I know I just wrote about flaws in the arguments against wind and solar power in my last post, but I recently saw a new attack so wrong I really need to say something about it. The attack appeared in a Substack called David Blackmon’s Energy Additions Substack. Blackmon knows a lot about the oil and gas industry, and since I don’t know much about the oil industry, I find his articles on those subjects to be interesting. However, when Blackmon drives outside his lane, which he does on occasion, he frequently gets into a wreck. Such is the case with his August 27, 2026 article, X Post: Ross Fires Torch Massive Solar Array. The story is about how a solar power array in Texas was destroyed by a wildfire. But in the lead-in to this story, Blackman took a gratuitous kick at wind and solar generation facilities: “We all know that wind and solar are weather-dependent generation sources which will always, without exception fail us when we need the electricity the most.”

The example Blackmon gives to support his claim was the failure of wind and solar to produce much electricity in Texas during Winter Storm Uri in 2021. This is ironic, because as the official University of Texas report on Uri observed, during the storm, “[a]ll types of generation technologies failed.”^[1] Although it is true that little electricity was generated by wind or solar facilities, it also is the case that about 25,000 megawatts of natural gas-fired generation capacity dropped off-line during Uri just when it was needed most.^[2] The failure of natural gas-fired generation peaked on the night of February 15, when rolling blackouts were initiated and then steadily increased as large amounts of natural-gas fired generation capacity was forced offline by the storm.^[3] The real lesson taught by Uri has nothing to do with wind and solar generators. Rather, it is that the Texas wholesale market run by ERCOT (the Regional Transmission Organization that operates the Texas electric system) failed to provide the proper incentives for Texas generation owners to take the steps necessary to winterize their facilities so they can operate during bad winter storms.

And if you were to look at actual data instead of calling people names, you would see that solar power is likely to be better able than fossil-fired generators to supply electricity during the hot summer days when demand for electricity skyrockets and generation capacity is needed most. The sun’s rays are the strongest on those days, which means solar panels typically produce electricity at their highest output. This makes solar power well-suited to help supply the high demand that inevitably peaks when temperatures rise in the summer.

By contrast, fossil generators cannot produce at their maximum capacity on hot summer days because they cannot operate as efficiently during temperature extremes. Moreover, fossil generators are prone to suffer from forced outages when operated full out for several hours in the heat. The result was reported in a recent post in Kilovar’s Substack (one of my recommended Substacks). Kilovar presented an analysis of outage data showing that, in the summer of 2026, forced outages left the Midcontinent Independent System Operator (MISO) with as little as 56 percent of its total generation capacity in operation, and PJM with as little as 72 percent.^[4] MISO and PJM were forced to operate close to the edge several times in the summer, and would have been hard pressed to keep the lights on without electricity produced by solar panels and wind turbines.

Don’t get me wrong. I have written several times about the limitations on solar and wind’s ability to provide reliable service. For example, one big problem with solar in the summer is that, even though it can produce quite a bit of electricity in the middle of the day, the production of electricity by solar panels declines in the later afternoon/evening hours more quickly than the demand for electricity. If an electric system does not have enough fossil or other conventional generation capacity (including batteries) available, the rapid decline in solar output can lead to rolling blackouts. This is exactly what happened in California in the summer of 2020, as I described in my post, Whose Fault? Further, as I have also written, solar and wind are not as helpful in meeting peak demands on cold winter days, when the highest demand is in the early morning and late afternoon/evening.

I also want to be clear that I am not addressing whether the benefits provided by solar and wind are worth the costs. This is a far more complicated subject than I can address in this post (or any post for that matter). But the fact is that there currently is about 310,000 megawatts of solar and wind capacity^[5] operating in the US, producing about 20 percent of all electricity generated by the grid.^[6] And I think this capacity very clearly does provide the grid with electricity during the summer peaks when demand is at its highest and every megawatt-hour of electricity is needed.

This has been one of my shorter posts, coming as it does immediately after a longer post addressing similar arguments against solar and wind. Let me make one request of those readers who don’t think very highly of those technologies. Please don’t submit comments making general attacks on solar and wind. I have seen those before, and that is not what this post is about. Instead, please consider whether you think our existing solar and wind capacity is providing electricity that is useful in satisfying the peak demand for electricity on the grid. I would be interested in hearing what you think.

***

I hope you enjoyed this post. I enjoy writing them and will never charge for subscriptions or ask for donations. I only ask that, if you did enjoy it, you press the “like” button below. And please consider sharing my posts with other people you think might be interested. Doing so will help me evaluate interest in the book I am writing on grid operations. Of course, if you have a reaction to, or question about, this post, please leave a comment and I will be happy to respond.

^[1] University of Texas Energy Institute, The Timeline and Events of the February 2021 Texas Electric Grid Blackouts, at 8 (July 2021), UTAustin (2021) EventsFebruary2021TexasBlackout 20210714.pdf.

^[2] Id. at 33.

^[3] Id. at 27.

^[4]Kilovar’s Substack, Equipment Outages: The Hidden Drag on Reliability, (July 19, 2026), https://kilovar1959.substack.com/p/equipment-outages-the-hidden-drag

^[5] EIA, Electric Power Monthly February 2026, Table 6.1. Electric Generating Summer Capacity Changes.

^[6] EIA, Electric Power Monthly February, 2026, Tables 1.1 Net Generation by Energy Source, and 1.1.A. Net Generation from Renewable Sources.

On March 21, I wrote a post (What is the Real Problem With Renewable Solar and Wind Generation?) in which I explained how advocates of solar and wind generation who have pushed for replacing all fossil generators fail to recognize or account for the operational limitations of solar and wind technologies. The inevitable result of quickly removing all fossil units from the grid would be darkness and chaos. Today, I want to turn the tables a bit and write about fallacies in arguments made by opponents of solar and wind generation.

Failure to Recognize the Major Role Already Being Played by Solar and Wind

It was only a few years ago that solar and wind generation first arrived on the grid. At that time, it may have been appropriate to question the wisdom of deploying those technologies on a large scale. Today, however, the total amount of operational solar and wind generation capacity is about 310,000 megawatts,^[1] and this capacity generated about 20 percent of all electricity produced in the US in 2025.^[2] Despite this fact, I continue to see repeated strident attacks against the incorporation of solar and wind into the grid, claiming that these technologies provide no value and in fact are harmful to grid reliability.

Whatever the merit, if any, of these arguments, that horse has left the barn. Instead, we need to be talking about how to best integrate today’s considerable operating solar and wind capacity into the grid and the extent to which new wind and solar capacity should and could be added in the future.

Failure to Recognize the Benefits Provided by Solar and Wind

I don’t mean to suggest that questions about the operational and other problems associated with solar and wind generation are irrelevant to the grid’s future. The operational issues raised by these technologies are, of course, relevant to where the grid should go from here. But in order to consider wind and solar’s future role, it is necessary to consider both their benefits and their negative attributes. And most of the discussion about solar and wind facilities I see from their opponents focuses solely on the technologies’ problems with little or no attention paid to their benefits.

For example, many opponents dismiss climate change concerns—the primary driver for most solar and wind development—as being nothing more than the ranting of wild-eyed crazy environmentalists and a pretext for money grabs by greedy profiteers. That being the case, they see no reason to consider any benefit accruing from emission-free generation of electricity. My long-time readers know that I stay away from climate change issues in this Substack. That is because climate change is a complex issue outside my area of expertise, and I can add nothing of value to the discussion. But it strikes me that a discussion of the potential benefits of reducing carbon emissions through the use of solar and wind generation capacity deserves something more than just name calling.

Putting aside the question of carbon emissions, which I admit can be controversial, I have never seen any acknowledgement of a second benefit I have frequently written about. Once wind and solar facilities have been constructed, they require no fuel to operate, which leads to significant cost savings. US fossil-fired generators spent on average 72 billion dollars a year on fuel over the last five years.^[3] And the use of solar and wind generators in 2025 instead of fossil generators saved the grid about 22 billion dollars in fuel costs.^[4] These savings are significant and cannot be so easily explained away.

It would of course be legitimate to argue that the operational problems caused by solar and wind facilities could more than offset their benefits. But in order to make this argument, you first have to acknowledge the benefits.

