Notice: Informational analysis, not financial advice. Market data is frozen as of August 29, 2026, at 21:52 UTC (Hyperliquid public API, Aster API, Lighter API, xStocks catalog); quotes for 688836 and Meituan are closing prices from August 28, 2026, and the issuer count for Hong Kong lines within the xStocks catalog is from August 23. Funding rates change hourly and the figures in this article do not self-update. CleanSky does not receive commissions or referral payments from any of the cited platforms.
Holding a $1,000 long position in Unitree perpetuals during the three debut sessions, from August 19 to 21, 2026, paid you $28.54; the person on the other side was charged that amount. The commission for that same trade, entry and exit, was $0.18 with the operator's reduced fee, or $1.80 at the standard rate. The perpetual (a contract that tracks the price of an asset without expiration and without ownership of the underlying) is the only direct way a retail investor outside mainland China can gain exposure to Unitree Robotics, the robot manufacturer that debuted on Shanghai's STAR market that day under the ticker 688836. The three layers in which a traditional asset can exist on-chain distribute costs very unevenly, and CleanSky's own calculation based on the August 29 snapshot quantifies it: the term highlighted by advertising—the commission—is the smallest and the only one the user controls. The large one lies in the funding, and its sign is decided by the calendar.
Where can a retail investor get exposure to Unitree: perpetual, token, or stock?
A traditional asset can appear on-chain in three ways, and what remains in hand is a different object in each. Layer 1 is a cash-settled contract with no underlying shares: perpetuals on Hyperliquid via HIP-3, the standard that allows a third party to deploy their own market on top of the chain's book. Layer 2 is a transferable receipt that an issuer puts into circulation against a held security, without voting rights. Layer 3 is the stock in the investor's name, the only one with corporate rights. The distribution of rights among the three Layer 2 windows is broken down in the xStocks vs Ondo vs Robinhood comparison, and the chart of which asset exists in which layer is in the traditional assets on-chain table. For Unitree, the distribution is asymmetric to the extreme.
As of August 29, 2026, for Unitree Robotics (688836), only Layer 1 exists: eight markets spread across seven DEXs, with $25.15 million in aggregate open interest—99.5% concentrated in the four that this article measures hour by hour—plus Bybit and Binance (centralized), and Vantage, which offers it as a contract for difference; and zero issuers of 1:1 backed tokens among more than 3,100 products (714 from xStocks, over 440 from Ondo, over 2,000 from Robinhood).
That snapshot expires every hour. To see it by the minute—here or for any other traditional asset—there is the traditional asset DEX finder, with over twenty platforms read from their public APIs, and the specific profile for where to buy Unitree Robotics on-chain, featuring the eight markets and the country exclusions for each; its snapshot is from August 23 and refreshes on Mondays.
| Layer | What you hold | For Unitree (688836) | Open Interest |
|---|---|---|---|
| 1 · Perpetual without ownership | Cash-settled contract | Eight markets across seven DEXs: trade[XYZ] and Paragon (two HIP-3 operators on Hyperliquid), Aster, Lighter, edgeX, ApeX Omni, Extended, and Carbon; plus Bybit, Binance, and Vantage, the latter as a CFD (contract for difference) | $25.15 million (eight DEX markets) |
| 2 · 1:1 Backed Token | Receipt of a share in custody | None. Zero products among the 714 from xStocks, 440+ from Ondo, and 2,000+ from Robinhood | $0 |
| 3 · The Stock | Security with corporate rights | Stock Connect to the STAR market is reserved for professional institutional investors; the QFII (Qualified Foreign Institutional Investor) quota is also institutional | — |
The open interest for Layer 1—the live notional at each venue, i.e., each platform where the order is matched—is distributed as follows on August 29 at 21:52 UTC: trade[XYZ] concentrates about $22.3 million (254,167 contracts at a mark price of $87.71), Paragon $2.23 million, Aster $289,000, and Lighter $216,000. On trade[XYZ], there are more live contracts today than six days ago—254,167 versus 240,815—but less notional, because the mark price fell 8.4% in that span. On August 15, before the debut, CoinDesk counted $9.1 million across two markets: the jump comes with the over one million contracts that changed hands on August 19, the day of the STAR debut. The remaining four markets total $120,000, and one of them disrupts the reading: edgeX had $113,467 in open interest with $1.53 million in 24-hour volume, 13× more turnover than live position, while trade[XYZ] had 17× more position than turnover ($22.29 million against $1.29). Two markets for the same contract, on the same day, with opposite profiles.
