim on future equity value, offered globally through a centralized exchange, without a named issuing entity, without a license citation, without an explicit US exclusion. The geo-blocking question is glaring: if US residents are barred, say so. The silence reads like a compromise between marketing reach and legal avoidance.
The conflicts compound. Gate sets the subscription price. Gate charges the underwriting fee. Gate operates the venue and charges both sides. Gate assesses the performance fee. Gate decides post-IPO processing "based on actual circumstances." Issuer, market maker, listing venue, settlement agent — one entity, four roles, no firewall. In any regulated market, that overlap is illegal, or firewalled, or disclosed so thoroughly that the disclosure becomes the document. Here, it is a feature with a fee attached.
What Phase 1 and Phase 2 actually tell you.
The product is labeled Phase 3. That means Gate has run this playbook before, presumably with smaller private companies, and has kept no public audit trail of outcomes. No settlement histories. No realized distribution rates. No data on what Phase 1 certificate holders actually received. Call it what it is: a pilot program whose accounting records have never been opened to the public. If Phase 1 and Phase 2 settlements went cleanly, releasing the data would be free marketing. The absence of that data is itself a data point. In markets, silence is information, and here the information is that Gate is not eager to show historical holders' realized outcomes.
Now list what the announcement does not say. The legal entity issuing the certificates. The governing law of the instrument. The dispute-resolution forum. The custodian of the underlying equity. The existence of any insurance or fidelity bond. The allocation ratio for oversubscribed rounds. The historical settlement performance of Phase 1 and Phase 2. The identity of any third-party counterparty. The trigger definitions for the ROFR clause. Every one of these items is knowable. None is disclosed. In a traditional offering document, the absence of any one of them would keep the deal from pricing. Here, the absence is the business model. The buyer is asked to bridge the information gap with trust, and trust is precisely the asset the fee schedule monetizes.
Contrarian: Retail Sees Kimi. Smart Money Sees a Counterparty.
The retail narrative is simple: AI unicorn, $50 billion valuation, Kimi assistant, get in before the IPO. The smart-money read smells different. Walk the fee engine from Gate's side.
On a single $10,000 subscription, Gate books $500 in underwriting fees the moment the order fills. Guaranteed. If the certificate appreciates, Gate takes another 20% above the benchmark. If the holder wants out, Gate charges up to $250 on a hypothetical $10,000 sale — 1.5% taker plus 1% venue fee. Multiply across a hundred thousand subscribers, and Gate books $50 million in upfront fees before a single successful IPO proves anything. The product is a toll road where the tolls are paid regardless of whether the destination exists. That asymmetry — certainty of fees versus uncertainty of outcome — is the definition of a good business, and it is the definition of a bad investment for everyone on the other side.
Run the three outcome scenarios.
Scenario one: Moonshot AI IPOs successfully above $50 billion. Holders get paid. Minus the 20% carry. Minus exit fees. Minus whatever "follow-up arrangements" Gate decides. The announcement's phrasing — Gate will handle subsequent asset processing based on actual circumstances — does a lot of legal work. Actual circumstances for whom? Gate, or the holder? The ambiguity is not accidental.
Scenario two: the IPO is delayed two years. Capital sits in a captive venue with no external liquidity. The only exit is a sale into Gate's KYC pool, taxed 2.5% to leave. The 3.8% subsidy ended at subscription. Opportunity cost compounds against the holder. In a sideways market — the one we are in — this scenario is not rare. It is the base case.
Scenario three: the IPO never happens. Moonshot AI is Chinese-founded, operating in a geopolitical current. A regulatory intervention, a funding gap, or an AI correction ends the story. Refunds arrive "in accordance with applicable rules." Anyone who traded through 2022 knows what that language sounds like on the way down.
Now the valuation. $50 billion pre-money is not early access. OpenAI's 2025 valuation band hovered near $300 billion. Anthropic sat in a similar range. Moonshot AI is credible — but $50 billion pre-IPO is late pricing with early-stage execution risk. The fee stack makes the risk-adjusted case strictly worse than the headline.
Consider what a successful IPO actually does to this product. The day Moonshot AI lists on a real exchange, the mirror becomes obsolete. Why hold a synthetic claim — burdened with a 20% carry and captive-venue exit costs — when the actual stock trades on Nasdaq or the Hong Kong exchange with real depth and settlement? The certificate's only exit event is the event that destroys its reason to exist. If the IPO succeeds, the certificate's value converges to the underlying minus the fee stack, and the rational holder sells immediately. That rush to the exit collides with a venue that charges 2.5% for the privilege of leaving. The result is a predictable sell-off in the dedicated market at the exact moment everyone learns the news. That is not a contrarian insight. It is a prepared trade.
I have personal data on AI narratives failing. In late 2025 I gave $50,000 to an AI-driven options agent on a decentralized exchange. Three weeks later, a regulatory announcement the model could not anticipate hit its volatility assumptions. The drawdown hit 60%. I terminated the position manually. The software was skillfully engineered and catastrophically overfitted. The dominant risk of every AI trade is not that the technology fails. It is that the market prices in a future that is not the one arriving. The KIMI certificate is that same wager with a toll booth attached. If you believe in Moonshot AI, buy the actual stock after the IPO. The synthetic is a worse deal with a worse exit.
One last structural point, from the cryptographer's seat. There is no independent verification path for this product. ZK proofs don't verify balance sheets. Arbitrage is just efficiency with a heartbeat, but you cannot arbitrage a claim that lives inside a proprietary database. Your certificate, the underlying equity, the settlement terms, the custody arrangement — all of it exists inside Gate's infrastructure. The product is a trust assumption wearing an AI costume. Trust assumptions are the oldest exploit in finance, and they do not get younger with repetition.
Takeaway: The Three Signals That Matter
Gate has distribution, KYC machinery, and balance sheet. The question is whether the structure works for anyone else. Three signals determine which reality forms.
First: does Gate name the legal holder and custodian of the underlying Moonshot AI equity before the dedicated market opens? No disclosure means no offsetting position. The mirror reflects nothing.
Second: measure the bid-ask spread and order depth during the first month of the dedicated market. A venue charging 1.5% taker on a book that cannot absorb a five-figure exit is not a market. It is a pricing booth. Depth, spread, and fee-adjusted slippage are the only honest indicators of whether this product can ever offer real liquidity.
Third: check whether Moonshot AI's legal team speaks. Companies hate unauthorized paper. A public disavowal turns "in accordance with applicable rules" into the only rule that matters.
For the record, the all-in break-even sits near $119 per unit before carry. If the dedicated market opens around $115 and bleeds, the correct position is no position. There is no shame in missing a narrative. The cost of being early to a synthetic with a 7.5% sales tax is not paid in courage. It is paid in dollars.
Code is law, but gas fees are the reality. Here, there is not even code — just a ledger controlled by one party. The actual innovation is not pre-IPO access; Forge Global built that decades ago. The innovation is convincing 58 million crypto users that a mirror is a window. A mirror is not a window. And the fee schedule — 5% in, 20% of the upside, 2.5% to leave — tells you exactly who is standing on the other side, looking through.