Failure to Consider Operational Issues Associated With Conventional Generation Technologies

Numerous opponents of solar and wind point to the intermittent nature of those technologies to assert they are unreliable when compared to conventional generation facilities. This argument is based on the assumption, sometimes acknowledged and sometimes implicit, that conventional generation facilities can operate 24/7 and never break down. But all generation facilities must be regularly removed from service for routine maintenance. And all generation facilities suffer from unscheduled outages or degraded capacities during extreme weather conditions when they are needed most. This is what happened in the California rolling blackouts in the summer of 2020 and the more extended blackouts in Texas during Winter Storm Uri in 2021. A recent post in the excellent Kilovar’s Substack (one of my recommended Substacks) shows that, in the summer of 2026, forced outages left the Midcontinent Independent System Operator with as little as 56 percent of its total generation capacity, and PJM with as little as 72 percent.^[5]

Of course, solar and wind facilities have lower capacity factors than conventional generators, but the difference is one of degree; conventional generators are not immune from the possibility of being unavailable when needed. It is precisely because no generation technology is completely reliable that electric systems have capacity reserve margins. It is misleading to argue that solar and wind require backup generation capacity without acknowledging that all other generation technologies do as well.

Failure to Acknowledge that the Grid Can Readily Absorb a Sizeable Amount of Solar and Wind Capacity at Little Expense or Effect on Reliability

Another argument I frequently see is that it is expensive to incorporate solar and wind capacity into the grid because these technologies either need a substantial amount of overbuilding of capacity or else need to be backed up by fossil generation in order to provide reliable service. I have seen several cost estimates of what would be required, and those estimated costs are prohibitive.

The problem with these very high cost estimates is that they assume a hypothetical electric system with no existing generation capacity. But in the real world, electric systems do have existing operating capacity, and all electric systems also acquire significant amounts of reserve capacity. Systems with existing generation capacity have the ability to integrate some amount of solar and wind facilities into their system without affecting reliability. As an extreme example, an electric system with 30,000 megawatts of conventional capacity could easily add 100 megawatts of solar or wind capacity with no added cost and no effect on reliability. In fact, that system likely could accommodate significantly more solar or wind capacity with no adverse effects, the exact amount depending on the system’s configuration and a number of other factors. There are, of course, limits as to how much solar and wind can be added before system reliability is affected and new costs must be incurred. But I have seen no evidence that electric systems are incurring costs to integrate solar and wind at the levels shown in these calculations.

Failure to Acknowledge that the Nuclear Renaissance Has Not Yet Arrived and May Never Do So

If you read the other posts in my Substack, you will see I am an advocate of nuclear power as representing the best approach for achieving the twin goals of reducing the grid’s carbon emissions and improving grid reliability in the future. And I am not the only one. A growing number of people are advocating nuclear power as the best generation technology for the future. But I am surprised that many nuclear power advocates are also vocal opponents of solar and wind generation. After all, all three of these generation technologies can play an important role in reducing carbon emissions. They should be able to play together nicely. But for whatever reason, they don’t. In addition to the other arguments discussed in this post, there are two types of arguments against solar and wind I frequently see from nuclear supporters.

The first comes in the form of complaints about nuclear power’s treatment over the last 30-40 years, during which time the industry’s growth came to an almost complete standstill. If only we had used the money spent on solar and wind to construct new nuclear reactors instead, the argument goes, today we could have a large nuclear fleet providing carbon-free electricity without the need for any renewable generation. I love history and find “what if” scenarios (such as what if Lincoln hadn’t been assassinated, or what if England and France had stood up to Hitler at Munich) to be very interesting. And I agree it would have been a good thing if the US had not stopped constructing nuclear power plants. But I also find this argument to be irrelevant to the question of the benefits of solar and wind capacity. Whether or not it was a good decision, the US did almost completely stop constructing new nuclear power plants, and we have to deal with the consequences of that fact today.

The second argument asserts that nuclear power represents a much superior technology than solar and wind for producing emission-free electricity. Nuclear power plants can produce large amounts of electricity on a 24/7 basis with very little down time and represent a much more reliable generation technology than solar and wind.

Again, I tend agree, but again I find this argument to be not that helpful to the issues faced by the grid today. Yes, there is a growing momentum towards reviving the nuclear power industry. But just as we can’t avoid the fact that almost no new US nuclear reactors have been constructed in the past few decades, we also can’t avoid the fact that the often-proclaimed nuclear renaissance has not yet begun and may never occur. In a best-case scenario, it likely will be 10 years or more before significant new nuclear capacity can come online. By all means, let’s pursue the development of new nuclear power plants. But until those plants are operational, we have to figure out how to keep the grid running reliably without new nuclear power plants. The fact that nuclear power potentially could do this better in the future says nothing about the role solar and wind should play today.

Failure to Compare the Costs of New Wind and Solar With the Costs of New Conventional Generation Capacity

One area of running dispute between advocates for and opponents against solar and wind is the question of the relative costs of different generation technologies. This is a complicated subject, and there are a number of different ways to look at costs, each with its own advantages and disadvantages. I wrote about these different approaches in my March 23, 2025 post, Measuring the Costs of Generation, and will not repeat that discussion here.

One of these approaches is called the “Levelized Cost of Energy,” (which in the acronym-mad world of the electric industry has been reduced to “LCOE”). Advocates of solar and wind like this measure, because it tends to show that new solar and wind are cheaper than new fossil and nuclear generation. To counter this conclusion, opponents like to perform modified LCOE calculations, which compare the cost of new solar and wind capacity with the cost of existing fossil and nuclear facilities that are mostly or fully depreciated.

This is an apples to oranges type of comparison. A big part of the LCOE calculation includes capital costs, and it therefore is not surprising that new solar and wind comes out as being more expensive than existing fossil. That is because generation facility book values subtract depreciation from the original capacity costs, and the book values of fully or mostly depreciated facilities therefore show little or no capital costs. It’s like saying that my 10-year old Mercedes (if I had one) is less expensive than your brand new Honda because the book value of my Mercedes is lower.

I suppose cost comparisons between new and existing generation facilities could be appropriate if the question is whether an existing fossil unit should be retired and replaced by new solar or wind facilities. But these comparisons aren’t very helpful for considering the appropriate generation mix for the grid’s future. And that is an important issue because the demand for electricity is projected to continue to increase for the next several years. This means the grid will need substantial amounts of new generation capacity in the future, not only to meet the increased demand, but also to replace old generation facilities that are no longer operable.

New fossil and nuclear generation facilities can be very expensive, costing in the billions of dollars. Comparing the costs of different generation technologies can be very complicated, and controversial, and beyond the scope of this post. But in the context of looking at the grid’s future, any discussion of the cost of new solar and wind makes sense only when compared to the cost of the new conventional generation capacity that otherwise would need to be constructed if the new solar and wind capacity is not.

***

At the end of the day, I think advocates for and opponents against solar and wind capacity suffer from a similar form of blindness. Each side sees the question of incorporating wind and solar capacity into the grid as being like a light switch. Flip it up and all fossil generation is replaced by solar and wind capacity. Flip it down and we get rid of all wind and solar capacity and have a grid dominated almost entirely by fossil, nuclear, and other conventional generation. Neither result would be optimal.

But flipping the switch completely up or down are not the only options. Our light switch also has a dimmer that allows us to choose a lighting result somewhere between totally on and totally off. We can choose a generation mix for the grid that includes renewable solar and wind, fossil, and nuclear generation. Over time, we can move the dimmer up or down as appropriate to achieve the optimal mix of generation. To me, it is much more interesting, and relevant, to discuss what this optimal mix should be and how it can be achieved rather than to debate whether there should be any solar and wind, or fossil generation at all.

***

I hope you enjoyed this post. I enjoy writing them and will never charge for subscriptions or ask for donations. I only ask that, if you did enjoy it, you press the “like” button below. And please consider sharing my posts with other people you think might be interested. Doing so will help me evaluate interest in the book I am writing on grid operations. Of course, if you have a reaction to, or question about, this post, please leave a comment and I will be happy to respond.

^[1]EIA, Electric Power Monthly December 2025, Table 6.1.

^[2] EIA, Electric Power Monthly February, 2026, Tables 1.1 and 1.1.A.

