The Myth of an Energy Transition

The summer 2026 news cycle continues to be dominated by the Iran War and its vicissitudes and consequences. The disruptions to oil and natural gas supplies and the economic fallout are animating articles with headlines such as Bloomberg’s last month, “Iran Shock Jolts Asia and Europe to Speed Up Energy Transition.”

The energy transition is the idea that the United States and the world are already starting to see the wholesale replacement of oil, natural gas, and coal by three politically favored energy technologies: solar and wind, moderated by batteries.

The claim that an energy transition is underway, much less that it can be sped up, is an illusory but tenacious idea anchored in a claim of technological inevitability: fossil fuels are the buggy-whips of the energy industry and thus face inevitable obsolescence. And, of course, the transition ideology is further animated by the cult of climate catastrophism that demands the elimination of hydrocarbons.

Meantime, the promised solar-wind energy revolution seems to be losing appeal. A recent Pew poll found public support for prioritizing wind and solar has dropped by over 20 percentage points in the past six years.

Nonetheless, the energy-transition ideology has captured the political process, a fact evident in a key metric: money. Over the past two decades, the U.S. and Europe have collectively spent nearly $20 trillion to encourage, subsidize, and even mandate an energy transition. That rivals the total collective spending on all social programs on both continents over the same period. Whatever one thinks about social programs, they have accomplished far more than energy transition programs.

Consider oil. The core relevant metric—global oil use per capita—is essentially the same today as two decades ago despite gushers of capital and tens of millions of heavily subsidized electric vehicles. That single fact exposes the fatuousness of the narrative that society is decoupling from oil. Not only does oil remain the world’s largest single source of energy today, but its global use has risen with population growth.

The same is true for natural gas and coal. The world is using far more of both compared to two decades ago. The data show nothing vaguely resembling an energy transition. What we’ve witnessed instead is the reality of continual additions to energy supplies from all sources, including from the politically favored ones. (Even wood for fuel is in greater use now than at any time in history.)

The only consequential structural change in energy markets over the past two decades has been the emergence of the United States as the world’s biggest producer and exporter of natural gas and oil.

To quote the science fiction author Philip K. Dick, “Reality is that which, when you stop believing in it, doesn’t go away.” Policymakers and the penumbra of influencers can choose to believe otherwise—until the next inevitable disruption.

Today, we’re facing two simultaneous energy shocks: one of supply and one of demand.

The supply shock, the Battle of Hormuz, is centered on oil as well as its derivate products and natural gas. The demand shock comes from the explosive growth of artificial intelligence that has upended the best laid plans in electricity markets. Both epitomize a reality that hasn’t gone away. And each exposes the bankrupt ideology of an energy transition.

The Battle of Hormuz shattered any complacency about global oil supply chains. And it illuminated a specific reality. Oil is essential for all commerce because it powers the transportation of all goods and materials as well as the people who enable all services. Without the physical availability of oil, all commerce stops. Oil today powers over 95 percent of all global transportation machinery.

As Vaclav Smil chronicled in his book Prime Movers of Globalization, the proliferation of oil-fueled combustion engines was the prime driver of the twentieth century’s great economic expansion.

The decoupling narrative, however, eagerly highlights the metric showing that oil’s share of our economy is declining. That isn’t a decoupling—it’s a decrease in the sensitivity of the economy to the price of oil. Consider a relevant analogy. While farming has evolved to become a small share of a bigger overall economy, that doesn’t mean that we’ve decoupled from the need for food.

But recovery from the Hormuz shock will have a long tail as importing nations reprice security. The transitionists have seized on this as an opportunity to resuscitate flagging political support for their favored and subsidized alternatives, despite the evidence of failure after trillions of dollars deployed over the past two decades. But the transitionists will likely succeed to some extent because of the seductiveness of the tenet of technological inevitability.

Of course, old technologies often yield to new ones. “Clean tech” and “energy tech” are seductive synonyms for wind, solar, and batteries because the terms imply that energy-technology progress can follow the same kind of transformation seen in communications tech—going from the pony express to landline phones to wireless.

But energy materials in particular, and many materials in general, occupy a unique corner in the tech universe. Many foundational materials, including hydrocarbons, have either no replacements or none that are affordable at the scales needed. We not only still use such ancient materials as stone, wood, and glass, for example, but also in far greater quantities than ever in history.

A related tenet of the transition ideology is that technology, especially software now accelerated by AI, will deliver efficiency gains at a pace that will reduce overall energy demand. History, however, shows the inverse. Improved energy efficiency is associated with—arguably stimulates—continual increases in overall energy demand because efficiency has the intended effect of reducing the cost of using the associated service or product, which in turn stimulates demand. More importantly, the critical net effect of technological progress is that it drives wealth expansion which, in turn, also creates more energy demand.

