He who controls electricity will control intelligence. He who controls intelligence will define value.
In an era when artificial intelligence (AI) agents orchestrate the bulk of economic activity, the kilowatt-hour is emerging as the fundamental unit of account. Money, once a claim on goods or gold, is becoming a derivative of computational power, and computation itself is a derivative of electrons. This is not a metaphor. It is the inescapable physics of the AI age, and it is already reshaping the strategic architecture of global power.
The pivot is already legible in capital flows and infrastructure decisions of historic scale. Global data-center investment is projected to reach $6.7 trillion by 2030, with $5.2 trillion earmarked for AI-ready hyperscale facilities, according to McKinsey’s central scenario. A single hyperscale site consuming one gigawatt, sufficient to power roughly 800,000 homes, can cost up to $80 billion to build and equip, as IBM chief executive Arvind Krishna noted in late 2025. The International Energy Agency projects that data centers will consume 945 terawatt-hours of electricity by 2030, more than double today’s level. In the United States alone, data centers are expected to account for 8.9 percent of total electricity demand by 2030.
Goldman Sachs estimates that data-center power demand could rise as much as 165 percent by decade’s end, driven overwhelmingly by generative AI. They are electron factories whose output is intelligence.
The Return of Physics to Economics
For most of modern history, economics has been governed by abstractions (currencies, credit instruments, and financial claims) that ultimately represent promises about future productive capacity. Beneath those abstractions, however, lies a physical substrate (energy) that cannot be printed, inflated, or conjured by central bank decree. In the age of artificial intelligence, that substrate is reasserting its primacy with striking force.
As computation becomes the dominant driver of economic activity, the cost structure of intelligence converges on the cost of electricity. In competitive markets, the marginal cost of thinking machines tends toward the marginal cost of the kilowatt-hour. Training a frontier AI model can consume millions of kilowatt-hours. Inference consumes comparable total energy at scale. Elon Musk, whose xAI Colossus supercluster already operates at gigawatt scale, has crystallised this constraint. By mid-2026, electricity, not semiconductors, will be the binding limit on AI progress. Intermittent renewables, however abundant, cannot reliably satisfy the always-on, density-intensive demands of hyperscale computation. Physics, not finance, sets the ceiling.
The implication for economic measurement is profound. In the 20th century, value was denominated in money backed by sovereign credibility. In the 21st century, it may increasingly be denominated in energy-backed computation. Money becomes a claim on compute, which is ultimately a derivative of electricity. The loop is closed, and it runs through the power grid.
The Coming Agent-to-Agent Economy
Amplifying this shift is the rise of autonomous AI agents conducting machine-to-machine commerce on a scale that would have seemed fantastical a decade ago. In the emerging agent-to-agent (A2A) paradigm, autonomous systems negotiate contracts, execute financial trades, design products, and manage logistics in real time, without human intermediation. Forward-looking estimates suggest machine-to-machine transactions could reach ~$14.2 trillion by 2030, representing up to 15 percent of global gross domestic product (GDP).
In this environment, the relevant metric of economic activity is no longer hours of human labor but kilowatt-hours consumed per task. A legal document drafted, a software module written, a pharmaceutical molecule simulated: all become computational services priced according to the energy required to produce them. A2A agents will not merely exchange data or tokens—they will compete for compute capacity, which ultimately means competing for electricity. AI procurement bots will secure compute leases backed directly by reactor output. The cost of a negotiated contract, a generated design, a batch of code: all become energy-priced commodities in a marketplace where electrons are the settlement layer.
This reprices everything by embedded energy content. A data center’s valuation will be judged less by square footage than by gigawatt-hours reliably delivered. Commodities will trade partly on the basis of embodied energy from extraction through delivery. Saudi Arabia’s public investment fund (PIF)-backed HUMAIN partnership with xAI for a 500 megawatt (MW) AI data center hub in Riyadh illustrates the new geopolitics of electrons: oil-rich nations are converting hydrocarbon wealth into baseload power for silicon, repositioning themselves at the nexus of energy and intelligence.
From Petrodollar to Electrodollar: A Tectonic Realignment
The geopolitical analogy is historically resonant. In the 1970s, the United States forged arrangements with Saudi Arabia, ensuring that crude oil was priced and settled in dollars. That petrodollar system created structural global demand for the American currency and cemented financial dominance for half a century. A similar dynamic is now plausible in the age of AI.
Instead of oil flows determining monetary influence, electricity production capacity, particularly clean, reliable, dispatchable electricity, may become the foundation of economic power. If the United States secures dominance in small modular reactor (SMR) deployment, data center co-location, and AI infrastructure, the dollar could evolve into an “electrodollar,” the settlement currency of global AI compute. Nations would accumulate dollar-denominated energy tokens not to buy Treasury securities but to access the electricity and computational power on which their economic survival depends. The proposition is not symmetric. If China or another bloc becomes the dominant electricity hub, renminbi-denominated energy trade could accelerate de-dollarisation with structural consequences for American financial primacy.
The mechanism is straightforward: pasting American monetary dominance derived from “buy our debt, and you can access the global financial system.” The emerging proposition is starker and more physical. “Bring your dollars, and we will sell you the electricity and the compute you need to survive in the AI age.” Collateral shifts from Treasury paper to generated electrons, but if the United States controls those electrons, its dominance is not diminished; it is deepened, because physical infrastructure is far harder to replicate or sanction than financial instruments.
