Haver Analytics
Haver Analytics
Global| May 14 2025

The Age of Constraints, Part 2, AI and the Energy Reality Check

The Age of Constraints, Part 2, AI and the Energy Reality Check

If Part I of this series mapped the fractures in the global economy’s supply-side foundations—labor, capital, and energy—this second installment turns to the potential workaround: artificial intelligence.

AI has been frequently pitched as the ultimate macroeconomic escape hatch. It promises to plug labor shortfalls, redeploy idle capital, and inject new life into flagging productivity trends. And in some areas, it already is. From medical imaging to generative models in finance, AI is no longer theoretical—it’s live, scaling, and impressive.

But it’s not weightless. Or frictionless. And it is certainly not energy-free.

A look at the long-run relationship between real energy prices and global per capita energy consumption—illustrated in the chart below—potentially suggests that AI may not ease the Age of Constraints so much as intensify it. Particularly when it comes to energy.

1. The Post-WTO Energy Boom

The chart shows a clear inflection point after China’s WTO entry in 2001. From that moment, global per capita energy consumption began a sharp and sustained rise. Importantly this accompanied a steep climb in real energy prices as well. And that wasn’t a coincidence.

The globalization era -led by China - that followed was extraordinarily energy-hungry. Manufacturing shifted eastward. Supply chains stretched across continents. Infrastructure boomed. Global living standards improved. But all of it—cheap goods, rising incomes, export-led growth—was underwritten by a surge in energy demand and a rise in the real consumer-facing price of energy.

This is the energy context into which AI now enters.

2. AI as Labor Substitute—But Energy Amplifier

AI’s appeal stems in part from its role as a scalable labor substitute. Ageing societies, shrinking workforce, and rising service demand have created a macroeconomic labor crunch. AI helps solve that—at least superficially.

But while a human worker clocks off, AI doesn’t. It runs 24/7. It draws from hyper scale data centers, cooled by vast industrial systems, powered by gigawatts of electricity. By some estimates, AI workloads could account for up to 70% of all new data center energy demand by 2030.

In this sense, AI is not a lightweight digital solution. It is capital- and energy-intensive infrastructure, hidden behind a user-friendly interface.

3. Capital Light, Infrastructure Heavy

AI also sidesteps traditional capital frictions. Unlike bridges or car factories, algorithms scale instantly, cross borders effortlessly, and face fewer geopolitical headwinds. But this illusion of capital lightness conceals a deeper constraint.

As McKinsey has noted, the global push to build AI-ready infrastructure could require over $5 trillion by 2030—mostly to finance energy-hungry compute power and cooling systems. The constraint hasn't been eliminated. It's just been relocated—from steel and cement to silicon and electricity.

AI may circumvent tariffs and supply chains. But it can’t outrun the laws of thermodynamics.

4. Energy Efficiency ≠ Energy Sufficiency

To be fair, AI is also helping manage energy smarter. It optimizes HVAC systems, streamlines logistics, and forecasts renewable output. But the energy it saves is still outweighed by the energy it consumes.

And critically, unless we see breakthrough innovations in energy production and storage—fusion, high-density batteries, or quantum efficiency gains—scaling AI means scaling energy demand. That’s the same logic that powered the post-WTO boom—and the same outcome: upward pressure on real energy prices.

5. Brains Still Need Brawn

The 21st-century growth model will not be purely digital. If anything, the digital revolution is re-exposing our dependence on the physical enablers of prosperity: abundant energy, resilient grids, mineral supply chains, and geopolitical stability.

The chart tells us something simple but profound: productivity-enhancing booms are energy-intensive. They were in the 2000s. They are today. And unless energy systems evolve in tandem with AI, we risk hitting the same bottlenecks—just faster.

Recent analysis by the International Energy Agency (IEA) reinforces this point. Global electricity demand from data centers, driven largely by AI workloads, is expected to more than double by 2030, reaching 945 TWh—equivalent to Japan’s total electricity use today. In the United States, data centers could account for nearly half of all incremental electricity demand this decade. AI-related compute is rapidly becoming one of the most powerful new forces shaping global energy needs.

And yet, unlike labor or capital, energy is not easily substituted. That insight underpins a deeper macroeconomic concern: if energy is a unique and non-substitutable factor of production, then a sustained rise in its real price will have profound effects on productivity, growth, and income distribution.

When energy is cheap and abundant, it acts as a scaling force—amplifying the productivity of both labor and capital. But when it becomes scarce or expensive, it imposes a drag on the entire production function. Even the most advanced technologies cannot operate without electricity, heat, or cooling. In this sense, energy is not just another input—it is a meta-input, one that conditions the effectiveness of every other factor.

The implications are far-reaching. Trend productivity growth may remain weak not because of insufficient innovation, but because the energy infrastructure needed to scale that innovation is increasingly expensive, politically fraught, or physically constrained. The same forces could also reshape factor income shares. If more value accrues to energy providers and infrastructure owners, labor’s share of GDP could further decline, not necessarily because workers are less productive, but because more of their productivity is being absorbed by rising input costs.

Moreover, higher real energy prices tend to suppress real wage growth, particularly in energy-intensive economies and sectors. If firms face persistently higher overheads, they may pass on fewer gains to workers, or curb hiring altogether (see chart below). This creates a risk of structurally weaker income growth and wider inequality, even in the midst of a technological boom.

In short: absent major breakthroughs in energy generation, storage, or transmission, the AI revolution will not escape the gravitational pull of physical constraints. It may even intensify them.

Coming Up: The Geopolitics of AI

In Part III, we turn to the geopolitical implications: how chip nationalism, energy competition, and digital sovereignty are shaping the AI race—and whether open innovation can survive in a world that increasingly favors control over collaboration.

  • Andy Cates joined Haver Analytics as a Senior Economist in 2020. Andy has more than 25 years of experience forecasting the global economic outlook and in assessing the implications for policy settings and financial markets. He has held various senior positions in London in a number of Investment Banks including as Head of Developed Markets Economics at Nomura and as Chief Eurozone Economist at RBS. These followed a spell of 21 years as Senior International Economist at UBS, 5 of which were spent in Singapore. Prior to his time in financial services Andy was a UK economist at HM Treasury in London holding positions in the domestic forecasting and macroeconomic modelling units.   He has a BA in Economics from the University of York and an MSc in Economics and Econometrics from the University of Southampton.

    More in Author Profile »

More Viewpoints