Author: Arthur Hayes
Compiled by Deep潮 TechFlow
DeepChain Overview: Throughout the evolution of crypto finance, each paradigm shift in assets has been accompanied by the emergence of new foundational derivative components. In 2016, Arthur Hayes and his BitMEX team invented the perpetual swap, fundamentally reshaping the global crypto derivatives trading landscape; in 2023, he proposed the concept of a synthetic dollar hedged by spot long positions against inverse contracts in his paper "Dust on Crust," directly catalyzing the rise of Ethena (USDe), a synthetic dollar protocol now exceeding $10 billion in size. Today, Arthur Hayes returns with his third major vision: Flop Network.
Faced with the geometric surge in computing power demand in the age of artificial intelligence, alongside the fragmentation, opacity, and lack of a unified spot benchmark in the traditional cloud computing market, Hayes proposes anchoring to physical-level floating-point operations per second (FLOPS), establishing a decentralized inference verification network to define a global standardized spot exchange rate for computing power, and building, on this foundational asset, standardizable computing power forward contracts (FCUs) as well as cash-settled differential quantification futures and perpetual contracts.
This paper comprehensively derives the complete mathematical framework for the assetization and financialization of computing power, along with the market-making and hedging logic, revealing the ultimate blueprint for redefining macro liquidity and computing power pricing rights in the age of AI agents through crypto-native finance.

I want to achieve “Three for Three.” I’m not talking about basketball, but rather my ability to conceive entirely new financial derivatives or use cases and witness them evolve into industry standards.
My historical trading record:
Perpetual Swap: A Huge Success (in Polarity's tone)
In May 2016, my team at BitMEX invented the perpetual swap (commonly referred to as “perps”). Perpetual swaps quickly became the most heavily traded financial derivative in the history of the crypto industry. Due to their immense liquidity and popularity, this same mechanism now powers around-the-clock, 24/7 trading of RWA (real-world asset) derivatives on innovative platforms like Hyperliquid. [1]
Synthetic USD Stablecoin: A Huge Success (in Polat's tone)
In March 2023, I authored an in-depth blog post titled "Dust on Crust," systematically explaining how to construct a synthetic dollar through a basis hedge combining a spot Bitcoin long position with a short perpetual futures contract. This synthetic dollar could be packaged into a stablecoin and widely used across the crypto capital markets for payments and settlements. The Ethena team adopted my concept and executed it with perfect engineering. Ethena’s synthetic dollar, USDe, became the fastest-growing stablecoin in history, surging from $0 to over $15 billion in circulating supply in less than 18 months.
Flop Network Spot Hashrate Market: To Be Announced (TBD)
This summer, I first publicly shared the concept of the Flop Network. $FLOP is the "fuel" for your AI agent. Agents will use $FLOP in commercial interactions because $FLOP can be directly exchanged for physical compute power, defined as Floating Point Operations Per Second (FLOPS). If Flop becomes the native payment and settlement network among AI agents, it naturally follows that the exchange rate between $FLOP and compute power will serve as the benchmark anchor for the global spot price of compute.
A truly globally liquid spot computing power market will inevitably give rise to a derivatives market built upon this spot benchmark. May this article serve as the authoritative blueprint for all ambitious entrepreneurs, guiding them in building a computing power derivatives market tailored to the needs of the new AI agent era.
Core pain points
There is currently no globally unified benchmark for computing power pricing. Although various indices exist in the market attempting to track the sale and rental prices of specific CPU, GPU, and TPU chips, the data samples are severely incomplete. Additionally, many indices attempt to track the invocation prices of AI tokens on a per-model basis. However, both chip prices and model token invocation prices are merely secondary derivatives of computing power. If I could accurately determine the benchmark price of computing power (FLOPS), I could derive from the top down the intrinsic value of any chip and the fair cost of any model token. We should not waste effort collecting fragmented and imperfect chip and token price data; instead, we must focus on establishing a unified price discovery mechanism for the underlying physical computing power.
Solution: Flop Network Hashrate Marketplace
The Flop Network produces one block per second. Each block contains verifiable receipt proofs that confirm miners have successfully completed inference tasks for specific agents. For each inference task, we can obtain three core parameters to precisely calculate the exchange rate between $FLOP and physical computing power.
The block data object contains:
Delay time (in seconds): The actual time taken to complete this inference request.
Work completed: Total number of floating-point operations performed (FLOPS)
Transaction fee: The amount of $FLOP earned by miners for executing the task.
$FLOP / (hashes per second) = [3] / [2] / [1]
The spot price is only a true and valid market price when everyone has the freedom to trade. On the supply side, as long as you possess computing chips—whether GPU, CPU, or TPU—you can mine $FLOP by completing inference requests from agents. On the demand side, as long as your wallet holds $FLOP, you can submit inference tasks at any time. Flop Network is an open-source network accessible to all of humanity. Therefore, the spot computing price formed here will become the globally recognized benchmark for computing power.