^[3] Data for 2025 was taken from EIA, Electric Power Monthly for February 2026, Table ES2.A, Summary Statistics: Receipts and Cost of Fossil Fuels for the Electric Power Industry by Sector, Physical Units, 2025 and 2024. Data for 2020-2024 was taken from EIA, Electric Power Annual (2025), Table 7.1, Receipts, Average Cost, and Quality of Fossil Fuels for the Electric Power Industry, 2014 through 2024.

^[4] This calculation assumes that the average fuel cost of the fossil plants replacing solar and wind in 2025 was the same as the average fuel cost of fuel used by the fossil units that actually operated in 2025. The calculation used the same data as in the calculation of the total fuel costs for 2025 cited in the preceding footnote, as well as data on the total amount of electricity generated by fossil, wind, and solar units from Electric Power Monthly for February 2026, Table ES1.B, Total Electric Power Industry Summary Statistics, Year-to-Date 2025 and 2024.

^[5]Kilovar’s Substack, Equipment Outages: The Hidden Drag on Reliability, (July 19, 2026), https://kilovar1959.substack.com/p/equipment-outages-the-hidden-drag

lAST December, the PJM Regional Transmission Organization (RTO) conducted its capacity auction for the delivery of capacity in 2027-2028 (the December auction). And as I explained in my January 18 post about that auction (PJM Failed to Acquire the Capacity it Needs to Ensure Reliability. What Does it Mean?), the results of that auction were unprecedented. For the first time in the 22-year history of PJM’s capacity auctions, PJM failed to acquire the full amount of generation capacity it had determined is needed to ensure the reliability of its electric system. PJM operates the largest electric system in the US and one of the largest systems in the world, so this result in the December auction was significant.^[1]

This past Tuesday, July 14, PJM published the results of its most recent capacity auction (the July auction). And guess what? It happened again. PJM once again failed to acquire the amount of generation capacity it determined is needed to ensure system reliability.^[2] In fact, the deficit between what PJM determined it needed to acquire, and what it actually acquired, was greater in the July auction than in last December’s auction. In December, the auction left PJM 6,516 megawatts of capacity less than it sought to acquire. In the July auction, the shortfall increased to 6,831 megawatts. And because the amount of capacity offered into the auction was short, the resulting price for capacity was high. For the third straight auction, the auction price was set at the FERC-approved price cap, which was $325 per megawatt-day.^[3]

I am pasting below a chart from PJM’s auction reports summarizing auction results for the last 11 years (Table 2), which is useful for understanding how these results compare with previous PJM auctions.

Here are my answers to some questions about the meaning of these results.

1. What Effect Does the July Auction Have on the Price of Electricity?

As you can see from the above chart, the $325 per megawatt-day auction price established in the July auction translates to total payments of $16.4 billion for capacity in all of the 13-state PJM region. This continues a series of four straight auctions with the highest prices in PJM’s history. Total capacity payments in the 13 PJM states in these four auctions range from $14.7 billion to $16.4 billion. These high prices followed a series of very low prices, with total payments ranging from $2.2 billion to $3.9 billion. It was the sudden increase of capacity prices from the very low levels to the very high levels that was largely responsible for the recent controversy during the 2025 New Jersey gubernatorial election about electricity price increases.

It’s important to keep in mind, however, that this price will not go into effect until the 2028-2029 “Delivery Year,” which will run from June 1, 2028 through May 31, 2029. The capacity price currently applicable in PJM is the $329.17 price applicable to the 2026/27 Delivery Year. Consequently, although the July auction price is high in an absolute sense, it should not cause an appreciable increase in electricity prices from the previous two delivery years.

PJM capacity costs are factored into the wholesale price of electricity. Those wholesale prices must be included in the retail prices for electricity paid by consumers, but the utilities selling to consumers at retail have considerable latitude to allocate wholesale costs unequally among different customer classes (residential, commercial, industrial, etc). And state public utility commissions have the ability to review and revise those cost allocations. Consequently, it is difficult to project exactly how the prices set in PJM’s July capacity auction will affect the retail rates paid by ultimate consumers. But I feel confident in saying that electricity prices will be higher today and through at least the 2028/2029 Deliver Year than they were prior to June of 2025.

2. How Did the Price Cap Affect the July Auction?

One of the longstanding disputes about the RTO wholesale electric markets has been over the use of price caps. The economic purists have claimed that price caps distort the signals sent by market prices and will fail to attract sufficient new generation capacity. However, regulators have found extremely high prices to be untenable. As a consequence, all RTO markets have implemented price caps.

For the most part, I have tended to side with those who question the use of price caps. In the case of PJM’s capacity auction, however, I come to the opposite conclusion. In Appendix A to its report on the July auction, PJM provided a chart showing hypothetical auction results if the current FERC-approved price cap had not been applied (Appendix A, Table 1). This chart shows that, without the cap, PJM would have acquired only about 77 megawatts of additional capacity, which is an increase of only 0.05 percent over the capacity acquired with the price caps. PJM calculated that this miniscule increase in capacity would have caused the auction price to rise to $554.72 per megawatt-day, an increase of over 70 percent. This price increase would have resulted in a $13.4 billion increase in the already highest ever total capacity payments in the PJM region ($16.4 billion). I believe that all but the most ardent economic purists would agree that $13 billion is too much to pay for an additional 77 megawatts (0.05 percent increase) in capacity.

3. Will PJM’s Failure to Acquire All the Capacity Needed in the July Auction Necessarily Result in Blackouts?

Although PJM again was unable to acquire all of the capacity it said it needed to ensure reliability, this does not mean PJM will inevitably be subjected to blackouts. As was the case in the December auction, PJM was able to acquire significantly more capacity than it anticipates will be necessary to serve its maximum demand for the year. Instead, the shortfall affects PJM’s reserve margin, which is the amount of capacity PJM acquires above its projected peak demand. The reserve margin allows PJM to supply the peak demand even if some capacity is unavailable due to problems with equipment or for needed maintenance, and/or if demand is higher than expected. Although the auction did not yield the 20 percent reserve margin PJM wanted, PJM still is left with a 14.4 percent reserve margin. This is a healthy margin, and close to PJM’s target reserve margin in many previous auctions. And even if a 20 percent margin is needed to meet PJM’s one-event-in-10 year reliability standard, there is only a 10 percent chance that once in 10 years circumstances will occur in the year in which PJM failed to acquire enough capacity to achieve a 20 percent reserve margin.

4. What Steps is PJM Taking to Make Up the Shortfall?

PJM recognized that it needed to take further action to increase its generation capacity in light of its failure to acquire all needed capacity in the July auction. In its press release announcing the results of the July auction, PJM detailed a number of steps it plans to take.

Perhaps most importantly, PJM stated that it “plans to seek FERC approval to hold a special ‘Backstop Procurement’ in September to help address the near-term shortfall in electricity supply.” This sounds like one of the steps identified in the January 15, 2026, “Statement of Principles Regarding PJM” entered into between the White House National Energy Dominance Council and the 13 governors of the PJM States.^[4] (I discussed this agreement in my February 13, 2026 post The Trump Administration and the Governors of the PJM States Have Reached an Agreement in Principle to Address PJM’s Capacity Deficit. Will it Work?). There is a potential difference, however. The Statement of Principles provided that the capacity commitments to be acquired in the Backstop auction should have a 15-year term, to facilitate financing (financing is more difficult with the one-year terms issued by PJM in its annual auctions). Nothing in PJM’s statement mentioned this longer 15-year term. I am curious to see what PJM will come up with when it files its proposal for approval by FERC.

Whatever form it takes, it seems extremely unlikely to me that such an auction can happen in September of 2026, as PJM states. PJM is required to go through a cumbersome process with its stakeholders (customers, regulators, investors, generation and transmission owners, etc.) before submitting its proposal. Although PJM initiated a stakeholder process after the issuance of the Statement of Principles this past January, it does not appear to me that PJM is prepared to immediately file a proposal at FERC, which it would have to do to have any chance of having the proposal approved in September. And even if a proposal were to be filed soon, FERC frequently takes several months before ruling on filings this complex and important. I could be wrong, I frequently am. But I think a Backstop auction is more likely to be conducted in 2027 than 2026. But even if I am correct about the timing, conducting a Backstop auction in 2027 to help make up the capacity shortfall resulting from the July auction (as well as the December auction) would be an important step.

