Will Data Centers Drive Energy Demand, or a Generation Bubble?

Or both?

Challenging times for electricity producers.
The move toward electrifying everything, including transportation and building HVAC, in addition to the rise of AI and massive Data centers, and onshoring of US manufacturing for everything from EV batteries to computer chips, to solar panels, has caused load forecasts to jump, jeopardizing some climate goals.
Some increased loads seem inevitable, but worth remembering the lessons of history.
We’ve seen several periods when massive load growth seemed inevitable, but failed to materialize – due in large part to total neglect of the potential for energy efficiency.

Missed this piece when it first came out in the Wall Street Journal.

Wall Street Journal:

The last time independent power producers were this excited about an electricity-boom cycle was in the late 1990s and early 2000s, when demand-growth expectations were fueled in part by the growth of Silicon Valley. It didn’t end well. 

In a piece published on Forbes.com in 1999, Peter Huber and Mark Mills wrote: “Southern California Edison, meet Amazon.com. Somewhere in America, a lump of coal is burned every time a book is ordered online.” The two authors, who co-wrote books about energy, including “The Bottomless Well,” estimated that one billion PCs on the Web would represent electrical demand equal to the total power capacity of the U.S. at that time. The piece drew much attention and pushback, including from scientists at the Lawrence Berkeley National Laboratory, who said the authors were overstating the impact.

In a 2001 PBS interview, Peter Cartwright, who was then chief executive of Calpine, said, “Silicon Valley, as everybody knows, is well aware—is one of the fastest-growing demand centers in the state, and we have very—hardly any power generated in this area.” Calpine was one of the most aggressive independent power producers of that era. Its installed base grew at a compound annual growth rate of 63% between 1998 and 2002 through both new construction and acquisitions. In addition to demand growth, Calpine and other developers believed that their new gas-fired power plants would come to replace older, less-efficient generators that were built by monopolistic utilities rather than competitive developers.

Similar to today, stocks of independent power producers at the time fetched much higher multiples of expected earnings than regulated utility peers, according to Chris Seiple, vice chairman at Wood Mackenzie. Seiple was a power-industry consultant at the time and authored a report warning that the power-plant construction frenzy could result in a bust. He turned out to be right: Major developers went bankrupt in the aftermath, including Calpine and NRG Energy.

Of course, there were other factors driving the power-plant boom at the time. For one, that was an era when many states—including California—started deregulating power generation, allowing independent developers to build large-scale power projects in a competitive market.

Hugh Wynne, co-head of utilities and renewable-energy research at SSR, worked at a power-project development company from the mid-1990s to 2001. He said developers at the time assumed that the long-term power price would match the long-run marginal cost of new power-plant capacity, covering both variable operating costs as well as capital costs. In fact, Wynne said, after the aggressive build-out resulted in a surplus of capacity in many markets, power prices fell sharply, covering only the operating costs of these power plants—not enough for the companies to pay off the debt they raised to build them. 

Could the current market face similar problems? Likely not in quite the same way. For one thing, the drivers of electricity-demand growth are more tangible this time around. Tech companies are spending real money on building out AI infrastructure, and the Chips Act contains clear incentives for nearshoring chip manufacturing. At the same time, energy demand from transportation and industrial applications—including fracking equipment—is steadily shifting to electricity.

Secondly, the industry today is more familiar with how competitive power markets work. Independent power producers and the financiers backing their projects were burned enough by the last gold rush that they aren’t likely to invest in new power plants without some kind of long-term power-purchase agreement, Wynne said. Always-available power is important for data centers, and tech companies have been willing to sign long-term contracts at premium prices.

While power demand looks set to inevitably grow, the magnitude and timing are still an open question. For instance, the utility Arizona Public Service in its latest planning document forecast that its demand from high-power users—such as data centers—could range anywhere from less than 100 MW to more than 1,000 MW in the coming years.

Notably, utilities’ long-term demand forecasts have changed a lot over the past few years and are likely to keep being revised. In early 2021, utilities in aggregate expected load to grow 8.2% by 2035 compared with 2021 levels, according to analysis by the Rocky Mountain Institute. As of June 2024, the expected growth was 23.9%.

So while an outright bust looks unlikely, there is still the risk that independent power producers’ stocks get ahead of themselves

Grid Strategies:

Wood Mackenzie sampled several utilities’ announced data center demand, estimating 93 GW total through 2030. Based on bottom-up analysis of tech industry capacity, Wood Mackenzie forecasts a more realistic national five-year AI buildout to be just 23 GW.

Data center developers may have such a large appetite for growth that their projects could use up any and all currently-unused grid capacity over the next five to ten years.
On the other hand, business revenues to cover the costs of the artificial intelligence investments have not yet been proven.

This combination of exuberance and uncertainty raises the question of whether these projects could fail to sustain anticipated power demand. Such a failure could leave other customers with responsibility to cover costs for transmission and, in some regions, generation investments. There are economic and energy reliability risks, both from under- and over-investment.

—

With this as background, worth remembering that wind and solar, backed by batteries, are the quickest, cheapest way to add capacity to the grid.
In addition, more distributed energy resources mean that utilities and power producers do not have an overwhelming number of eggs in one basket, like, say a Vogtle-style nuclear money pit.
Solar and wind projects in the 50 to 200 MW range, once permitted, can go up in 18 months or so, come on line quickly without adding burdensome long term interest costs, and allowing planners to flex as demand projections become clear.