Editor's Note: In October 2025, Hyperliquid launched HIP-3, allowing developers to deploy perpetual contract markets on its core trading infrastructure, HyperCore. Subsequently, TradeXYZ introduced XYZ100 and gradually brought traditional assets such as NVIDIA, Micron, gold, and oil onto the chain. While traditional markets have fixed trading hours, on-chain perpetual contracts can operate 24/7; the differences between the two sets of markets in price updates, liquidity, and funding costs have given rise to a new wave of arbitrage opportunities.
The real question worth discussing is whether this profit comes from trading skill or from an immature market structure. Equity perpetual contracts have seen rapid growth, but on-chain pricing still relies on traditional markets, and funds cannot be transferred instantaneously between different venues as with crypto assets. Price spreads, funding rates, and speed advantages can generate high returns, while data interruptions, hedge delays, and bank transfer limitations can quickly create significant one-way exposure.
In this article, author CBB reflects on his and his brother's experience of building a stock perpetual contract arbitrage bot from scratch. According to his account, within 10 months, the duo executed approximately $32 billion in volume through Hyperliquid HIP-3 and Interactive Brokers (IBKR), making over $10 million in profit. However, during this period, a market data failure caused the bot to unexpectedly accumulate a $120 million short position in gold, resulting in a $1.1 million loss. The related profit, volume, and incident data are all disclosed by the author and have not yet been independently audited.
Nowadays, professional market makers are entering the space, and Ethena also plans to expand its arbitrage trading to stock perpetual contracts. While on-chain stock perpetual contracts may not have become a mature market yet, the phase where early participants rely on speed and market frictions to gain excess returns is narrowing. This article documents both a set of arbitrage strategies and a competition regarding execution speed, infrastructure, and risk control.
Key Points from the Original Article:
In October 2025, CBB and his brother began searching for new trading opportunities.
For the previous eight months, they had been running a leading-scale arbitrage bot on HyperEVM. However, as professional firms like Wintermute joined the competition, the returns from their original strategy significantly decreased. For a small team of only two people, continuing to compete head-on with institutions possessing more capital, engineers, and risk management resources was not realistic.
Their past advantage came from speed: deploying strategies quickly after identifying opportunities, racing to capture early gains before large institutions completed compliance, approval, and system integration.
This time, they set their sights on the newly launched HIP-3.
HIP-3 Launches, Revealing a Pricing Discrepancy between On-chain and Traditional Markets
HIP-3 launched on October 13, 2025, on the Hyperliquid mainnet. It allows developers meeting staking requirements to deploy perpetual contract markets on HyperCore, where they can set parameters such as assets, oracles, leverage, and fees. Three days later, Unit/TradeXYZ introduced the first stock-like market, XYZ100.
CBB initially did not have a complex trading thesis. They believed that there were still plenty of HYPE tokens awaiting distribution in Hyperliquid, and accumulating trading volume early on HIP-3 might bring potential incentives. Eight months ago, they had used a similar approach when entering the HyperEVM arbitrage market.
The two decided to build a set of arbitrage bots connecting HIP-3 to the traditional financial markets: using IBKR's stock and futures prices as reference, establishing positions on HIP-3 when there was a discount or premium, and then hedging the positions through IBKR.
If the price of the NVIDIA perpetual contract on HIP-3 was lower than IBKR's, the bot would long on HIP-3; once the on-chain order was filled, it would short the equivalent amount of NVIDIA stock through IBKR. If HIP-3 experienced a premium, the reverse operation would take place: shorting on-chain and longing on IBKR.
Structurally, this was a cross-market basis trading: traders held positions in opposite directions, primarily profiting from the price convergence spread and funding rate benefits. It is commonly referred to as a "market-neutral" strategy, but market neutrality only describes the directional exposure in an ideal state and does not imply that the trading is risk-free.
How Does the Arbitrage Strategy Operate with IBKR Hedging On-chain Positions?
Prior to launching the project, neither of them had traded stocks before, and they had almost no experience in the futures market.
In the first few days, CBB studied the IBKR platform while sending various interface screenshots to Claude, inquiring about contract meanings, trading setups, and how to hedge XYZ100. Meanwhile, his brother started researching the IBKR API.