In addition to committing to file for approval of a Backstop auction, PJM also mentioned several other steps it was taking to expedite the construction of new generation capacity or otherwise meeting the expected increased data center demand for electricity. This includes: (a) clearing the backlog of requests to interconnect with its electric system; (b) continued development of a “Connect and Manage” proposal that contemplates allowing data centers to connect to the grid in exchange for agreeing to “operate flexibly”(i.e. disconnect from the grid) when necessary to avoid other customers from being subjected to blackouts; (c) creating an expedited interconnection process for large state sponsored generation facilities; (d) taking steps to maximize the capacity and availability of existing generation facilities; and (e) moving forward on a long-term market reform. These are all good ideas, and PJM’s pursuit of them shows it is taking seriously the potential threat to system reliability caused by its capacity shortfall.

5. Have Higher Capacity Prices Resulted in Construction of New Generation Capacity?

One purpose of the RTO competitive wholesale markets, including PJM’s capacity auction, is to send price signals about the need for new generation capacity. Low prices mean that no new capacity is needed and perhaps existing capacity should be retired. High prices mean new capacity is needed. As we have seen, capacity prices in the three auctions prior to the July auction were substantially higher than capacity prices in all previous years. How, if at all, did those high auction prices affect the amount of capacity offered into the July auction? To help consider this question, I have pasted below Table 6 from PJM’s report, which shows the amount of generation capacity offered into the July auction and the three prior auctions.

A few numbers from this chart jump off the page. First, the total amount of capacity offered into the July auction (over 150,000 megawatts) increased by about 3,000 megawatts from the prior auctions. That’s not a large increase, but it is over twice as much as the projected increase in demand during the 2028-2029 Delivery Year, which PJM stated was 1,375 megawatts (Report page 4). On closer inspection, however, the reported size of this increase is a bit misleading.

For one thing, another number also jumps off the page. After holding steady in the prior three auctions, the amount of coal-fired generation capacity offered into the July auction decreased by 3,500 megawatts. PJM states that this decrease is the result of “planned deactivations.” (Report Page 12). In other words, the high capacity prices in the last three auctions did not prevent the owners of some coal plants from planning to deactivate their plants.

The third number that immediately caught my attention was the 5,550 megawatt increase in the amount of natural gas-fired generation capacity offered into the July auction. This increase came after the amount of natural gas-fired generation offered into the three prior capacity auctions had declined each year. At first blush, this increase looked like maybe the high capacity prices in PJM’s last three capacity auctions had incentivized the construction of a significant amount of new natural gas-fired capacity. But then I read PJM’s explanation for this number. (Report at Page 12). PJM attributed the increase not to new construction, but to three factors: (1) changes in the way PJM calculated the capacity provided by existing facilities; (2) participation in the auction by existing facilities that had been exempted from participating in prior auctions; and (3) the conversion of a coal plant to natural gas (this conversion might help to explain some of the reduction in coal-fired capacity into the auction). In other words, no significant new natural gas-fired facilities are being planned to go in service during the 2028-2029 Delivery Year.

There are minor increases in the amount of wind, solar, and battery capacity offered into the July auction. But these increases appear to follow the trend of constructing new renewable generation capacity rather than any response to the high PJM capacity prices. The other generation sources listed by PJM make relatively minor contributions to total system capacity levels. Consequently, minor changes in the amounts of capacity offered by these sources into the July auction do not appear to indicate one way or the other whether high PJM capacity prices have led to the construction of material amounts of new generation capacity.

I didn’t know how this analysis would turn out before I started it. I was hoping it would show increased construction of new capacity in response to PJM’s high capacity prices, since that is what PJM needs to eliminate the shortfalls of capacity acquired in its annual auctions. But that does not appear to have happened yet. Perhaps it supports my thesis that PJM’s approach (and that of other RTO capacity auctions) of acquiring capacity in one year or less increments does not support the financing of significant amounts of new capacity no matter how high the price. Or perhaps a longer period of high prices is needed to incent new construction. Perhaps time will tell.

***

Well, despite my best efforts, including leaving a lot of words on the cutting room floor, this ended up not being a short post. I’m reminded of an old lawyer’s joke (it’s a joke I have heard many times over the years, not a joke told by an old lawyer). “I’m sorry for writing such a long brief. I didn’t have time to write a short one.” I will try to do better next time, for my sake if not for yours.

***

I hope you enjoyed this post. I enjoy writing them and will never charge for subscriptions or ask for donations. I only ask that, if you did enjoy it, you press the “like” button below. And please consider sharing my posts with other people you think might be interested. Doing so will help me evaluate interest in the book I am writing on grid operations. Of course, if you have a reaction to, or question about, this post, please leave a comment and I will be happy to respond.

^[1] I provided a brief description of PJM’s capacity auction and capacity auctions in general in my January 18 post, and I recommend that you read that post first if you want a better understanding of how capacity markets work.

^[2] PJM issued a news release that provides a description of its capacity auction results: “PJM Capacity Auction Procures 138,318 MW of Generation Resources as Work Continues To Address Growing Electricity Demand,” as well as a much more detailed report; “2028/2029 Base Residual Auction Report.” The news release and detailed report can be found on PJM’s website, PJM.com.

^[3] A price per megawatt-day means the generation owner is paid that price for each day of the contract period. Since the PJM capacity contracts are for one year, this means that the generation owner is paid the price per megawatt-day times the 365 days of the year. So, for example, if the price per megawatt-day was $10, the generation owner would be paid $10 * 365, or $3,650 for each megawatt of capacity sold. The July auction price of $325 translates to an annual payment of $32,500 for each megawatt of capacity sold.

^[4] The Statement of Principles can be accessed at PJM SoP 011526 Secretary Wright and Burgum Signed (2026-01-15 1453 ET)-DOI to DOE 9 Signed.pdf.

I have decided to change up my publishing schedule a bit. It seems, to me at least, that I am posting articles too infrequently. One problem has been that my recent articles have been long and have taken a long time to write. I now intend to post shorter articles more frequently, at least for a while until I run out of ideas. Today’s post will be about an important new development in the nuclear power front. Next week I hope to write about the results of PJM’s annual capacity auction, which are supposed to be announced at the end of the day on Tuesday, June 14.

As I have written before, only two new nuclear reactors have been constructed in the US since 1996. These reactors, Vogtle Units 3 and 4, took 15 years to construct before coming online in 2023 and 2024 respectively. Originally projected to cost about $7 billion each, the combined cost of the reactors exploded to $34 billion over the 15 year construction period. Construction of two other reactors, the V.C. Summer reactors in South Carolina, was abandoned after eight years of construction in 2017, when costs reached $25 billion.

One of the primary causes of the originally high cost estimates, and the staggering cost increases of these reactors has been the onerous Nuclear Regulatory Commission (NRC) licensing process. The Biden and Trump administrations have both attempted to reform this process. President Biden supported the passage of the Accelerating Deployment of Versatile, Advanced Nuclear for Clean Energy (ADVANCE) Act of 2024 and last year President Trump issued Executive Order 14300, which required the NRC to update its regulations licensing regulations. I wrote about these efforts in my June 2025 post, Will the New Nuclear Executive Orders Jumpstart the Construction of New Nuclear Reactors?

The NRC issued a series of three proposed rule changes in one week, starting on July 1, 2026, intended to implement the requirements of both the ADVANCE Act and Executive Order 14300. These rule changes, if adopted, could go a long way towards reducing the cost of new reactors.

The first of these proposals, issued on July 1, revised the NRC’s radiation protection regulations setting public and worker dose limits.^[1] The NRC applies an “as low as reasonably achievable” (ALARA) principle for determining radiation dose limits. This seems like a reasonable principle, and 111 nuclear reactors were constructed from 1960 to 1992 in compliance with the NRC’s regulations implementing that principle. In 1994, however, the NRC implemented a new approach requiring the use of a model, known as the “linear no-threshold” model (LNM) for setting radiation exposure levels. The LNM model essentially required exposure to radiation to be completely eliminated. Ostensibly, the LNM model implemented the ALARA principle, but in fact it established an extremely low limit that was not reasonably achievable and was not required to ensure public safety. It is no coincidence that only three nuclear reactors have been placed in service—including the two Vogtle reactors costing $34 billion--since the use of the LNM model was required.