Small Modular Reactors and the New Reserve Asset
One technology stands apart in this strategic landscape is nuclear power, and specifically small modular reactors. Unlike solar or wind generation, nuclear reactors provide constant baseload electricity, precisely the always-on, high-density power required by data centers operating around the clock. SMRs promise co-location with compute infrastructure, long-term predictable electricity costs, and energy density unmatched by any renewable alternative. For hyperscale operators, a co-located reactor transforms electricity from an operating expense into a proprietary strategic asset.
The implications extend deep into monetary architecture. Considering the trajectory already visible in American policy, the Guiding and Establishing National Innovation for US Stablecoins Act (GENIUS) of 2025 requires payment stablecoins to maintain 1:1 reserves in USD, short-term Treasuries, or equivalent assets. In an energy-as-currency world, the next logical evolution is the energy-backed stablecoin, in which kilowatt-hour reserves produced by SMRs operating alongside data centers supplant T-bills as the low-risk anchor of digital liquidity. The logic is compelling. Unlike sovereign debt, electricity is physics-constrained and cannot be inflated. A kilowatt-hour produced is a kilowatt-hour that exists. In this framing, the progression is clear. Past safe assets were the T-bills, and future safe assets could be the verified, tokenised kilowatt-hour (kWh).
In decentralized finance, this means a new generation of protocols built around real-world energy assets, kWh-tokenised vaults, AI compute leases as collateral, and smart-contract-mediated electron allotments negotiated autonomously by A2A agents. Chainlink-style energy oracles could price electricity in real time, and decentralized exchanges could trade compute leases backed directly by reactor output. The blockchain, in this scenario, becomes the operating system of the energy grid.
A New Bretton Woods: The Compute-Energy Standard
The 20th century monetary order rested on the Bretton Woods system, anchoring currencies first to gold and subsequently to the US dollar. The 21st century analog may be a Compute-Energy Standard, in which the credibility of national currencies derives increasingly from the reliability and scale of their energy-to-intelligence pipelines. A country’s strategic reserves would not be measured in gold bars or Treasury holdings but in gigawatts of dispatchable, always-on electricity.
This framework is not merely theoretical. Governments and sovereign wealth funds are already racing to co-locate generation with compute. Tech giants, including Alphabet, Microsoft, Amazon, and Meta, collectively spent approximately $268 billion on capital expenditures in 2025, with projections reaching up to almost $700 billion in 2026, according to recent analysis, transforming data centers from infrastructure into fortified energy vaults. Microsoft has restarted Three Mile Island in Pennsylvania, securing 835 MW of dedicated nuclear output. Amazon has committed to X-energy’s SMR programme at a scale of 5 gigawatts (GW). Google has contracted with Kairos Power for 500 MW. JPMorgan’s technology spending approaches $20 billion in 2026, much of it underwriting AI infrastructure with direct energy implications. These are not technology investments. They are energy sovereignty investments dressed in silicon.
In Washington and Gulf capitals alike, the strategic question has evolved. It is no longer “Will energy become the basis of money?” It is “Which nations will secure enough reliable electricity to function as monetary superpowers in the AI century?” The nations that answer correctly will define the economic hierarchy of the coming decades.
Strategic Imperatives: Four Urgent Actions For Leaders
The window for decisive action is narrow. Four imperatives demand immediate attention from policymakers and business leaders in equal measure.
First, accelerate SMR permitting and deploy multi-billion-dollar public-private investment in nuclear baseload capacity. Regulatory streamlining is not a concession to the industry, but it is a prerequisite for monetary sovereignty.
Second, mandate or powerfully incentivise data center energy self-sufficiency through co-location of generation and compute. Dependence on external grid power is a strategic vulnerability.
Third, forge international SMR and power-grid alliances among like-minded democracies, creating an energy equivalent of semiconductor export controls, an architecture that shapes who can access AI-grade electricity and on what terms.
Fourth, evolve the GENIUS Act framework to permit energy-backed stablecoins while preserving financial stability, creating the regulatory foundation for the electrodollar before competitors do.
Winners in this regime will hoard electrons. Data centers will become fortified vaults, and kilowatt-hours will serve as the ur-currency of the digital economy. Those who delay will find themselves clinging to fiat instruments whose purchasing power erodes relative to the physics-constrained intelligence produced by their competitors. Energy is no longer merely a cost of doing business in the digital economy. It is the throne upon which economic power in the AI age will be seated.
The Kilowatt World Has Arrived
The tectonic shift in global finance is not in the form of money but in its substrate. The economic system of the coming decades may be defined by a simple chain of logic. Energy generates computation, and computation produces intelligence, and intelligence creates value. If artificial intelligence becomes the primary engine of productivity, then the ultimate source of economic value returns to a physical constant that cannot be printed, manipulated, or inflated away.
Those who grasp this first, securing reliable electron flows, pricing intelligence at its true marginal cost, re-engineering monetary architecture around power grids and SMR pipelines, will define the economic hierarchy of the AI century. The dollar’s future as the world’s reserve currency depends less on the credibility of the Federal Reserve than on whether American industry and policy can deliver the gigawatts that global AI demand will require. The kilowatt world is not approaching on a distant horizon. It has arrived.
About the Author: Stella Kim
Stella (SuHee) Kim is an investment and nuclear strategy expert with over a decade of experience bridging global finance and deep-tech, with a particular focus on small modular reactors (SMRs). She is the CEO of Pandia Bridge, a Singapore-based advisory firm that connects global investors and leading Korean conglomerates with top SMR developers, including NuScale Power, to facilitate cross-border investments and strategic partnerships. Her work centers on the intersection of energy security, technology innovation, and strategic finance on a global scale. She holds a BA from Ajou University in South Korea and participated in a study abroad program at the Luleå University of Technology in Sweden.