They quickly discovered that the traditional financial trading system was significantly different from a cryptocurrency exchange. Cryptocurrency exchanges usually offer quick access through a unified API, while IBKR involves multiple systems such as market data subscriptions, contract specifications, order permissions, TWS, IB Gateway, and interface restrictions.
A week later, the arbitrage bot started taking shape.
To control trading costs and execution risk, they set parameters separately for each market. For example, with NVIDIA, the bot would not hedge immediately on every minor spread opportunity but would wait until the net position accumulated to 55 shares before executing, to mitigate the impact of IBKR's minimum $1 per trade fee; the maximum single hedge limit was set to 400 shares to avoid excessive slippage from large orders.
One of the most crucial parameters was the maximum position delta. If the position delta between Hyperliquid and IBKR reached 800 shares, the bot would cease trading that market to prevent further exposure escalation after a hedge failure.
The settings on the HIP-3 side were more detailed, including order size, bid-ask spread relative to the fair price, cancellation threshold, minimum price difference required for taking liquidity, maximum position limits, and additional risk premium during pre-market hours. Due to weaker pre-market liquidity in traditional markets, the bot would widen the quote spread to cover higher hedging costs.
By the end of October 2025, the system was officially launched. In the initial days, the IBKR API frequently disconnected, but the two soon confirmed that there was indeed a significant price difference between HIP-3 and traditional markets.
According to CBB reports, the bot completed around $850 million HIP-3 trading volume in November, generating a profit of over $500,000; in December, the trading volume was around $550 million, with still substantial profits.
However, at that time, they did not consider it a "gold mine" worthy of long-term investment. The real game-changer was the subsequent surge in precious metals prices.
$120 million Short Position Error: Why Did Market Neutrality Suddenly Fail?
In January 2026, there was a surge in gold and silver prices, with a notable increase in commodity long positions on Hyperliquid.
As most on-chain traders favored long positions, the arbitrage bots needed to continuously establish short hedges on IBKR. As the trading volume grew, the two had to constantly replenish funds into the IBKR account. Traditional financial accounts could only receive deposits through the banking system, and funds could not transfer as swiftly between trading platforms as with crypto assets, making liquidity management a key constraint on the strategy.
According to the author, they conducted a $1.7 billion transaction volume on Hyperliquid in January, with only the funding rate revenue exceeding $600,000.
The rapid growth also caused the original risk control design of the system to start failing.
On January 27, CBB received a margin call alert from IBKR shortly after arriving in Dubai. Upon logging into the account, he discovered that the bot had accumulated a $120 million gold futures short position on IBKR, while the price of gold was still rising rapidly.
The two immediately closed the bot and manually closed out all short positions in the next 15 to 30 minutes. The author calculated that this incident ultimately resulted in approximately $1.1 million in losses.
A post-incident investigation revealed that the position data from the IBKR API did not refresh correctly. As a result, the bot mistakenly believed there was a perpetual unhedged exposure between Hyperliquid and IBKR, leading it to repeatedly short sell gold in an attempt to correct for a non-existent delta.
This incident exposed one of the most easily overlooked risks in cross-market arbitrage: when two opposing positions fail to sync in execution, the original market-neutral strategy transforms into a directional bet. Market data distortions, API disconnections, order rejections, or insufficient funds can all lead to similar outcomes.
Following this, the two implemented data latency checks, position growth limits, and more pre-trade validations. However, upon the system's redeployment, they briefly lost confidence and began reevaluating the strategy's risk-reward ratio.
The next day, silver experienced a significant pullback from its all-time high, and there was a temporary 3% price divergence between Hyperliquid and IBKR. According to the author, the bot seized this opportunity and made a profit of approximately $600,000, recouping over half of the previous losses.
From Price Discovery to Speed Race, Arbitrage Competition Intensifies
With the sustained increase in profits, the two individuals judged that more institutional players would soon enter this trading space. To maintain their edge, they began focusing on two key challenges: capital and market data speed.
Asset re-balancing within the crypto market typically takes only a few minutes, but fund transfers to and from IBKR accounts rely on bank transactions. If available funds on one side of the traditional market dwindle, the bot might be unable to further expand hedging positions.
As a solution, they established a dynamic fund management system.
When IBKR's available liquidity is low, the bot increases the spread required for opening positions, while accepting some losses to close existing trades, thus releasing margin; when account liquidity is sufficient, it lowers the spread requirement and deploys funds more aggressively.
The second issue is market data speed.
