One-Page Summary
This report examines whether AI computation tokens (tradeable digital units that grant access to AI services such as model training, inference, and data processing) are becoming a new kind of asset. For decades, electricity served as the dominant metaphor for understanding how a general-purpose technology becomes a priced input to every sector of the economy. AI computation tokens move that idea from metaphor to practical reality: they allow the right to use AI computing power to be bought, sold, and held across borders in ways that electricity never could.
The global shortage of high-performance GPUs, the rise of decentralised compute networks such as Bittensor and Render, and the entry of institutional investors through vehicles such as the Grayscale Decentralized AI Fund all point toward the same question: should tokens that grant access to AI computation be treated as prepaid expenses, intangible assets, or financial instruments? The answer affects how billions of dollars are reported on corporate balance sheets, and the accounting standards are only now catching up.
In the early twentieth century, some economists proposed that electricity could serve as a future currency. The logic was compelling: electricity is a universal input to economic production, its consumption can be metered precisely, and everyone needs it. The idea never materialised, largely because electricity cannot be stored economically and is tied to physical transmission infrastructure within national borders.
AI computation tokens realise a version of this idea that electricity never could. Unlike a kilowatt-hour, a token representing access to AI inference or model training can be held in a digital wallet, traded on global exchanges, and used whenever and wherever the holder chooses. A researcher in Auckland can purchase a token that grants access to a GPU cluster in Frankfurt, and that token carries the same rights regardless of which jurisdiction the holder or the hardware sits in. This borderless, storable property is what separates AI tokens from the electricity parallel and gives them the character of a financial asset.
The key difference is functional. Electricity serves one role: it is consumed. AI tokens serve at least two. They act as a medium of exchange (used to pay for AI computation) and simultaneously as a store of value (held in the expectation that demand for AI compute will drive prices higher). This dual role raises accounting questions that never arose with utility meters. If a company holds tokens partly to run its AI workloads and partly because it expects the tokens to appreciate, how should those tokens appear on the balance sheet? Should the portion held for use be treated as a prepaid expense while the portion held for appreciation is marked to market? The answer is not obvious, and the distinction between use and speculation blurs in practice.
The electricity parallel is therefore useful less as a prediction and more as a way of thinking about it. It shows that when a general-purpose input becomes scarce and valuable, markets develop instruments to price and allocate access to it. AI tokens are that market mechanism for computation. Whether they stabilise into a recognised asset class or remain a volatile niche depends on how accounting standards, regulatory frameworks, and institutional infrastructure evolve over the next several years.
AI tokens sit at the intersection of three major structural trends: the explosive growth in demand for AI computation, the tokenisation of access to real-world assets and services, and the maturation of accounting and regulatory frameworks for digital assets. Each of these trends is individually consequential; their convergence around a single financial instrument makes AI tokens a rewarding subject for scholars across accounting, finance, and economics.
For accounting scholars, the central question is classification. Are AI tokens held for service access prepaid expenses, intangible assets, or financial instruments? The answer affects whether fair value volatility flows through earnings, whether holdings are disclosed in financial statements or only in risk factors, and how auditors verify the existence and valuation of token positions. For finance scholars, the correlation and volatility characteristics of AI tokens raise questions about portfolio allocation, treasury management, and whether token-based compute access can serve as a hedge against rising AI infrastructure costs. For regulators, the convergence of AI governance (EU AI Act) and crypto regulation (MiCA, SEC enforcement) creates compliance challenges that no single framework currently addresses. These are not speculative questions for a distant future: $25 billion in market capitalisation, growing institutional investment, and the imminent full effect of ASU 2023-08 make them questions for today.