The perpetual for the STAR 50 ETF, xyz:KSTR, appears on trade[XYZ] as delisted with zero open interest; on Aster, KSTRUSDT remains active with 297 contracts, about $7,200. Not even the indexed proxy has a live market, and the STAR 50 index does not yet include Unitree.
What does it really cost to hold $1,000 of Unitree in Layer 1?
The calculation is proprietary and its assumptions are published so anyone can replicate it: 6.74 yuan per dollar, the implicit rate of the 150.80 yuan = $22.37 pair from the opening price; base-level commission in trade[XYZ] growth mode according to its documentation, with growthMode: enabled confirmed by the API; hourly rates as applied by Hyperliquid; and the book spread in the August 29 snapshot. No leverage, $1,000 nominal long position.
The published fee for trade[XYZ] is 0.090% taker and 0.030% maker at the standard rate, and 0.009% / 0.003% in growth mode, the reduced commission mode the operator activates in most of its stock markets—including NVDA and AAPL. Half of that commission goes to Hyperliquid and the other half to the market operator.
| Term | Calculation | Result per $1,000 |
|---|---|---|
| Round-trip commission (taker, growth mode) | 2 × 0.009% × $1,000 | $0.18 paid |
| Round-trip commission (taker, standard rate) | 2 × 0.090% × $1,000 | $1.80 paid |
| Crossed spread (Aug 29 book) | 0.026% × $1,000 | $0.26 paid |
| Funding, pre-IPO phase (Aug 5 to 18) | sum of hourly rates, 14 days (+0.0092%) | $0.09 paid |
| Funding, three days post-conversion (Aug 19 to 21) | −1.462% −0.821% −0.571% | $28.54 collected |
| Funding, full window (24.3 days) | −4.127% × $1,000 | $41.27 collected |
| Worst observed hour for the short (long collects it) | 0.3719% × $1,000 | $3.72 in one hour |
The two terms the user sees before opening the position—commission and spread—total $0.44 in growth mode. The term they don't see moved $41.27 in the same position: 94 times that sum of commission plus spread, and 23 times the round-trip commission at the standard rate, which is the single denominator used in the rest of the article ($1.80 per $1,000, or 0.18% of the nominal). The window is less than 30 days: the market opened on August 5 at 15:00 UTC and the last rate is from 21:00 on the 29th, totaling 583 hourly rates, or 24.3 days.
Funding is a payment between the two sides of the contract: if the perpetual trades above the underlying, longs pay; if it trades below, it's the reverse. Hyperliquid settles it every hour and retains none of that flow, which goes entirely to the user on the opposite side. The mechanism is explained in how funding rates work in a perpetual DEX.
Who collected the Unitree perpetual funding after the August 19 debut?
The long side collected it and the short side paid it. The reason has nothing to do with the quality of either side's analysis: it was decided by the debut calendar.
In the preceding fourteen days, with the market in pre-IPO mode—5x leverage and a funding multiplier of 0.005, a hundred times smaller than the 0.5 for SpaceX or Cerebras—no session left the range between −0.004% and +0.007% daily: $0.09 accumulated per $1,000, half the growth mode commission.
On August 19, the stock debuted in Shanghai and closed at 845 yuan. The perpetual, which until then had traded without an external reference, remained significantly below the price converted to dollars. And then the scale of the funding sign shifted.
