What will be the most valuable resource in an economy powered by artificial intelligence?
Data? Computing capacity? Algorithms? Or electricity?
As data centers expand and electricity becomes an increasingly critical production input, a provocative idea is gaining relevance: the megawatt-hour may be emerging as a new unit of economic value.
But could a megawatt-hour actually function like money?
Not Money in the Traditional Sense
Economists typically identify several fundamental functions of money, including serving as a unit of account, a medium of exchange, and a store of value. In the digital economy, electricity is beginning to perform functions that resemble some of these roles. First, the megawatt-hour provides a physical reference point for understanding the cost and productive capacity of digital infrastructure. Electricity consumption can be related to AI model training, inference, computing services, and data center operations. Second, companies increasingly secure future electricity supplies through long-term power purchase agreements (PPAs), investments in generation facilities, and dedicated energy infrastructure. These arrangements do not literally store economic value in megawatt-hours. Rather, they protect future access to a critical production input and help manage exposure to energy costs and supply constraints. Microsoft, Google, and Amazon have all developed extensive energy strategies involving long-term electricity procurement, low-carbon generation, and partnerships supporting new power infrastructure. These initiatives are not simply sustainability programs. They are increasingly central to strategic risk management.
Why One MWh Is Not Always Equal to Another
Physically, one megawatt-hour always represents the same quantity of energy. Economically, however, its value can vary dramatically. Electricity available at noon during a period of abundant solar generation is not equivalent to guaranteed electricity delivery during an evening demand peak. The economic value of electricity depends on several factors: delivery time, location, emissions intensity, dispatchability, grid conditions, and reliability. This distinction is crucial. Unlike money, megawatt-hours delivered at different times and locations are not freely interchangeable. Electricity therefore cannot serve as a universally accepted medium of exchange. It can, however, function as a form of functional quasi-money: a unit through which economic value is increasingly measured, priced, and secured within specific parts of the digital economy.
The Risk Management Perspective
If access to electricity determines whether a company can expand its operations, securing that access begins to resemble protecting a strategic productive asset. Energy strategy must therefore extend beyond managing price volatility. Companies also need to assess volume risk, delivery profiles, counterparty exposure, infrastructure constraints, and the alignment between contracted electricity and actual operational requirements. A contract that stabilizes electricity prices does not necessarily guarantee uninterrupted delivery to a particular facility. Likewise, owning computing hardware does not guarantee that the equipment can be operated economically.
From Digital Assets to Physical Constraints
The economic value created by AI ultimately rests on a physical foundation. Algorithms require processors. Processors require electricity. Electricity requires generation, transmission infrastructure, and system reliability. In this context, the megawatt-hour becomes more than a technical measurement or an operating expense. It becomes a way to understand the physical scarcity underlying digital production. The megawatt-hour will not replace money. But it may become one of the most important units for understanding the productive capacity and investment value of the AI economy.
Perhaps the most important question for future AI investments will not be, "How many processors do we own?" It may be, "How much reliable electricity have we secured to operate them?"