The NRC’s proposed rule would eliminate the LNM model. In its place, the NRC proposes to set specific radiation dose limits. Although reasonably achievable, these limits are below levels “associated with known health effects.” As such, the proposed limits would seem to satisfy the ALARA principle.

The second proposed rule, also issued on July 1, aims to modernize the NRC’s reactor licensing, safety oversight, and siting practices.^[2] These proposed rules are intended to implement Executive Order 14300’s requirement that the NRC revise its rules to: (1) establish stringent thresholds for circumstances in which changes to reactor design can be required once construction is underway”; (2) adopt revised thresholds to ensure that reactor safety assessments focus on credible, realistic risks”; and (3) reconsider the regulations governing the time period for which a renewed license remains effective, and extend that period as appropriate. The NRC also proposed a number of additional changes to its regulations intended to improve the efficiency and efficacy of its licensing process.

The third proposed rule, issued on July 7, implements a comprehensive revision to the NRC’s environmental review of proposed nuclear reactors under the National Environmental Policy Act (NEPA).^[3] These proposed changes are consistent with the Trump Administration’s general policy of streamlining NEPA environmental reviews. The previous review process caused delays of several years in completing the NEPA reviews that, by law, must precede construction.

These proposed rule revisions represent an important first step in speeding up the NRC’s review process and reducing the cost of new nuclear reactors. But they are only a first step. Under the process for agency rulemaking, the NRC has given the public 45 days to submit comments. The NRC will then have to take the time to review the comments, make any appropriate revisions, and write up a lengthy response to the comments that must withstand judicial review. This process likely will take up at least the rest of 2026 and could take much longer. For example, it took over a year for the proposed revisions to be issued in the first place. Once final rules are issued, they will then be subject to judicial review in the US Courts of Appeals. This review could easily take 1-2 years.

But finalizing the NRC’s new rules represents only part of the equation of making the construction of new nuclear reactors more affordable. Even with the new rules, it can be expected that every proposed new reactor will be subject to bitter opposition. I am aware of recent polls showing less public opposition to new nuclear plants. But does anyone doubt that a number of people, including those who previously said they are not opposed to new nuclear power plants in the abstract, will not be happy with the idea of having a new nuclear power plant in their own neighborhood. Opposition will likely include local politicians and could also include politicians at the state and federal levels. And even though the NRC rule revisions would make opposition more difficult, there is no doubt that smart committed lawyers will be able to find ways to delay the permitting process and make construction more expensive.

Even without cost overruns resulting from permitting challenges, nuclear reactors are not cheap. For example, the initial cost estimate of the two Vogtle reactors in 2008 was about $7 billion each. It is hard to imagine that the cost of a new reactor today could be less that this. I recognize there are claims that cost efficiencies potentially could be achieved by the adoption of a single design for use in a number of reactors, such as the AP1000 design employed by the Vogtle reactors. Use of a common design would eliminate the cost of coming up with a new design for each reactor, would help to develop a supply chain, and would permit efficiency gains over time. Public opposition also likely would decline as more plants are constructed and operate safely.

But all of those benefits are dependent on two or three companies committing to be first movers before any of these efficiencies have been achieved and their effect on costs demonstrated. I am not an expert in the nuclear industry, but as far as I can tell this hasn’t happened yet. And it seems to me it is unlikely to happen at least until the NRC’s proposed rules are finalized.

If you have read my other posts, you know that I am a big advocate for the construction of new nuclear capacity. I think development of nuclear power represents the best path to achieving a low carbon grid while ensuring the grid’s reliability. But although I am heartened by the NRC’s proposed rules revisions, I don’t see how they could lead to a significant amount of new nuclear capacity being placed in service any sooner than 10 years from now. And it could take longer or even never happen at all. I hope I will be proved wrong.

***

Before I go, I want to recommend a post in Kilovar’s Substack, entitled “The Record We Won’t Know for Sixty Days,” (July 8, 2026). This very interesting article explains how the eastern and midwestern Regional Transmission Organizations got through the recent heat wave without having to implement any power outages. Kilovar’s posts are more technical than mine. But they are understandable and it is well worth working through his explanations to see what happened.

***

I hope you enjoyed this post. I enjoy writing them and will never charge for subscriptions or ask for donations. I only ask that, if you did enjoy it, you press the “like” button below. And please consider sharing my posts with other people you think might be interested. Doing so will help me evaluate interest in the book I am writing on grid operations. Of course, if you have a reaction to, or question about, this post, please leave a comment and I will be happy to respond.

^[1] Reforming and Modernizing the NRC’s Radiation Protection Framework, Docket NRC-2025-1140 (July 1, 2026).

^[2] Modernizing Reactor Licensing, Safety Oversight, and Siting Practices, Docket NRC-2025-0975 (July 1, 2026).

^[3] Implementation of the National Environmental Policy Act, Docket NRC–2025–0478 (July 7, 2026).

This is the second part of my two-part post on mitigating electricity price increases caused by data centers. My first post, dated May 12, explained the various ways that high data center demand for electricity does, and does not, increase the price of electricity. Today, I will provide my take on various options that have been proposed. Spoiler alert. Some of these are better than others.

1. Retail Rate Freeze. During the 2025 New Jersey gubernatorial campaign, Mikie Sherrill—the eventual winner of the election—promised to freeze electric rates in New Jersey. But, although retail electricity prices can’t go up if they are frozen, a rate freeze is not the answer to the data center demand problem. The electric utilities in New Jersey have very little control over the costs of the electricity they sell. They don’t own any generation facilities, and they purchase all of their capacity and energy from PJM’s wholesale markets at prices established in PJM’s auctions. A rate freeze would simply force New Jersey’s electric utilities to eat the increasing costs they have no control over but would have to pay to acquire electricity and deliver it to their retail customers. This would not be a good thing.

To see why, we can look to the California Energy Crisis of 2000-2001, when wholesale electricity prices skyrocketed due to defects in California’s competitive market design. Most people think that California consumers had to pay the high electricity prices prevailing at the time. In fact, however, the high prices were limited to the wholesale market. The California utilities that sold electricity to retail customers had previously agreed to freeze their retail rates for a period of time. This rate freeze was part of the complicated deal involving the formation of the California wholesale market and the introduction of competition to California.[1] Their retail rates were still frozen when the energy crisis hit, and as a result the utilities were forced to sell electricity to retail customers at prices significantly higher than the prices the utilities had to pay to purchase that electricity. The utilities incurred huge losses, threatening their financial viability.

Similarly, if the retail rates of utilities were frozen during a period of rapidly climbing wholesale prices, the financial viability of those utilities would be threatened. Since the utilities have little control over the costs of the electricity they acquire at wholesale, the only way the utilities could survive financially would be to reduce other costs, such as vital operations and maintenance costs. And just as your car inevitably will break down if you don’t perform the necessary maintenance, the inevitable result of a rate freeze at a time of rising wholesale costs would be increasing equipment failures followed by blackouts. Each state’s circumstances are different, but forcing utilities to operate at a loss would only deteriorate the quality of service utilities provide without addressing the cause of the problem.

Conclusion: Rate freezes are not a viable approach.

2. Banning the Construction of New Data Centers. In a similar vein, a number of states and smaller localities are considering implementing a ban on the construction of new data centers. But this is rather like trying to put the genie back into the bottle. AI is with us today, and it is foolish to think that we can put a stop to it. Some bans might be put in place in certain states or localities, but I do not think this will be widespread.

Moreover, banning new data centers in a locality will not necessarily prevent electricity costs from increasing in that locality. The grid is becoming increasingly interconnected, and grid costs are becoming regionalized. For example, every utility located in the PJM Regional Transmission Organization, the largest electric system in the country, has to pay its share of PJM’s capacity auction costs. It is the spike in those costs that is largely responsible for the electricity prices in New Jersey.[2] Banning new data centers in a state or other locality in PJM will not insulate consumers from paying their share of PJM’s capacity costs.

