As US stocks close, Nvidia still trades on-chain: liquidity has shrunk by 33%, so why are seemingly ubiquitous arbitrage opportunities unprofitable?

Author: Cooper Duschang (Coin Metrics)

Compiled by: TechFlow

TechFlow Summary: As the wave of real-world asset (RWA) tokenization accelerates, tokenized US stocks represented by companies like NVIDIA and Tesla are becoming the core entry point for global non-US investors to gain US equity exposure, with their total market capitalization rapidly surpassing the $3 billion mark. However, unlike traditional US stock exchanges with strictly limited trading hours, the on-chain market operates 24/7, giving rise to unique after-hours and overnight pricing dynamics.

Coin Metrics reveals through deep on-chain and high-frequency order book data that tokenized stocks often experience a sharp 19% to 33% drop in order book depth during traditional market closures. This liquidity vacuum not only amplifies daytime premiums but also leads to significant spreads for the same underlying asset across different exchanges and issuers. This research report comprehensively unpacks the lead-lag effects, execution slippage, and arbitrage frictions between tokenized US stocks and their underlying assets, fully illustrating the foundational evolution of the on-chain capital markets.

Between October 27 and November 2, spot market trading volume for tokenized US stocks across major exchanges surpassed $2 billion. With the sustained expansion of demand for on-chain US equity exposure, custodians continuously increased their holdings of underlying stocks and issued more tokenized shares, pushing the total market capitalization of tokenized US stocks above $3 billion.

Fig.: Total market cap of tokenized US stocks and growth trends across major issuers

Although the total market capitalization of tokenized stocks continues to climb steadily to approximately $3 billion, spot trading volume remains relatively limited compared to their perpetual contract derivatives.

In the past week alone, perpetual contracts on US equity indices and single stocks on the Hyperliquid platform generated approximately $12.3 billion in trading volume, equivalent to six times the $2 billion spot market volume for tokenized US stocks.

Fig.: Comparison of spot trading volume for tokenized stocks and on-chain US equity perpetual contracts

In this latest network state analysis, we will systematically unpack the premium/discount patterns of tokenized stocks relative to their underlying real stocks, on-chain pricing behavior during US stock market after-hours and pre-market sessions, and the fragmented liquidity status of these assets across major decentralized and centralized platforms.

Pricing Mapping Between Tokenized Stocks and Underlying Physical Stocks

Tokenized stocks are typically issued for non-US investors, providing them with on-chain price exposure outside regulated markets, including during US stock market closures.

These tokens are usually issued by institutional entities holding the underlying real stocks, with assets segregated and custodied through compliant custodians. Operating in a 24/7 trading market, tokenized stocks allow assets to maintain continuous pricing and liquidity even when traditional financial markets are closed.

Fig.: Comparison of continuous pricing trends for tokenized NVIDIA stock and the traditional US stock market

Currently, only specific institutions and market makers that have passed strict whitelisting approvals hold the direct right to mint new tokens or redeem underlying stocks from the issuer. This mechanism poses structural redemption risks and potential liquidity discount hazards for ordinary holders without whitelisted permissions.

From October 27 to November 2, NVDAx issued by xStocks traded at a premium for 58.6% of regular US stock market opening hours, with an average premium of 3.1 basis points (bps).

Meanwhile, NVDAon issued by Ondo Stocks traded at a premium for as much as 88.9% of regular trading hours, with an average premium reaching 14.1 basis points. Following the broad market sell-off last Friday, quotes for tokenized NVIDIA stock overall fell below Friday's closing price, reflecting preemptive risk-averse selling by offshore capital.

Fig.: Premium rate distribution of NVDAx and NVDAon relative to the underlying benchmark stock

Market segmentation between exchanges leads to price deviations for the same tokenized stock across different platforms, regardless of whether it is during regular trading hours or market closures. Taking the NVDAx market on the Kraken exchange as an example, its average spread relative to the underlying NVDA during regular opening hours was -0.9 bps, showing a slight discount.

Lead-Lag Transmission Effects During Market Closures

Using a quantitative lead-lag analysis model, we compared the price volatility trajectory of NVDAx tokens during US stock market closures with the actual opening price of the underlying real NVDA stock on the following trading day. Covering 210 complete trading days from January 2026 to date, we calculated across three different time windows:

After-hours session (ET 16:00 to 20:00): The evening volatility of tokens immediately after US market close shows extremely weak correlation with the next morning's actual opening price, with a correlation coefficient of only 0.24. This means that in the initial post-close period, on-chain token volatility provides almost no effective warning signal for the next day's market direction.

Overnight deep session (ET 20:00 to 04:00 next day): The correlation coefficient between overnight token trends and the next day's opening price rises to 0.30. Although the signal remains weak, it is noticeably stronger than the after-hours session, and signal strength gradually accumulates and amplifies as the US pre-market trading session approaches.

