Godzilla comes for PJM
If you’re at all involved with the electric power industry, I highly recommend you join Energy Central. Full disclosure: It’s much better for an organization to join EC on behalf of its employees than for employees to join as individuals, since the cost for joining as an individual isn’t trivial. I am a writer for EC, since I post almost all the posts I write for Substack on EC, where they get almost equal numbers of readers. I’ve been posting on EC for about seven or eight years. I don’t get paid for doing it, but I do get free access to all the content produced by EC and its writers.
What I find most beneficial about EC is its newsletters, which over the past six or seven years have progressed from links to random news articles about the power industry to well informed original commentary on developments in the industry, backed up by links to news articles. Of course, one of the most important of those developments is the huge growth in projected demand by data centers that are either planned or under construction in the US. As we all know, this has led to growing opposition in many regions of the US, due to the projected (and actual) impact of the data centers on power bills and quality of life in those regions.
These confrontations are intensifying. While there have been some relatively small victories by opponents of the data centers (and the big AI companies whose insatiable thirst for computing power is behind the need for those data centers), they have often proven to be illusory. It’s now clear to me that what we have seen so far are just the opening skirmishes of a much larger war, which is beginning to look more and more like an irresistible force meeting an immovable object.
That’s why I was surprised to read the title of this past Thursday’s Energy Central daily news roundup newsletter: “A grid break-up?” Of course, I’m sure that apocalyptic language like this is used daily by other news sources, but given Energy Central’s reader base, this was startling. The email began:
American Electric Power warned it may ditch PJM and SPP as interconnection delays pile up.
The power giant says the grid operators aren’t moving fast enough to bring a flurry of new data centers online. AEP’s utilities now have 63 GW of contracted large load set to come online by 2030 (nearly 90% is tied to data centers). This announcement arrives less than a year after PA Gov. Josh Shapiro threatened to flee PJM.
AEP’s options: A) Leave the RTOs, B) stay put, or C) a mysterious third thing that CEO Bill Fehrman referred to as “alternative structures.” He’s especially concerned about getting new power online in PJM, which he thinks won’t clear things up anytime soon.
Meanwhile, PJM hears the frustration…and is weighing its next move. 🤔
The grid operator is mulling over a major market overhaul as hyperscaler demand outpaces supply and capacity prices skyrocket. Among the options: A) requiring long-term energy contracts B) letting states cap capacity costs…while accepting less reliability or C) shrinking the capacity market.
Here is my rewording of this:
1. Utilities that are part of American Electric Power – one of the nation’s largest electric utility holding companies – have promised to deliver 63 gigawatts (63,000 megawatts) of power to planned or under construction new data centers in the PJM and SPP footprints[i] by 2030.
2. For comparison purposes, the plant at Grand Coulee Dam in North Dakota, the largest (by nameplate rating) power plant in the US, can generate about 7.1 GW. Thus, it would require almost exactly nine Grand Coulee Dams to power those new data centers. Moreover, since the average natural gas-fired plant generates 92 MW annually, it would take about 685 gas plants to deliver that much power. Needless to say, these are astronomical numbers.
3. Another factoid: 63 GW is a little less than one seventh of average continuous consumption in the US. In other words, AEP has committed to provide new data centers with power equal to one seventh of current US consumption by 2030. When you add the commitments of all utilities in the country, this probably means that hundreds of gigawatts of new power plant capacity will be required by 2030. This is of the same order of magnitude as total US annual power consumption (around 500 GW).
4. Currently, about 86 GW of utility-scale generation is under construction in the US. The good news is this is a record. However, the bad news is that this is probably less than a quarter of the total new generation that will be required by 2030. This is required for utilities to deliver on their current commitments for new power, let alone the new commitments they’re making all the time.
It doesn’t seem likely that the US will be able to power all these new data centers. Something will have to give. What might that be? The article lists three options for PJM, but they apply more generally to the entire US grid:
A. Requiring long-term energy contracts. The problem with doing this is it makes it much harder for residential, commercial and industrial property developers to ramp up development to address unanticipated increases in demand.
B. Letting states cap capacity costs…while accepting less reliability. In other words, states could agree to have a less reliable power supply, so they could pay less money to power producers to overbuild production capacity as a buffer against greater-than-expected demand growth. Of course, this policy isn’t likely to be popular, since it will probably lead to future power shortages.
C. Shrinking the capacity market, in which power producers are paid to invest in new plants, even if it’s not certain the output of those plants will ever be required. Of course, doing this will also reduce costs, at the risk of future power shortages.
All three of these options make it likely that future economic growth will be curtailed. The question now becomes whether we let the free market determine which growth is curtailed. Since it’s likely that, at least in the next five to ten years, data centers used for AI purposes will be able to outbid other end users – hospitals, schools, heavy manufacturing, even residences – for scarce power, we may have to decide whether we’re willing to accept that situation, perhaps because we are convinced that curtailing power use for AI poses a greater risk to our economic security than curtailing use for hospitals.
But even if we decide that’s an acceptable trade-off, we should only do that if we also decide to impose a significant excise tax on AI use (just like we have excise taxes on tobacco, alcohol, some luxury goods, etc.), with the proceeds going to fund important goals like health care and education.
Tom Alrich’s Blog, too is a reader-supported publication. You can view new posts for two months after they come out by becoming a free subscriber. You can also access all my 1300 existing posts dating back to 2013, as well as support my work, by becoming a paid subscriber for $30 for one year (and if you feel so inclined, you can donate more than that or become a founding subscriber for $100). Whichever option you choose, please subscribe.
If you would like to comment on what you have read here, I would love to hear from you. Please comment in my chat or email me at tom@tomalrich.com.
[i] PJM and SPP are both regional transmission organizations (RTOs), which are responsible for operating the transmission grid, including collecting tariffs, in a designated area of the country. PJM is the world’s largest competitive wholesale electricity market. It serves over 65 million people across 13 states and Washington, D.C.