| Date (UTC) | Milestone | Daily Funding | Effect on $1,000 Long |
|---|---|---|---|
| Aug 5 to 18 | Pre-IPO phase, no external oracle | between −0.004% and +0.007% daily | $0.09 paid in 14 days |
| Aug 19 | STAR debut; close 845 yuan | −1.462% | $14.62 collected |
| Aug 20 | Close 687 yuan (−18.70%) | −0.821% | $8.21 collected |
| Aug 21 | Close 672.41 yuan (−2.12%) | −0.571% | $5.71 collected |
| Aug 22 and 23 | Weekend, internal price setting | −0.032% and −0.039% | $0.71 collected |
| Aug 24 to 29 | No milestones; stock falls to 585.00 yuan on the 28th | −1.212% in six days | $12.12 collected |
The worst individual hour of the series was −0.3719%. Sustained for a full day, it would have been 8.9% of the nominal: $89 per $1,000 in 24 hours, collected by the long and paid by the short. It wasn't sustained, but the magnitude shows what can happen in one hour in a market with $5,027 of depth on the buy side and $175,396 on the sell side, which is what the xyz:UNITREE book had in the Saturday night snapshot. For comparison, the SpaceX perpetual on the same platform had $555,031 and $535,917, distributed almost equally between both sides. The tightness has an hourly explanation: STAR markets on trade[XYZ] only take an external oracle on business days from 9:30 to 11:30 and 13:00 to 14:57 Beijing time, converting offshore yuan to dollars in real-time; outside that window—most of the time, and the entire weekend—the price is set by the book itself, and this snapshot is from a Saturday night.
The pattern did not end with the debut, but it did fade. On August 24 and 25, funding still totaled −0.483% and −0.663%; the following four days were a trickle. August 28 was the first day since conversion that the sign flipped and longs paid: +0.093%. The reason is arithmetic. The stock closed that day at 585.00 yuan—$86.79 at the implicit rate of 6.74—and the perpetual closed at $88.03, 1.4% higher. As soon as the contract stops trading below the stock, the funding flips sides.
That the long collected is an effect of the debut calendar; what persists is the gap between what the commission costs and what the funding moves. Whether the perpetual correctly predicted the stock price is another question, which we measure separately in the pre-IPO perpetual accuracy scoreboard for Unitree.
The same pattern appears in the other two venues with public history—Paragon and Aster; Lighter does not publish it—with the caveat that units are not interchangeable: Hyperliquid settles every hour and Aster every eight, except for UNITREEUSDT, which switched to four on August 19, so its 104-period series mixes both intervals. Aster also applies a cap of ±2% per period, and a single period in the entire series hit that floor: August 19 at 08:00 UTC. The totals for windows of similar length do compare: −4.127% on trade[XYZ] over 24.3 days, −5.413% on Paragon over 25.6, and −5.857% on Aster over 23.8.
Does funding also outweigh commission in NVDA and Apple perpetuals?
Funding also outweighs commission in NVDA (3 times) and in AAPL (2 times), but nowhere near the 23 times seen in Unitree, and saying so matters because that is where most people trade. In stocks with years of history, arbitrage against the spot is cheap and funding stays close to zero: commission does compete. These four rows are the anchors of the range; the full ten markets, with their absolute sum and worst hour, are in the comparative table by layer.
| Market on trade[XYZ] | Net 30d Funding as of Aug 29, 2026 | Effect per $1,000 Long | Times the round-trip commission (standard fee 0.18%; in growth mode ×10) |
|---|---|---|---|
| CXMT (Chinese memory, pre-IPO) | −13.821% | $138.21 collected | 77× |
| UNITREE (recently debuted, 24.3 d) | −4.127% | $41.27 collected | 23× |
| NVDA | +0.498% | $4.98 paid | 3× |
| AAPL | +0.366% | $3.66 paid | 2× |
In Apple, net 30-day funding was $3.66 per $1,000 compared to $1.80 in standard round-trip commission: double, not two orders of magnitude. In SpaceX, with −0.393% net, $153.3 million in open interest on trade[XYZ], and $5.0 million in daily volume, the balance was $3.93 collected, 2.2 times the commission. And SpaceX has been trading on Nasdaq since June 12, which defines the thesis better than any label: what matters is the distance to the debut. In the two markets that do not yet have a stock behind them or have just had one—CXMT and UNITREE—funding ranges from 23 to 77 times the commission. Cerebras, converted on May 14, has dropped to 6. SpaceX, converted on June 12, to 2.2, already in the territory of liquid megacaps where the rule ceases to apply.
The table is calculated based on the standard fee. In growth mode, where the round-trip costs 0.018%, each multiple jumps tenfold, and even NVDA reaches 28 times. The reduced commission does not shrink the problem: it enlarges it, because it shrinks the known term without touching the unknown one.
