End-of-year wraps often recap what happened over the last 365 days and make predictions about what will happen over the next 365. In our world at the intersection of energy and AI, we’ve seen enough to know that even the most well-informed predictions can fall short — and, much like a bad email, the world doesn’t need another “what we learned” post. Instead, for our concluding edition of Machine Readable, we’re going to anchor on the state of play in energy right now. 

Why? Even those well-embedded in the U.S. energy industry should be reminded just how big one element of it is: the electricity system.  

Just the hundred or so investor-owned utilities in the United States allocate $200 billion dollars a year in capital expenditures. Their relatives — electric cooperatives, municipal utilities, and various Federal power authorities and administrations — invest tens of billions of dollars more. In nominal terms, investor-owned utility capex has doubled in just over a decade. 

Source: EEI

While retail sales of electricity in the U.S. are (still) flat in energy terms for almost two decades, utility revenue is not. Last year, investor-owned utilities sold more than $400 billion of electricity. 

Total sales, including other electricity retailers, are higher still. In the past 12 months, US rate payers have spent more than $500 billion dollars on electricity. The collective U.S. power bill has increased by more than $100 billion in just the past three years. 

U.S. investor-owned utilities were worth a collective $890 billion at the end of last year, and the one-year return of the major US utilities index is up more than 20% since December of last year. 

And while U.S. utility valuations approach a trillion dollars, the value of their assets is even higher: more than $2.1 trillion at the end of 2023. U.S. electricity is a big sector, with big players, growing revenue, and a growing demand for its services as demand from AI and manufacturing comes to outweigh the industry’s multi-decade improvements in efficiency. 

There’s another important number here too, one which I think is often overlooked when considering the sector — its human footprint is significant as well. U.S. utilities employ almost 600,000 people, and total headcount has increased by almost 50,000 since 2020. Most of those employees work in the physical business of providing utility services, but 120,000 US utility jobs are managerial or supervisory.

Source: BLS

We don’t often think of the US electricity system in this aggregated and integrated way, but it’s worth doing so. To wrap these figures all together: almost 600,000 people invest $200 billion a year, receive $500 billion in revenue, and operate a $2 trillion machine that provides the essential energy input of our built environment, our technology system, and increasingly, our transportation and industry systems as well.  

This big system also needs to keep getting bigger. It needs to grow so that it can accommodate another century of volume growth, expansion in scope, innovation in business model and technology, and new priorities such as system resilience and decarbonization. It needs to do so safely, equitably, efficiently, and quickly — a daunting set of essential priorities. 

If you are reading this note, you are probably hard at work on one of these priorities yourself! Halcyon is focused on two of those priorities: efficiency and speed. Embodying those priorities also requires a greater understanding of customer needs. 

One important learning from our last year of building is that AI interfaces, while often magical, do not always lend themselves to the managerial and strategy workflows of those 120,000 utility employees I mentioned above. We find that extracting numbers from words, to put it plainly, is a necessary but not always sufficient part of the workflow. An LLM and the capability to query relevant data is a fantastic, but not necessarily scalable, workflow for analysts who need to know not one thing about one thing, so to speak, but the same thing about everything.  Accomplishing this means  combining machine learning and data science, not just one or the other. 

Another learning — and one that we are delighted to learn! — is that AI is not a replacement for teams and strategy.  It is, rather, an amplification of teams and strategy. Applied smartly, it is a multiplier of what companies do best, and it allows them to do more of it. Human-in-the-loop, as the saying goes, is a feature of the best AI technology applied to business strategy challenges. AI is a technology, but it is also a service, and more importantly it is in the service of essential work of organizations busy building the future. 

Next year, we will build more, and show more of what we are building. Thank you to those already building the future with us, and thanks in advance for those interested to do so. Let us know how we can build with you, and for you, in the service of extending and improving the $200 billion / $500 billion / $2 trillion / 600,000-person machine that is the US electricity system.  

And finally, thank you for reading and following our work this year. We’re now shipping alerts and are grateful for the engagement so far. Please keep sharing, and keep the feedback coming! We’ll be back in 2025.

Comments or questions? We’d love to hear from you - sayhi@halcyon.eco, or find us on LinkedIn and Twitter