“‘Nother ten minutes, no longer. And then I’m turning around, ‘round.” -Led Zeppelin, from Fool in the Rain
In a testament to staying relevant, this isn’t our first Fool in the Rain reference within Alt Blend. But relevance comes naturally when you’re talking about one of the coolest drumbeats ever, by the greatest rock drummer of all time, from the “Best Rock Band of All Time.” Maybe I’m biased. 😊
To the point of today’s quote, this whole Iran War situation was supposed to have been “short-lived,” just like Robert Plant’s commitment to waiting for his no-show date. Yet, here we are, at least 3 months later, with new escalations amidst continued oil supply disruptions. Looking past the current chaos, today I’m setting out with the goal of bolstering our collective energy optimism, guided by Zeppelin’s lyrics of this iconic song. Here we go!
Fun fact: “Bahrain was the first place on the Arabian side of the Persian Gulf where oil was discovered,” and it has been producing oil since 1932. Oil is a major export for the island nation. And, while the Strait of Hormuz wasn’t part of the common American lexicon just 6 months ago, this now infamous body of water has become a center of global attention, along with becoming a massive oil-transport chokepoint for countries including Bahrain, Saudi Arabia, the UAE, Kuwait, Qatar, and Iraq (as well as Iran).
Why don’t you show up and make it alright?
Now I’m of the opinion that the world needs oil (because it does), and I’m also sure that this Iranian conflict will end at some unknown point in the future; it may have even ended between the time I’m typing this and the time you receive it! Regardless, because human ingenuity is unstoppable, oil logistics are “challenging,” and we may require a wee bit more energy for AI/data-centers in the coming years, I’m betting that people are working on pragmatic future energy solutions, instead of simply crossing their fingers and wishing the Strait soon reopens (which wouldn’t remotely solve the AI energy need, anyway).
Deloitte forecasts that global data center electrical demand could be in excess of 1000 TWh by 2030 – roughly double what it is today. McKinsey arrives at a lower estimate (606 TWh), but these numbers put the US data center usage somewhere in the mid-teens of total US power demand. And let’s not forget that our ex-data-center’s need for electricity has already been increasing because, well, more things run on electricity these days. Not only do we have more electrical devices, but some things that used to run exclusively on oil are now electric (e.g., cars, lawnmowers, etc.). Goldman Sachs anticipates total power consumption will increase by 160%+ by 2030, so that’s really what we need to solve for. I was going to joke that soon my t-shirts and baseball hats will need electricity, but then I found these electrical clothes!
As of now, oil fuels transportation, petrochemicals underpin manufacturing, and natural gas plays a central role in power generation across large portions of the globe. At the same time, we’re clearly in the middle of an energy evolution: renewable energy, battery technology, the grid, and new nuclear options are improving. Thus, to me, the question boils down to two segments/phases:
- How meaningfully can non-fossil energy sources offset the expected increased power consumption of AI, and
- When will alternative energy reduce the need for oil, gas, and coal utilized to solve today’s baseline energy needs (i.e., not only offsetting future energy inflation).
Well, there’s a light in your eye that keeps shining
Supporting an optimistic view, there are a variety of solutions moving the ball forward to alleviate some of the power-consumption issues we face. One that could be a total game-changer is greater efficiency. As this article points out (worth the read for more details on this topic), the efficiency of data centers is measured by Power Usage Effectiveness (PUE), with perfect efficiency being 1.0 (meaning all power is used for computing rather than for cooling, lighting, and other power needs).
“This dramatic improvement results from multiple innovations: free cooling, hot/cold aisle containment, optimized airflow management, high-efficiency power distribution, and advanced building designs that minimize cooling requirements.” Just since the early 2000s, PUE has dropped from 2.0 to typically 1.5-1.6 today, but Microsoft and Google are already hitting marks around 1.10. Servers are more efficient, cooling has been reduced by 60-70%, and Water Usage Effectiveness (WUE) has improved 85-95%. If we keep improving efficiency across the board (devices, equipment, heating/cooling of homes, etc.), it can go a long way to solving this whole pending predicament.
I’ve got no reason to doubt you, baby
Expected to play a major role in addressing my first question, IEA expects renewables to have annualized growth of “22% between 2024 and 2030” and meet nearly 50% of the electricity demand growth of data centers. There is some reason to doubt these solutions, and that comes down to energy storage. Wind and solar resources are intermittent, which is a mismatch for the 24/7 needs of data centers. Even so, they could meet about 80% of a data center’s needs when paired with storage and baseload generation (such as nuclear or natural gas) to smooth out round-the-clock demand.
Whatever happened to you?
Whether the science is finally sinking in for people, we’re far enough removed from historic nuclear disasters (off the top of my head: 3-mile Island, Chernobyl, Fukushima), or the power need just seems so overwhelming that it’s politically palatable to pursue all options – or all of the above – nuclear is finally being embraced again. But bringing it online is slow, so it will provide only about 10% of the needed capacity by 2030. I think there are several reasons for longer-term optimism, here, with developments like:
- This PSU-Westinghouse partnership (as an engineering major, I got to tour the referenced Penn State Breazeale Reactor in the early 2000s, which was cool), and Microsoft reopening Three Mile Island.
- Small modular reactors (SMRs) or microreactors that can address local needs more quickly
- Advanced fission technology (sodium-cooled, high-temperature gas, molten salt)
- Nuclear fusion: the “holy grail” promise of incredible power generation, but currently estimated to be 20-30 years out from being a real contributor.
And you thought it was only in movies
Also from PSU, if you want something that sounds futuristic and may be very relevant regarding the aforementioned storage issue, how about “turning renewable energy into a storable fuel?” The TLDR is that they have created technology that “efficiently converts carbon dioxide and renewable electricity into methane – the primary component of natural gas,” which can then be stored and transported!
Just a fool waiting on the wrong block
At the moment, we have many power-generation challenges ahead of us, including the power grid itself, which we covered about a year ago. But maybe someday in the not-too-distant future, we’ll look back and realize we were making a mountain out of a molehill. Don’t get me wrong – the current state of worry itself may be a necessary factor in motivating mankind to collectively pursue solutions.
Like Robert Plant freaking out that his date was a no-show, while the whole time he was on the wrong street corner, perhaps we just don’t yet have the right perspective to uncover the easy answer to this riddle. As above, there are game-changers that can quickly make this situation less problematic – such as improved efficiency and storage solutions. And has anyone considered the ultimate irony of AI improving all the solutions we are pursuing, or solving its own power problem altogether, effectively rendering “Fuel in Bahrain” obsolete?
Until next time, this is the end of alt.Blend.
Thanks for reading,
Steve