There is also an asymmetry that the published fee leaves out: the net hides the volatility along the way. In UNITREE, the sum of hourly absolute values is 5.790% against a net of −4.127%; in ZHIPU (the Chinese LLM developer, also with a pre-IPO market), 7.690% against +2.150%. Anyone entering and exiting mid-window takes whatever they were assigned in their span, which in ZHIPU can be three and a half times the final balance of the entire window.
Why isn't Unitree (688836) on xStocks, Ondo, or Robinhood?
The figure is the result of extracting the full catalogs. xStocks published 714 products on August 29—715 six days earlier; Ondo Stocks declares "over 440"; Robinhood, "over 2,000." Neither 688836 nor any other STAR stock nor any Shanghai or Shenzhen A-share appears in any of the three. Neither does the STAR 50 ETF nor any mainland China ETF.
The reason lies in the product terms themselves. Ondo describes its catalog as stocks and ETFs "all listed in the United States." Robinhood issues derivatives under MiFID II, not ownership, and its only off-exchange lines are OpenAI and, since before its IPO, SpaceX, both via special purpose vehicles. Backed, the issuer of xStocks, is a Jersey entity distributing through Payward—Kraken's parent company—from Bermuda and Cyprus, and its product requires buying and custodying the security: for an A-share, that means a QFII quota or a Stock Connect channel that no token issuer has open to retail.
Where there is a Chinese catalog is in Hong Kong. Among the 714 xStocks products, at least fifty lines of Hong Kong-listed securities appear—48 identified by counting issuer by issuer on August 23, as the catalog does not mark which exchange each symbol trades on: Tencent (TCENTx), Meituan (MEITx), Xiaomi (XIAOx), BYD (BYDCOx), ICBC (ICBCx), AIA (AIAGRx), HKEX (HKEXCx), Kuaishou (KUAIx), and Horizon Robotics (HRZRBx), which manufactures assisted driving chips. The limit is set by the listing type: Hong Kong lines enter the catalog, while A-shares are left out of all three.
There is one last place to look outside Layer 1, and it's not there either: Jarsy, a platform for access to private markets, maintains a Unitree page—consulted on August 29, dated the 27th—with no price, no ticker, and no buy button, with a warning that creating an account "does not guarantee access to any company." The layer-by-layer comparison of all these assets, including issuer, commission, funding, and jurisdictions, is in the traditional assets on-chain table.
How much Unitree is inside a Meituan share?
There is an indirect Layer 2 route, and it can be quantified. Meituan held 9.6488% of Unitree before the placement, according to the ACN Newswire notice on August 10. The placement was for 10% of the expanded capital, so the diluted stake stands at 8.68%. And Meituan is tokenized: MEITx, listed on ten chains in the catalog, with contract addresses published for nine (Tron's is missing).
The calculation, with its assumptions visible:
| Step | Data | Result |
|---|---|---|
| Meituan's diluted stake in Unitree | 9.6488% × 0.9 | 8.68% |
| Unitree Market Cap (Aug 28, close 585.00 yuan) | Sina Finance; 404.46 million shares | 236.61 billion yuan |
| Value of the stake | 8.68% × 236.6 billion yuan | 20.54 billion yuan ≈ $3.05 billion |
| Meituan Market Cap (Aug 28, 77.50 HKD) | 478.32 billion Hong Kong dollars (StockAnalysis), which at 77.50 HKD implies 6,172 million shares | ≈ $61.32 billion |
| Weight of Unitree within Meituan | 3.05 / 61.32 | 4.97% |
| Effective exposure per $1,000 of MEITx | $1,000 × 4.97% | $49.70 |
The starting market cap matches the share capital: 236.61 billion yuan at 585.00 per share implies the same 404.46 million securities, consistent with the 40.4464 million placed as 10% of the expanded capital. In the week of August 21 to 28, both legs fell—Unitree by 13.0% and Meituan by 8.8%—and the ratio barely moved: from 5.21% to 4.97%. The 6,172 million Meituan shares come from the market cap published by StockAnalysis against the August 21 close; rates are 7.80 HKD/$—within the 7.75 to 7.85 band—and 6.74 CNY/$. The figure values the stake at market price, does not discount latent capital gains tax or Meituan's net debt, and the market does not have to recognize 100% of an unconsolidated stake. It is a ceiling, not an estimate.