Pre-market session (ET 04:00 to 09:29): Upon entering the pre-market window, NVDAx quotes begin to rapidly and steadily converge toward the expected opening price of the underlying NVDA, with the R-squared value from regression analysis surging significantly to 0.42.

Complete closure cycle (ET 16:00 to 09:30 next day): From a global perspective spanning the entire closure period, on-chain tokens almost perfectly tracked the actual movement of the benchmark stock, yielding an R-squared value as high as 0.98. As anticipated by market arbitrage mechanisms, at the moment the bell rings for the open, NVDAx has fully converged and eliminated any spread with the underlying asset.

The price of tokenized stocks becomes highly anchored to real stocks at the US market open, but this convergence effect is primarily achieved through intense arbitrage during the pre-market session. Observing token volatility during the first few hours after close alone is insufficient to determine the next day's trend.

Fig.: Regression analysis of correlation between tokenized stock volatility during various closure periods and US stock opening prices

Liquidity Vacuum and Depth Gaps During Market Closures

Tokenized stocks are highly sensitive to the price impact of fund inflows and outflows, especially during traditional stock market closures. After traditional markets close, liquidity within the 1% midpoint depth range for NVDAx markets on Bybit and MEXC drops by an average of 19%.

On the Hyperliquid platform, limit orders within the 1% midpoint depth range evaporate by approximately 33% during closure periods. During this vacuum period of sharply shrinking liquidity, any moderately sized buy or sell order can severely impact the price, thereby exacerbating price distortions for the same underlying asset across different trading venues until US stocks reopen.

Fig.: Comparison of 1% limit order depth shrinkage across major platforms during traditional open and closure periods

When Do Tokenized Stocks Concentrate Their Trading Activity?

The core value of tokenized stocks lies in providing non-US investors with an around-the-clock channel to invest in US equity assets. By mapping the intraday volume distribution across all hours, it is clear whether investors remain active even during thin-volume closure periods.

Fig.: Average trading volume distribution of tokenized NVIDIA stock over a 24-hour period

Data shows that the vast majority of NVDA xStocks trading activity remains concentrated around the two key time windows of US market open and close. On weekdays between ET 09:00 and 10:00, NVDAx averages an hourly trading volume of $225,000.

While there is some trading volume during pre-market and after-hours, primarily driven by bets on opening prices and digestion of after-hours earnings or major news, this is insufficient to reflect the true geographic distribution of traders. Offshore exchange platforms supporting these markets indicate that international users are indeed buying and selling tokenized assets, but their trading rhythms remain heavily tethered to the conventional US market clock.

Cross-Platform Execution Costs and Slippage Dynamics

Pricing discrepancies for the same tokenized stock across different platforms mean that the true cost for investors to gain US equity exposure depends entirely on where the order is executed.

Fig.: Comparison of execution slippage across major trading platforms for different order sizes

For small orders, multiple exchanges can provide reasonably ideal execution depth. Taking $1,000 orders between October 27 and November 2 as an example, Bybit and Hyperliquid served as the optimal execution platforms with the lowest slippage during 58% and 23% of time windows, respectively.

However, when individual trade sizes scale up to the $100,000 level, market liquidity quickly concentrates at the top, with Bybit or Hyperliquid acting as the optimal execution venues for up to 98.2% of the time.

Fig.: Trend of average execution slippage for tokenized NVIDIA issued by different providers

Comparing token variants issued by different providers reveals that execution costs depend not only on the exchange but are also closely tied to the issuing brand. From October 27 to November 2, NVDAb issued by Backed on the Binance platform demonstrated the lowest slippage levels among all NVIDIA token variants in 96% of observed samples when measuring against $10,000-sized orders.

Can Arbitrage Spaces Cover Friction Costs?

Although price differences exist among token variants from different providers across various exchanges, theoretically birthing arbitrage opportunities, actual trading must strictly account for liquidity depth and slippage costs.

Looking back at historical data from June 2026 to November 2, before deducting fees, the average bid-ask spread and slippage friction alone completely offset the cross-market spreads observable even in $1,000 small trades.

Fig.: Distribution comparison between cross-market spreads and comprehensive trading slippage friction

Conclusion and Outlook

Tokenized stocks have opened the door for global non-US investors to access US core equity exposure around the clock. However, current tokenized assets still heavily rely on traditional trading hours and face significant liquidity gap challenges during US market closures.

The systematic collection and monitoring of order book depth, execution slippage, and fair reference benchmark rates for tokenized assets provide irreplaceable quantitative reference value for investors to horizontally evaluate execution costs and avoid potential pricing traps in fragmented on-chain and offshore markets.