Initially, the robot relied entirely on IBKR's data to determine the fair price. However, as competitors increased, the update speed of IBKR's market data became a limitation. The two then connected to Databento and applied for Nasdaq market data authorization to access faster direct market data.
This change indicates that the competition of arbitrage strategies will eventually shift from "price discovery" to "who sees the price difference first, who trades first, and completes the hedge first." When participants use the same pricing model, latency, fees, capital costs, and order execution quality will determine the final returns.
Gold, Crude Oil, and Semiconductor Relay, Amplifying Profit Space through Volatility
Entering February, the precious metals market remained active. According to the author, the robot completed approximately $1.5 billion in trading volume that month.
Subsequently, geopolitical conflicts escalated market volatility, with oil prices surpassing $100 per barrel. The author stated that during the most active market periods, arbitrage spreads and funding rates could bring in daily income of about $60,000 to $120,000. While traditional financial markets are closed on weekends, on-chain perpetual contracts continue to trade, further widening the pricing difficulty due to the disparate trading hours of the two markets.
At this point, Claude was also included in the trading analysis process. The two individuals provided the model with Hyperliquid and IBKR's trade records to analyze which trades incurred losses, identify any issues with parameter settings, and explore optimization opportunities for the strategy.
In this scenario, AI is not directly responsible for predicting market direction but is mainly used to process trade logs, identify anomalies, and assist in post-trade analysis. CBB continuously adjusts parameters based on market data, while his brother updates the code almost daily.
By the end of April, as the oil market cooled down, the two thought the high-yield phase was coming to an end. However, semiconductor stocks then became the new center of volatility, with targets such as Sandisk (SNDK) and Micron (MU) experiencing intense market movements similar to crypto assets, continuing to provide spreads for cross-market arbitrage.
According to the author, from May to July 2026, the robot's monthly trading volume remained between $1.5 billion and $2.5 billion, with weekly profits around $400,000 to $500,000.
The author admits that luck played a significant role: they happened to complete the system setup before the rise in volatility and encountered active cycles in precious metals, oil, and semiconductors. However, in their view, early bets on HIP-3, stock perpetual contracts, and TradeXYZ were also crucial prerequisites for the viability of this strategy.
10-Month Profit of $10 Million: How Long Can the Early Arbitrage Window Last?
As of early September 2026, CBB reported that this bot had executed approximately $32 billion in trading volume over 10 months, including transactions between HIP-3 and IBKR, accounting for about 1.5% of TradeXYZ's total volume, with accumulated profits exceeding $10 million.
All these figures are self-reported by the author, with no disclosure of account size, transaction records, fee calculation method, taxes, or audited income proof. The author also stated that based on varying capital utilization efficiency over time, the bot's annualized ROI from actual deployed capital ranged from 35% to 45%.
Such high returns rely on significant deployable capital and an immature market structure. On-chain stock perpetual swaps have seen rapid growth, but pricing still needs to anchor to the traditional market; the mismatch between 24/7 trading and traditional trading hours, liquidity discrepancies across platforms, and high funding rates collectively contribute to early arbitrage profits.
These conditions are evolving. Professional market makers are stepping in, competition is heating up in market data and execution speed, and Ethena has announced plans to expand its basis trading strategy to stock perpetual swaps. Citing Ethena data, CoinDesk reported that since 2026, the average funding rate for certain stock perpetual swaps has been significantly higher than Bitcoin perpetual swaps, a key reason institutional funds have started eyeing this market.
The variables to watch next are quite clear: can the open interest of stock perpetual swaps continue to grow, will the deviation between on-chain prices and traditional markets persist, how much will funding rates drop after institutional entry, and whether trading platforms can enhance oracles, risk controls, and cross-market fund allocation.
If the speed of liquidity growth outpaces the influx of arbitrage capital, this trading opportunity may still hold profit potential; however, if professional market makers swiftly narrow spreads and funding rates, the early team's phase of gaining excess returns through speed will conclude.
For CBB, this trade is nearing its end. For the on-chain stock market, the real test has only just begun: after the early arbitrage dividends fade, can it sustain growth based on genuine trading demand.
Welcome to join the official BlockBeats community:
Telegram Subscription Group: https://t.me/theblockbeats
Telegram Discussion Group: https://t.me/BlockBeats_App
Official Twitter Account: https://twitter.com/BlockBeatsAsia