Even accepting the ceiling, the order of magnitude rules: it takes about $20,100 of MEITx to obtain $1,000 of economic exposure to Unitree, and with it comes $19,100 of exposure to Meituan's food delivery business. The other listed shareholder, Shoucheng Holdings (0697.HK) with 3.8262% before the placement, is not tokenized in any of the three catalogs.
Does MiCA regulate tokenized stocks or does MiFID II?
MiFID II regulates them. Article 2, paragraph 4, point (a) of Regulation (EU) 2023/1114—MiCA—excludes from its scope crypto-assets that qualify as financial instruments, and a tokenized stock falls squarely into that exclusion. The four windows reviewed on August 29, 2026, confirm this one by one: Robinhood issues its stock tokens as derivatives under MiFID II with a Lithuanian license, only for the EU and EEA, with zero commission and a 0.1% currency exchange fee per trade—0.20% round-trip; xStocks is distributed from Cyprus by a Payward entity, Kraken's parent, and is not available in the United States, Canada, United Kingdom, or Australia; Ondo restricts it to non-US persons; and in Layer 1, trade[XYZ] operates as XYZ Ltd under British Virgin Islands law and excludes the United States, Canada, and the United Kingdom, with an express VPN ban.
The pattern of all four is not one of crypto regulation: it is the securities law of each jurisdiction, which requires a prospectus, an authorized intermediary, and recognized custody to put a stock in the hands of a retail investor. MiCA regulates stablecoins and crypto-asset service providers—we cover this in the electronic money token register; for tokenized stocks, the framework is different. The path that would point to a European product with a prospectus is that of exchanges entering the business, as in the ICE and OKX joint venture.
What would have to happen for commission to outweigh funding again in Unitree?
Two things, both verifiable against a public source in an afternoon.
- That some issuer of 1:1 backed tokens—Backed/xStocks, Ondo, Robinhood, Dinari, or Binance's bStocks—lists 688836 or any STAR A-share. That would open Layer 2 and turn the asymmetry of the first table into something transitory. Of the five, three have verified catalogs—xStocks on August 29 based on the full list published on its products page, Ondo and Robinhood on the 23rd—and none list it; Dinari and bStocks are pending direct verification.
- That in the three trade[XYZ] markets with the most extreme funding—UNITREE, CBRS, and CXMT, of which only the first two already have a listed stock behind them—the net accumulated 30-day funding drops, in absolute value, below double the round-trip commission at the standard rate, i.e., 0.36% of the nominal. Then the commission would be the dominant term and the second half of the argument falls. On August 29, the lowest of the three, CBRS at 1.019%, is 2.8 times above that threshold; UNITREE, 11.5 times; CXMT, 38.4 times. Two of the three have approached the threshold in six days—CBRS came from 4.1 times and CXMT from 53.7—which is the direction in which this second condition would be met. UNITREE has moved away, from 8.2 to 11.5, because its window is still growing toward 30 days and is not yet comparable to itself. The first full 30-day window for UNITREE closes on September 4, 2026: until that date, its figure covers 24.3 days out of 30.
In the meantime, what remains is an unequal distribution of information. The commission appears in the interface before signing and is real: $0.18 per $1,000 is cheap compared to the $2.00 for a taker round-trip on Kraken Pro, the platform that distributes xStocks. But for Unitree, that comparison does not exist because Layer 2 does not have the asset. What the user buys when opening the position is a flow of hourly payments two orders of magnitude above the commission, with a sign set by the gap between the perpetual and a stock that trades four hours a day in another time zone. Whoever was long in Unitree from August 19 to 21 won that lottery. Whoever was short paid it without having been wrong about the direction.
Related articles: Traditional assets on-chain: perpetual, 1:1 token, or stock, the live table with issuer, commission, and funding per asset. Unitree and the three signals to read a pre-IPO perpetual, which measures price error instead of cost. How trade[XYZ] pre-IPO perpetuals work and what happened to SpaceX after its conversion. Monitor your positions on CleanSky — wallet tracking, lending, and portfolio; derivatives are outside the scope of the tool.