Commodification of Compute

TL;DR

The paper proposes the Global Compute Exchange (GCX) to commodify compute by turning compute hours into a tradable asset using a layered blockchain-enabled marketplace. It defines Compute Hours (CH) as the standard unit, supported by standardized benchmarking, a reference system, and ESG-based resource grading, enabling fair pricing and reliable delivery. GCX combines spot and derivatives trading (futures, options, perps) with offchain/onchain components, clearing pools, insurance, and risk management to hedge volatility and optimize utilization across global providers, including DePINs and data-center operators. By drawing analogies to oil and power markets and detailing a scalable architectural framework, the GCX aims to democratize access to high-performance compute, reduce idle capacity, and catalyze innovation across AI, fintech, media, and science.

Abstract

The rapid advancements in artificial intelligence, big data analytics, and cloud computing have precipitated an unprecedented demand for computational resources. However, the current landscape of computational resource allocation is characterized by significant inefficiencies, including underutilization and price volatility. This paper addresses these challenges by introducing a novel global platform for the commodification of compute hours, termed the Global Compute Exchange (GCX) (Patent Pending). The GCX leverages blockchain technology and smart contracts to create a secure, transparent, and efficient marketplace for buying and selling computational power. The GCX is built in a layered fashion, comprising Market, App, Clearing, Risk Management, Exchange (Offchain), and Blockchain (Onchain) layers, each ensuring a robust and efficient operation. This platform aims to revolutionize the computational resource market by fostering a decentralized, efficient, and transparent ecosystem that ensures equitable access to computing power, stimulates innovation, and supports diverse user needs on a global scale. By transforming compute hours into a tradable commodity, the GCX seeks to optimize resource utilization, stabilize pricing, and democratize access to computational resources. This paper explores the technological infrastructure, market potential, and societal impact of the GCX, positioning it as a pioneering solution poised to drive the next wave of innovation in commodities and compute.

Figures (8)

• Figure 1: The Global Compute Exchange (GCX) Platform architecture is divided into several layers, each with specific functions. The Market Layer includes participants such as hedgers, traders, and market makers. The App Layer hosts various applications and user interfaces for interaction with the GCX. This also contains the GCX App and has an API and SDK for other apps to be built in this layer offering trading of compute to users. The Clearing Layer consists of guarantors providing proof of compute capacity and ensuring delivery. The Risk Management Layer features a risk engine with various components. The Exchange Layer (offchain) operates the core trading functions for compute resources. The Blockchain Layer (onchain) contains the smart contract ecosystem, including staking, liquidation engines, DeFi connectors, compute validators, customer collateral, and insurance pools. This multi-layered structure ensures a secure, efficient, and transparent market for trading compute resources.
• Figure 1: Compute trends across three eras of machine learning. The table presents the periods, the data set size, the scale of compute from start to end, and the corresponding doubling times. In the Pre Deep Learning lecun2015deep era (1952 to 2010), all models exhibit a scale increase from 0.00003 to 200 TFLOPs with a doubling time of 21.3 months. The Deep Learning era (2010 to 2022) saw regular-scale models increase from 700 to 2,000,000 TFLOPs with a significantly faster doubling time of 5.7 months. Finally, the Large-Scale era (September 2015 to 2022) involved large-scale models scaling from 4,000,000,000 to 800,000,000,000 TFLOPs with a doubling time of 9.9 months. From sevilla2022compute.
• Figure 2: Trends in FLOPs for 121 milestone ML models between 1952 and 2022. The graph illustrates the growth in computational requirements (FLOPs) over time, highlighting three distinct eras. A noticeable change in slope occurs around 2010, corresponding to the advent of Deep Learning. Additionally, a new large-scale trend emerges in late 2015, indicating further advancements in ML model complexity. Replicated from sevilla2022compute.
• Figure 2: This figure illustrates different strategies used by various market participants in the compute market. Founder Alice locks in the price of compute for her AI startup using a futures position, effectively hedging against future price increases (consumer hedge). Speculator Bob buys put options, anticipating a market correction to secure a price floor and minimize potential losses (trade). Datacenter Owner Carol sells both call and put options, collecting premiums. When these options expire worthless, she profits from the premiums, generating additional yield and smoothing her revenue stream (producer hedge). The figure demonstrates how these strategies can benefit different actors in the market by managing risk and optimizing financial outcomes hull2018options.
• Figure 2: Comparison of Power and Compute Markets: This table outlines the key differences between power and compute markets, including trading units, storage, volatility (see AWS_SpotGoogle_PreemptibleAzure_Spot for the reference to 90% spot discount), market fragmentation, peak vs. off-peak pricing (PPA is a Power Purchase Agreement mendicino2019corporate), and typical futures contract terms.