Flop Computing Unit (FCU)
Before delving into cash-settled financial derivatives, let’s continue exploring the physical-delivery compute market. A key feature of Flop Network is its ability to allow demand-side users to pre-book and reserve computing power for delivery up to 30 days in advance. To enhance liquidity and tradability, the network standardizes computing power into “Flop Computing Units” (FCUs).
Imagine you know with certainty that you’ll have a large computational task to process in 7 days, and you’d like to lock in the cost of compute power today. You can place a bid to reserve a specific amount of compute power for delivery 7 days from now. After the 7-day period, miners have a 24-hour window to fulfill the delivery. Once your bid is matched and accepted by a miner, the reservation generates a standardized 7-day FCU (Compute Power Unit) object, which is essentially equivalent to a deliverable forward contract. If the spot price of compute power rises significantly before delivery and you no longer need the capacity, you can directly resell this FCU on the Flop Network’s native marketplace to other participants who need it.
Here are specific business examples of FCU in actual operation. Please continue to monitor updates to Flop’s official documentation, where all technical details will be fully presented.
Single FCU standard object specification:
Delay time: 1 second
Computing power: 1 PetaFLOP (quadrillion floating-point operations per second)
AI model: Specified by the purchaser
Delivery time: 12:00 UTC, 7 days after creation
Note: FCU covers only inference scenarios and is not applicable to model training. The overall value of a model training compute cluster far exceeds the simple sum of its individual components; in contrast, inference tasks are inherently highly divisible, which is the fundamental premise for the existence of a single benchmark unit, FCU.
For each AI model, up to 30 FCU contract tenors may be actively traded on the market, each corresponding to a unique expiration date. If you require 1 ExaFLOP of computing power, you simply purchase 1,000 FCUs. As a buyer, to attract miners to accept your order, you must publish a $FLOP bid indicating the price you are willing to pay for the FCU, and you must fully pre-fund your order for network verification. As a seller (compute provider), you publish a $FLOP ask indicating the price you wish to receive, and you must stake an amount of $FLOP weighted by your stake as performance collateral, ensuring that if you default, other miners can step in to fulfill delivery at a profitable cost on the spot market. This staked amount is dynamically determined by the network based on the total recent inference volume processed, incorporating a baseline $FLOP cost and penalty multipliers to deter malicious behavior and delivery defaults. This mechanism creates a fully funded, two-sided order book that eliminates junk orders entirely: funded buy and sell orders are settled on-chain, while a high-frequency, continuous two-sided order book and matching engine operate off-chain—identical in mechanism to the spot market.
The forward market inherently carries a risk of failure to deliver. As previously mentioned, to mitigate this risk, FCU compute providers must post a substantial $FLOP performance bond significantly exceeding the inference fees they receive. If a miner fails to deliver the agreed-upon compute power within 24 hours after expiration, their entire $FLOP bond will be forfeited. At this point, backup miners will bid to take over the order, as the per PetaFLOP fee they earn for fulfilling the contract far exceeds the prevailing spot market price. Should the spot price surge above the bond amount and no miners are willing to fulfill the contract, the system will deem a default to have occurred, entitling the buyer to a full refund plus the entire amount of the forfeited bond as additional compensation.
FRCR futures contract
FRCR: Flop Reference Compute Rate
Hourly volume-weighted $FLOP / GigaFLOPS settlement fee
Above, I broke down how real demand parties can procure physical compute power through $FLOP in both spot and forward markets. Now, let’s turn to exploring how to construct cash-settled financial derivatives that can be listed on both traditional finance (TradFi) and cryptocurrency exchanges. I will outline a complete contract design and meticulously walk through how market makers can hedge long or short futures exposures.
Contract Specifications:
Contract Name: FRCR Quanto Quarterly Futures Contract; Delivery Expiration: Last Friday of March, June, September, and December at 12:00 UTC; Underlying Index: FRCR.
Contract multiplier: $1 per point
Example: Assume the FRCR index is at 100 points.
Contract value: multiplier × FRCR = $1 × 100 = $100
Margin currency: USD
If you went long 1 contract at an entry price of $100 FLOP and settled at $200 FLOP, your profit and loss (PNL) is as follows:
Leverage × Number of Contracts × (Settlement Price - Entry Price) = $1 × 1 × (200 - 100) = $100
If you shorted 1 contract at an entry price of $200 FLOP and settled at $100 FLOP, your profit and loss would be:
Leverage × Number of Contracts × (Settlement Price - Entry Price) = $1 × (-1) × (100 - 200) = $100
Quanto derivatives are a type of exotic derivative. A Quanto structure is necessary because it is assumed that the vast majority of market participants wish to deposit margin in USD and withdraw profits in USD, while the underlying spot market is priced in $FLOP. Therefore, we must employ a Quanto structure to achieve the desired payoff profile that traders seek. To price Quanto derivatives accurately, market makers must dynamically adjust risk premia based on the covariance between the FRCR index and the $FLOP/USD exchange rate.