Conclusion: Bans on new data centers are not a realistic approach.

3. Promoting the Construction of New Generation Capacity. Once candidate Sherrill became Governor Sherrill, she apparently realized that a rate freeze is not a viable option. Instead, Governor Sherrill proposed some rate credits and issued an executive order (Executive Order 2) intended to jumpstart new construction of generation capacity employing three different generation capacity technologies: (1) solar power coupled with batteries; (2) natural gas-fired generation; and (3) nuclear power. (The press release for this executive order omitted any mention of the order’s provisions intended to increase natural gas-fired generation capacity).[3] Similar efforts are being made in other states, and the US Department of Energy (DOE) is encouraging the construction of large amounts of new nuclear power plants, which I described in my June 1, 2025 post, Will the New Nuclear Executive Orders Jumpstart the Construction of New Nuclear Reactors?

Unlike Candidate Sherrill’s proposed rate freeze, Governor Sherrill’s attempts to jumpstart the construction of new generation capacity is a good idea. Bringing new capacity online is really the only way out of the grid’s current dilemma. Increased supplies of capacity should bring down the high capacity market prices that currently prevail and should reduce energy market prices as well. The problem is that, even if the construction of new generation facilities is accelerated, it will take a number of years for enough new capacity to be placed in service. It took over 15 years for the most recent nuclear power reactors, Vogtle Units 3 and 4 in Georgia, to be constructed and placed in service. Permitting reforms hopefully will greatly reduce construction times (and costs), but it will be several years before any significant amount of new nuclear power capacity will be operating. Natural gas combustion turbines and combined cycle facilities can be constructed more quickly but still need three or more years to be completed. And if new natural gas pipeline capacity is needed to supply new gas-fired generation, that can take considerably longer. It likely will be quite some time before new generation capacity will have any significant beneficial effect on electricity prices.

I should note that about 120,000 megawatts of wind and solar capacity is planned to be constructed between 2026 and 2029.^[4] And Governor Sherrill’s Executive Order 2 also proposes to accelerate the construction of additional solar generation capacity in New Jersey. Wind and solar capacity typically can be added more quickly than conventional generation capacity. However, due to their intermittent nature, wind and solar generation facilities are not well suited to provide the 24/7 supply of electricity required by data centers. And the capacity value provided by wind and solar starts to diminish as the total amount of wind and solar capacity in an electric system increases. Construction of new wind and solar facilities may help to reduce price increases resulting from increased data center demand to some extent, but what is mostly needed is new conventional generation capacity.

Conclusion: Facilitation of the construction of new generation capacity is an essential long-term approach, but is not likely to mitigate price increases in the short term

4. Retaining Existing Capacity Slated for Retirement and Recommissioning Recently Retired Capacity. In my recent post, What is the Real Problem With Renewable Solar and Wind Generation? I noted that 145,000 megawatts of coal and oil-fired generation capacity was retired between 2015 and 2025. A large amount of natural gas-fired capacity also was retired, but my data source netted out additions and retirements and as a result I was unable to determine the amount of natural gas-fired retirements. The Energy Information Administration’s 2024 Electric Power Annual also shows planned future retirements for both coal and natural gas-fired capacity, and this data shows that close to 37,000 megawatts of coal and natural gas generation capacity is planned to be retired between 2026 and 2029.^[5]

As I have written in recent posts, the North American Electric Reliability Corporation (NERC) and DOE have released “resource adequacy assessments” showing the potential for capacity shortfalls in the future that could lead to widespread blackouts.^[6] Among the data that NERC and DOE rely on in doing their assessments are the data on planned generation capacity retirements. Some of this capacity is nearing the end of its useful life and must be retired. But much of that capacity is being retired for economic and/or environmental reasons even though it is capable of continued operations. Keeping generation facilities online past their planned retirement dates would go a long way towards ensuring that enough capacity is available in the near term until sufficient new capacity can be constructed. And because these facilities already are operational, there is very little incremental cost required. Consequently, keeping existing generation facilities online past planned retirement dates is one of the best ways to meet increasing data center load without causing the price of electricity to increase. This is something the Department of Energy has been attempting to do since May of 2025 through the issuance of emergency orders, as described in my June 8, 2025 post, Is Bringing Generation Facilities Back From the Grave a Good Approach to Grid Reliability?

A closely-related approach is to recommission recently-retired generation facilities and bring them back online. A widely-reported example is Constellation Energy’s agreement to recommission Constellation’s Three-Mile Island Unit 1 nuclear power plant (now called the “Crane Clean Energy Center) and sell its output to Microsoft data centers.^[7] Two other nuclear reactors, the Palisades Nuclear Plant in Michigan and the Duane Arnold Energy Center in Iowa are also being recommissioned. Large coal and natural gas generation facilities likewise can be recommissioned. This is not cheap. For example, the estimated cost for the three nuclear reactors currently undergoing recommissioning is over $1 billion apiece. But that is significantly less expensive than the cost of constructing new nuclear reactors, and the recommissioning can be accomplished in considerably less time. Similarly, the cost of recommissioning retired coal and natural gas plants is considerably less than constructing new plants and recommissioning can be achieved more quickly.

Conclusion: Extending the life of older generation facilities and recommissioning retired facilities is a good approach that can help in the near term, but the amount of old capacity that can be kept in service or recommissioned is somewhat limited

5. Making Data Centers Pay All Costs. On March 4, 2026, President Trump announced that Amazon, Google, Meta, Microsoft, OpenAI, Oracle, and xAI (AI Companies) signed a “Ratepayer Protection Pledge” in which these companies agreed to pay all the costs of the infrastructure needed to supply their data centers.^[8] This pledge provides that the AI Companies will “ build, bring, or buy the new generation resources and electricity needed to satisfy their new energy demands, paying the full cost of those resources.” Further, they ”will pay for all new power delivery infrastructure upgrades required to service their data centers, including adequate network upgrade costs.” Some states similarly have attempted to require new data centers to pay all of the costs required to serve their loads.

This sounds good, and signing the pledge was a great PR move by the AI Companies. But what does it actually mean? The pledge does not require the AI Companies to completely separate themselves from the grid and self-provide all of their electricity needs (more about that later). And if the AI Companies remain hooked up to the grid, general ratemaking principles already require that customers pay the costs incurred to provide them with electricity. What is added by this pledge?

Let’s start with that part of the pledge stating that the AI Companies will “ build, bring, or buy the new generation resources and electricity needed to satisfy their new energy demands.” First of all, data centers are already required to “buy . . . the electricity needed to satisfy their new energy demands.” No one will give them electricity for free, I promise. The relevant part of the commitment relates to the pledge to acquire generation resources. But if an AI company can satisfy this commitment simply by “bringing” or “buying” existing generation capacity, that doesn’t really address the capacity shortage problem. What is needed is new generation capacity to address the grid’s existing capacity deficit, not simply a change in ownership or a new power sales agreement to sell electricity from existing capacity.

Even if a data center does “build, bring, or buy” new capacity, how much new capacity will it be required to build, bring, or buy? Will the data center also be responsible for providing backup generation capacity to supply its electricity during those times when its generation resources are down for maintenance or suffer a mechanical problem and shut down? If not, and the data center expects to obtain its electricity from the grid when its own generation resources are down, that negates much, if not all, of the benefit provided by the data center’s acquisition of its own generation capacity. This is because the grid would still be required to acquire enough generation capacity to serve the data center’s demand any time its main generation is off-line.

Further, as we have already discussed, it takes time to build new capacity. Will the AI Companies wait for this capacity to be built before constructing new data centers? If not and they obtain their electricity from the grid until the new capacity is online, then that doesn’t really help the immediate problem facing the grid.

The AI companies’ commitment to ”pay for all new power delivery infrastructure upgrades required to service their data centers, including adequate network upgrade costs” is more straightforward. But cost recovery for transmission and distribution facilities is highly regulated, by the Federal Energy Regulatory Commission for transmission facilities and by the states for distribution facilities. As a general matter, data centers already should be allocated most, if not all of the costs of new power delivery upgrades required to deliver their electricity. If there are upgrade costs that cannot be allocated to data centers under current federal and state regulations, however, then those costs cannot be charged to data centers absent a change in the applicable regulations.

