If you care about energy, or technology, the growth in electricity demand from data center operations is inescapable. If you care about both, it is elemental. AI-driven data center demand is a, if not the, main driver of electricity demand growth in many US electricity markets, and is one of the prime movers of global electricity demand growth this decade. The combined capital expenditures to develop AI from the four ‘hyperscalers’ (Alphabet, Amazon, Meta and Microsoft) will exceed $200 billion in 2024.
US electricity demand hit its all-time high in 2021, at more than 4,200 terawatt-hours. The US has never used more electricity than it does right now.
But, it also uses barely more electricity today than it did fifteen years ago. The previous record year for electricity consumption was all the way back in 2007. To put it another way, US electricity demand was post-growth for most of this century. Even 2022’s record power consumption was less than two percent higher than it was 15 years earlier.
Conditions are changing, though, and so are expectations about future demand. While the benefits of AI are unclear for some; for electric utilities, powering data centers offer the tantalizing prospect of increasing demand for the first time in decades. As we pointed out last week, utilities like the Rappahannock Electric Cooperative are seeing potential demand from data centers that far exceeds their current system load. And, many of these centers have ambitions for using clean energy to power their systems. Managing that demand was a hot topic (really, the hot topic) at last week’s Deploy 24 Conference, the US Department of Energy’s annual showcase of clean energy technologies supported by the Loan Programs Office and other department offices.
At Deploy24, companies from along the data center value chain shared their experience in developing past projects and thoughts on the future. Fun fact: the first 200 MW of a “fully chipped” data center could cost $10 billion, and plans are underway for 2 gigawatt facilities to be built in phases. These are enormous projects that can change the economics of a large area. So what did we learn from the panels?
First, utility planning groups face a difficult challenge. While data center compute needs may be absolute, their locations are not, and data center developers often move their early-stage plans from one market to the next as changing rate structures and other siting complications change the calculus. Even worse, those developers often come back several years later with an even bigger project and tell the utility, “no, this time for real.” Utilities then need to aggregate all of these projects and get approval from their state public utility commissions (PUCs) to procure sufficient power, often through an Integrate Resource Plan (IRP) or similar proceeding.
Second, reforms are underway that could shift the balance more in favor of utilities. Several PUCs are considering measures like making data centers sign offtake agreements or underwrite additional transmission. In Ohio for instance, a new rate structure would differentiate between data centers based on size and could impose an “exit fee” on projects that are canceled.
Hopefully some of the AI magic driving this surge in electricity demand can also help those working to supply it. For instance, the IRPs we mentioned earlier are often more than 500 pages long and could contain multiple appendices (NV Energy’s had 29 different documents totaling thousands of pages). Halcyon has aggregated these unwieldy proceedings into a single location and organized them such that analysts can quickly find and evaluate current and past IRPs to identify trends in data centers (or really any topic you can think of).
Here’s an example of IRP analysis — each of the cells in this sheet hyperlinks directly back to the source material for easy verification.
And here’s a view into how we’ve organized these proceedings and documents. You can easily search to find IRPs, dockets, and other document types, and then either read them directly or query against them:
To stay current with the latest changes from the PUCs, Halcyon launched daily email alerts that capture all the new data we’ve added to our platform in the last day. Alerts add important context and organization to the high velocity stream of information produced by the PUCS and the broader energy industry. Read more about our intentions with Alerts and our current capabilities here.
Together, you can start to see how all these solutions come together holistically to save analysts time and improve productivity. Alerts proactively push novel information directly to interested users, they can click through to either read the source document directly, or into a platform where they can query against that document (along with others, either in tandem or isolation) and quickly verify their analysis via inline citations that directly link to source material.
In an industry that has its fair share of junkets, Deploy24 stood out as an informative, entertaining, and ultimately very productive event. While many energy and clean tech conferences largely focus on supply, the Electricity Demand growth track was a welcome addition that was right up our alley. And while I said there was representation from across the value chain, there could be more representation from offtakers like Google, Meta, and the other tech giants that are central to these demand growth conversations.
We will definitely be back at Deploy25 next year - hopefully in a position to make a bit more of a splash. We look forward to seeing you there.
Comments or questions? We’d love to hear from you - sayhi@halcyon.eco, or find us on LinkedIn and Twitter