If the market prefers to directly use $FLOP as the margin asset, the contract design can be simplified to:
Contract multiplier = 1 GigaFLOP
Contract value = Multiplier × FRCR = 1 × 100 $FLOP = 100 $FLOP
The success or failure of any financial derivatives contract hinges on whether market makers can easily replicate long and short futures positions using the underlying asset.
Long futures position replication:
To replicate a long exposure, market makers need to acquire $FLOP and purchase a series of consecutive FCUs. Below are the specific steps for market makers to synthetically create a 60-day FRCR long futures position:
Spend $FLOP to purchase a 30-day FCU; upon expiration in 30 days, roll over into another 30-day FCU.
On November 1, the 30-day FCU price was $100 FLOP; on November 30, you settled your original 30-day FCU at $100 FLOP and purchased a new 30-day FCU expiring on December 30 at $200 FLOP.
You have successfully synthesized a 60-day long futures contract, locking in an entry price of $200 $FLOP.
Short futures position replication:
To replicate a short position, market makers must acquire $FLOP and offer a series of continuous-term FCU computing power supply contracts. This requires additional $FLOP capital, as the staking collateral far exceeds the computing power fees collected. Market makers do not need to physically purchase chips, as they actively roll over and close their positions before settlement. Below is the specific process for synthetically creating a 60-day FRCR short futures position:
Post $FLOP as margin and place an order to provide a 30-day FCU; before the 30-day maturity, roll over this position into another 30-day FCU supply contract.
On November 1, the 30-day FCU price was $100 FLOP; on November 30, you closed your existing 30-day FCU by buying it back at $100 FLOP and placed a new 30-day FCU, expiring on December 30, for sale at $200 FLOP. As long as you roll over the FCU before its expiration or designate an alternative miner to fulfill the contract within the 24-hour delivery window, you will never default, and your margin remains secure.
This perfectly replicates a 60-day short futures contract with an entry price locked at $200 $FLOP.
FRCR Perpetual Contract
Once you fully understand how futures contracts work, extending them to perpetual contracts becomes straightforward. Remember that perpetual contracts are essentially a continuous chain of futures positions with infinite rollover.
Perpetual Contract Specifications:
Contract Name: FRCR Quanto Perpetual Contract; Underlying Premium Index: 8-hour Time-Weighted Average Price (TWAP) of [Perpetual Contract Price - FRCR Index].
The perpetual contract price will be determined using the hourly volume-weighted average price to strictly align with the FRCR calculation method, with eight samples collected over an 8-hour period and averaged arithmetically.
Contract multiplier: $1 per point
Example: Assume the FRCR index is at 100 points.
Contract value: multiplier × FRCR = $1 × 100 = $100
Margin currency: USD
The funding rate mechanism is identical to that of traditional perpetual contracts. If the premium index is positive, the funding rate is positive, and longs pay funding fees to shorts; if the premium index is negative, the funding rate is negative, and shorts pay funding fees to longs. As this is a Quanto derivative, all funding fee settlements are made in USD.
To create a synthetic long perpetual contract, traders simply go long on FCUs and continuously roll them over. For example, to establish a 120-day synthetic long perpetual exposure, a trader need only buy one 30-day FCU and roll it over four times. Conversely, to create a synthetic short perpetual contract, traders simply place sell orders for FCUs and continuously roll them over. For instance, to establish a 120-day synthetic short exposure, a trader need only place a sell order for one 30-day FCU and roll it over four times.
AI Model Adaptation Options
A subtle challenge facing the compute derivatives market is that the benchmark compute price underlying financial derivatives is model-agnostic; however, in reality, each distinct AI model corresponds to a separate FCU trading market. Therefore, to accurately synthesize futures and perpetual contracts, traders must buy or supply a basket of FCUs across individual models, weighted precisely to match the FRCR composite index.
Along with publishing the composite index, FRCR simultaneously discloses granular data for each model asset pool, ensuring complete transparency in the weighting of the basket of assets. The composite index drifts as the model invocation structure changes; for hedgers, the primary risk to manage shifts from physical depreciation of underlying hardware to the risk of model iteration and retirement.
Zero-friction trading
To foster and expand the derivatives market based on spot and FCU hash rate prices from the Flop Network, FCU trading incurs no additional platform transaction fees. Both parties need only pay the network’s base $FLOP fee to validation nodes for transaction inclusion. This fee is identical to the standard $FLOP transfer fee between wallets.
The Core Advantages of Flop Spot and Futures Markets
Anyone can participate in hash rate purchasing or hash rate supplying on Flop Network without any barriers.
All settlement transactions are executed in a fully open and transparent market.
All data calls generated by the Flop Network are completely free.
It’s that simple and clear. If you’re committed to building any form of hashpower derivatives, all exploration must begin with a deep understanding of on-chain activity and price discovery in the spot and forward markets of Flop Network.
Hail to the king!
[1] RWA: Real-World Assets (e.g., crude oil or gold).