Add to this the fact that humans continually invent new energy demand. There was no demand for gasoline until the invention of the car, nor for aviation fuel until the invention of the aircraft, and so on.

This brings us to the energy demand shock. It isn’t news that economic growth in the twenty-first century is being driven by silicon engines, not combustion engines, i.e., by computers, not cars—especially with the arrival of artificial intelligence at scale.

A decade ago, it was recognized by those inside the computing community that AI was rapidly evolving from a specialized tool to become a general-purpose technology. It was also well known that AI is astonishingly energy-hungry. It was similarly widely believed back then—though still hotly debated today—that the arrival of useful, affordable AI would boost economies. What was not known was what year, specifically, general-purpose AI would break out.

Now we know, and we’re relearning the lesson that economic growth is impossible without increasing energy demands.

The power needed per single AI data center is truly staggering, as is the number and velocity of data centers under construction. Dozens are already underway, and hundreds have been announced, many with a power demand well over 200 megawatts (MW), enough to serve a city of 100,000 homes. The 1,000 MW scale—gigawatt scale—data centers under construction each match the power demands of a city of half a million people.

For context on the grid-engineering implications, over the past 60 years, the United States’s public grids have added only about a dozen industrial facilities with electricity demands exceeding 200 MWs. All were huge chemical factories, oil refineries, steel mills, aluminum smelters, or LNG export terminals.

But in the past half-dozen years, eight data centers, each well over 200 MW, have already been connected to public grids. And that’s just the warm-up act. Hundreds more at that scale, and bigger, are under construction or planned. Most of today’s publicly regulated electricity industry is incapable of meeting that kind of growth for both technical and regulatory reasons. Demand is mounting faster than most utilities can make decisions, much less build power plants and transmission lines.

This demand shock is coming at a velocity and scale unanticipated by either the punditocracy or the transitionists. As with the supply shock from the Battle of Hormuz, there will be a long tail to the consequences.

One such consequence is the rapid rise of a data-center resistance movement. While much of the opposition is fueled by the professional grievance industry, some challenges are real. One of the central issues is figuring out how big the demand shock might yet become.

All tech revolutions involve forecasting challenges. Few imagined, for example, when the automobile emerged that we’d end up with more cars in America than drivers. Or that the proliferation of automobiles would stimulate entirely new derivative services and industries.

Similarly, few imagined circa 1984 that the invention of an affordable personal computer would mean we’d have more computers than humans. Or that it would stimulate the creation of entirely new derivative and transformational services, not just e-mail but also, for example, e-commerce and Uber-like shared services.

While history offers useful lessons, there is something different this time. With previous technologies, one could measure and forecast growth in terms of, for example, road miles, air hours, viewing hours, and so on. There are clear limits to the demand associated with all such quantities, dictated by the number of hours in a year and number of people using the service.

But data, the resource that AI refines and curates, is unique because it can be created by what we choose to measure and what we want to understand or control. For all practical purposes, those demands are infinite. We will always want more granular data about everything, from the operations of machines and processes to every feature of nature and our environment, including ourselves. And we’ll always want deeper, more granular, more useful analyses of that data.

Thus, rather than peak, the deployment of AI will accelerate because of a two-factor phenomenon. The efficacy and efficiency of the underlying AI software is improving faster than it did for conventional computing half a century ago. And the deployment of AI is accelerated by building on the already-existing cloud.

Before AI came on the scene, the United States already had thousands of data centers to provide the vast array of digital services everyone now takes for granted. The cloud is on track to be the largest infrastructure ever built. That is not hyperbole, but measurable in dollars, material tons, network route miles, and by the energy that will be needed to build and animate it all.

You can take two things to the bank.

First, AI is an economic accelerant. If AI is successful in merely restoring U.S. productivity growth to its previous average, simple arithmetic shows that this would add a cumulative extra $10 trillion to the U.S. GDP over the coming decade. Many believe the productivity boost will be bigger. And a bigger economy induces much more overall energy demand than the specific energy needs to power the data centers.

Second, natural gas will supply most of the electricity needed to power data centers in this first phase of the AI revolution. That’s not speculation; it’s already happening. But energy markets prefer fuel diversity. Odds are that in the near future, nuclear energy will emerge as the next dominant new form of energy addition. Solar (amplified by grid-scale batteries) will doubtless add useful quantities of energy on the margin. That’s already happening, too, and it epitomizes the reality of energy additions—not the ideology of an energy transition.