Let me end this discussion by noting two additional problems with the Ratepayer Protection Pledge. First, the pledge is voluntary and unenforceable. Nothing prevents any of the AI Companies from violating any provision of the pledge if doing so is deemed too impracticable or too expensive. It is not unlikely that this could happen. Some states are attempting to incentivize the location of new data centers by offering them, among other things, special lower rates for electricity. Are the AI Companies committing not to accept special rates that fail to cover all of the costs of the infrastructure needed to serve data centers?

Second, not all data center owners have signed on to the Ratepayer Protection Plan. Unless states impose a similar requirement applicable to all data centers, the Ratepayer Protection Plan will provide only limited benefits.

Conclusion: Although they sound good, the Ratepayer Protection Plan and similar proposals are so vaguely worded that it is difficult to know whether, if at all, they will mitigate increases in the price of electricity.

6. Requiring Data Centers to Be Completely Separated From the Grid. There is one way to ensure that data centers pay all the costs of meeting their demand for electricity. That is to require data centers to be completely separated from the grid, with no interconnection whatsoever. If they are not connected to the grid, data centers necessarily would have to be responsible for the cost of all generation resources, including backup capacity, needed to supply electricity on a 24/7 basis. They also would need to be responsible for the costs of all facilities used to deliver electricity from the generation facility or facilities to the data center. Because the data center’s electric supply would never interface with the grid, the costs of supplying the data center with electricity could never be passed on to other retail customers.

This approach also potentially provides several benefits to data centers. It provides a quicker path towards acquiring a power supply. It allows data centers to bypass the need to obtain the necessary regulatory approvals to interconnect with grid. It avoids the delay caused by the need to study the interconnection and install transmission and/or distribution upgrades if there is inadequate capacity to deliver electricity to the data center. It also minimizes public opposition because an islanded data center cannot affect the price of electricity paid by local residents or affect the reliability of the grid.

There are, however, practical difficulties with the islanding approach. Because backup capacity from the grid is not available, data centers must install backup generation capacity equal to the capacity of the primary generation facilities to ensure the 24/7 supply of electricity they need. They also need backup transmission and distribution wires. And even with backup generation and transmission capacity, the reliability of a data center’s supply of electricity cannot be as reliable as if it was connected to the grid. The grid has hundreds, if not thousands of generation facilities employing multiple different generation technologies and thousands of miles of transmission and distribution lines. Further, islanded data centers must be located in locations that also are appropriate sites for their generation facilities, including facilities for fuel supply, such as natural gas pipelines.

These difficulties are not insurmountable. But separating from the grid doesn’t make sense for just a single data center. The cost of the duplicative facilities needed to ensure 24/7 service most likely would be insurmountable. But what could work is the development of microgrids that are scaled to support multiple large data centers. For example, Pacifico Energy has proposed a private-grid power generation campus located in West Texas;[9] and FO Permian Partners & HiVolt Energy are developing a similar site in Midland Texas.[10] These microgrids will employ multiple natural gas-fired generators and are located near plentiful supplies of natural gas. It seems likely that this approach could provide an alternative electricity supply option for some data centers.

Conclusion: Separating data centers from the grid can protect customers from cost increases. However, this approach does not appear to be practical for most data centers.

7. Requiring Data Centers to Accept Curtailment of Electricity Supply From the Grid. There is one simple way to protect against electricity price increases from data centers. It involves requiring data centers to curtail their purchases of electricity from the grid during periods of high demand when there is not enough generation capacity on the grid to supply all electric demand. When this happens, the grid stops providing electricity to the data center.

As I mentioned in previous post describing the ways in which data center demand can affect prices, data center demand does not affect the grid’s reliability in most hours of the year. Rather, the problem is that the grid may not have enough generation capacity to meet the highest demand for electricity on the hottest days of the summer or the coldest days of the winter. I won’t repeat that discussion here but will use one of the graphics included in that post, plus one new graphic, to help explain how curtailment would work.

Those of you familiar with my Substack will recognize my old friend, THE CURVE, one of my not very well done graphics intended to illustrate how demand for electricity changes during a summer day. In THE CURVE, demand in each hour of the day is represented by the dotted red line. As you can see, demand is low at night when most people are asleep, increases as people wake up, go to work or school, peaks in the afternoon when the temperature is at its highest, and then declines as the temperature drops, people go home, have dinner and eventually go to bed. (The shape of the demand curve varies among different electric systems and changes in the winter, but this does not affect the point I am making here). This version of THE CURVE also shows a typical mix of different types of generation facilities used to meet the changing demand. There are large “baseload” units that run all day, somewhat smaller “intermediate” units that run for most of the daylight hours, and small “peaking” units that only run at times of peak demand.

The solid blue line represents the grid’s increased system demand once a new data center, which demands the same amount of electricity 24/7, comes online As you can see, because total demand is well below the peak demand in most hours, the grid is easily able to handle the increased data center demand in most hours without any problem. However, THE CURVE shows that the data center’s demand causes total demand to exceed the system’s generation for amount three hours in the middle of the day.

This next version of THE CURVE, which is making its Substack debut, shows what happens if the data center’s supply of electricity is curtailed when the grid does not have enough capacity.

As this version of THE CURVE shows, curtailing the data center’s demand from noon to 3 pm lowers the demand curve back down to levels where it can be served by the grid’s existing operational generation capacity. Curtailment is not required in any other hours of the day because total system demand is well below the grid’s total generation capacity.

You might wonder whether curtailment is a viable option. After all, data centers need a reliable source of electricity on a 24/7 basis and will never agree voluntarily to curtailment. But curtailment of the supply of electricity from the grid does not need to cause data centers to shut down. Most data centers already have backup generators on site to supply them with electricity when storms or other events damage the transmission or distribution system, causing temporary loss of deliveries of electricity from the grid. These generation facilities, or moderately upgraded ones, should be adequate to keep data centers up and running if their supply of electricity from the grid is temporarily curtailed during periods of peak demand.

The curtailment approach is essentially the converse of requiring data centers to pay for the generation capacity used to serve them, but allowing the data centers to take electric service from the grid when their generators are offline. Instead, data centers purchase their electricity from the grid in most hours of the year but instead rely on their own backup generators to supply electricity when the grid is unable to do so.

In addition to ensuring that the grid can meet all demand for electricity even in times of peak demand, making new data center demands curtailable also can mitigate cost increases. This is illustrated with my other favorite not so good graphic, THE STACK.

This is the basic version of THE STACK, which illustrates how prices are set in the RTO capacity and energy markets. Owners of generation facilities submit offers to sell, in this version of THE STACK, into a capacity market. Those offers are placed into a “stack,” with the lowest price offer at the bottom. The stack is then compared to the total amount of capacity being purchased, which is based on an estimate of the maximum amount of generation capacity that will be needed at the highest peak hour of the year. The highest-priced offer in the stack needed to supply the amount of capacity being purchased is given a capacity award, as are all lower-priced offers.[11] In the scenario illustrated by this version of THE STACK, the grid operator determines that it will need to purchase 850 megawatts of capacity. The generator submitting the $50 offer, and all lower-priced capacity offers, are needed to supply this amount of capacity. They each are awarded a contract to supply capacity for $50.

Now, let’s see what happens if total system demand increases by 200 MW to serve new data centers.

Now, the capacity auction yields less capacity than the 1050 megawatts the grid operator determined it needed. All generators are awarded capacity contracts, all are paid the $90 price submitted by the highest-priced offer, and the system operator is left scrambling to figure out how to deal with the shortfall. This is not a far-fetched scenario. It is in fact essentially what actually happened in the most recent capacity auction conducted by the PJM Regional Transmission Organization (as described in my January 18, 2026 post, PJM Failed to Acquire the Capacity it Needs to Ensure Reliability. What Does It Mean?).

How does the ability to curtail data center load affect the capacity market pricing? This is illustrated by the following version of THE STACK.

Because the system operator has the ability to curtail the supply of electricity to the data centers, it only needs to acquire enough capacity to serve the demands of its other customers in the peak hour of the day with the peak demand. This not only means that the right to curtail the data centers ensures that the electric system will have enough capacity, but also that the most expensive unit needed to supply the 850 MW of capacity needed by the grid is the red unit, which was offered at $50. Consequently, the capacity price paid to all suppliers of capacity is reduced from $90, without curtailment (as shown in the prior version of THE STACK) to $50 with curtailment.

Requiring data centers to agree to curtailment also could be used as a temporary measure that applies only until additional capacity equal to a data center’s demand can be constructed and placed in service. Allowing new data centers to commence operations, subject to curtailment, before the capacity they are constructing is online allows new data centers to commence operations more quickly. It also would protect them from any delays in completing their new capacity and placing it in service.

I don’t want any of you to think that I came up with the idea of making data centers subject to curtailment. This idea has been around for a while. For example, the PJM emergency capacity auction contemplated by January 15, 2026 agreement between the Trump administration and the governors of the 13 PJM states contemplated that the auction would apply only to supply data centers that have not agreed to be curtailed. (This agreement described in my February 13, 2026 post The Trump Administration and the Governors of the PJM States Have Reached an Agreement in Principle to Address PJM’s Capacity Deficit. Will it Work?) Further, PJM is planning on an alternative approach that includes, among other things, an option for data centers to agree to make their demands subject to being curtailed.

***

At the end of the day, it seems unlikely that any single approach to mitigating electricity price increases will be adopted throughout the US. Rather, different approaches described above, and maybe others, will be implemented in different parts of the country. And some states may decide that electricity price increases are an acceptable price to pay to attract the jobs and economic activity that will come with new data centers, and as a result will not attempt at all to mitigate price increases resulting from increased data center demand.

Conclusion: Requiring curtailment of data center demand in periods of peak demand could reduce costs, but only if data centers are able to provide for backup sources of electricity when they are being curtailed

***

I hope you enjoyed this post. I enjoy writing them and will never charge for subscriptions or ask for donations. I only ask that, if you did enjoy it, you press the “like” button below. And please consider sharing my posts with other people you think might be interested. Doing so will help me evaluate interest in the book I am writing on grid operations. Of course, if you have a reaction to, or question about, this post, please leave a comment and I will be happy to respond.

[1] The expectation was that the wholesale market prices at which the utilities purchased electricity would be lower than the retail price at which it was sold. The profits the utilities earned would be used to offset the “stranded costs” that the utilities were expected to bear as the result of competition.

[2] PJM’s high capacity costs are described in my January 18, 2026 post, PJM Failed to Acquire the Capacity it Needs to Ensure Reliability. What Does It Mean?

[3] Governor Sherrill also issued an executive order (Executive Order 1) that directed New Jersey’s public utility commission to “consider pursuing a pause, abeyance, or modification of the schedule governing any proceedings in which electric distribution utilities seek approvals for rate increases or cost recoveries to the extent permitted by law.” Most of the rest of the order addressed different ways that various rate credits could be given to retail customers. These rate credits do mitigate the effects of price increases caused by data centers, but do not address their root causes.

[4] Energy Information Administration’s 2024 Energy Information Electric Power Annual, Table 4.5 (Oct. 16, 2025).

[5] Energy Information Administration’s 2024 Energy Information Electric Power Annual, Table 4.5 (Oct. 16, 2025).

[6] See, A Closer Examination of the Effects of the Explosion in the Demand for Electricity Caused by Data Centers, Part One; What Can We Learn From NERC’s New Winter Reliability Assessment?

[7] Originally, there were two reactors at the Three Mile Island nuclear power plant. The Three Mile Island Unit 2 reactor was involved in the infamous accident in 1979 in which the reactor lost its cooling water and suffered a partial meltdown. Unit 2 has been permanently decommissioned and will not be restarted.

[8] The press release announcing the Ratepayer Protection Pledge can be found here: Fact Sheet: President Donald J. Trump Advances Energy Affordability with the Ratepayer Protection Pledge – The White House; and the Ratepayer Protection Pledge itself can be found here: Ratepayer Protection Pledge – The White House.

[9] (GW Ranch | Off-Grid Power Generation | Pacifico Energy)

[10] news.

[11] My October 27, 2024 post, Explaining the Grid Part Four, explains why the lower priced offers are paid the same price as the highest cost offer needed to supply the amount of capacity being purchased.

I know that, notwithstanding my hypothetical monthly publication schedule, it’s been seven weeks since my last post. Sorry about being so late. My life has been very hectic, with lots of travel and other activities, including work on my book. And this has been a hard post to write; a lot of words were left on the cutting floor of my editing room. I will post my next article in a month or less, I (hypothetically) guarantee.

In the meantime, a number of new people have subscribed or followed Explaining the Grid since my last post. Welcome to all of you. I try to write largely jargon and acronym free posts that are understandable to reasonably intelligent people, which by definition you all are. since you are reading my Substack. This post, like all others, is based on my best analysis and not on politics or name-calling. Let me know if you think I got something wrong and, if I agree with you, I will make a correction.

The inspiration for today’s post comes from J.P. Morgan’s 16^th Annual Eye on the Market Energy Paper, written by Michael Cembalest, entitled Fighting Words. (Check it out, you can find it here. Fighting Words) One of the topics in J.P. Morgan’s paper is entitled “Data centers: the pitchforks are gathering regarding their impact on US power prices.” As the name implies, the discussion describes the growing concern that recent increases in retail electricity prices in many states have been caused by data centers’ high and still growing demand for electricity. That topic inspired me to write a post about different proposals to address the effect of data center demand on electricity prices.

As I started to write my post, however, I realized that before I can discuss price mitigation proposals, I first have to explain how, if at all, data centers are causing electricity prices to increase. Anyone who reads the news knows that data centers are not very popular right now, in part because a large segment of the population is skeptical about AI. One easy way to attack AI is to claim that data centers are responsible for electric price increases. In writing my post I quickly learned that any discussion of the effects of data center demand on electricity prices is complicated. I decided that the best course is to write a two-part article, the first (this one) addressing the question of how data centers affect electricity prices, and the second (the next one) on proposals to mitigate electricity price increases caused by data centers.[1]

Let me start by noting that, to date, data centers’ demand for electricity does not explain most of the currently high electricity prices. Those prices have resulted primarily from other factors, including: (1) 13 straight years of record breaking investment in transmission and distribution facilities—more than $1.3 trillion over the last 10 years—primarily to harden the grid against wildfires, floods, and hurricanes and to interconnect with renewable generation facilities; (2) the inclusion in rates the cost of subsidies for renewable generation facilities as well as the costs of social programs completely unrelated to the grid; (3) the use of large amounts of expensive oil-fired generation in the winter as a consequence of inadequate natural gas pipeline capacity; and (4) higher prices for natural gas following Russia’s invasion of Ukraine. I am sure that data centers have received at least some of the blame for the price effects of these unrelated factors.

The effect of data center demand on the price of electricity was discussed in a recent October 10, 2025 joint study by the Lawrence Berkeley National Laboratory and the Brattle Group (Factors Influencing Recent Trends in Retail Electricity Prices in the United States: What do we know? Where are the gaps?). This study concluded that, contrary to the claim increased demand from data centers is causing retail electric price increases, the states with the highest demand growth over the previous five years actually saw a decline in their electric prices. By contrast, states whose demand declined often saw increases in the price of electricity. The study concluded that the greater use of the grid’s assets resulting from meeting higher demand allowed the costs of those assets to be spread over a greater amount of electricity sales, thereby lowering the cost per unit of electricity sold. It’s like inviting a family to share your beach house for a week. Splitting the cost between the two families significantly reduces the cost each family has to pay.

But at some point, your beach house won’t hold any more people, and another house will have to be rented if more families want to come. Similarly, the cost of producing and transmitting electricity will go up when the grid needs to add new infrastructure to serve data centers’ increased demand. And significant new infrastructure is starting to be required. The potential resulting cost increases can be separated into three categories: (1) the cost of new generation facilities; (2) the cost of new transmission and distribution facilities; and (3) the wholesale price of electricity. This post will discuss each of these three categories. I also will discuss a fourth factor, perhaps the most important factor of all: how these costs incurred by utilities selling electricity at retail to consumers are reflected in the price they charge for electricity.

Cost of Generation Capacity

It is intuitively obvious that increased data center demand for electricity will, at some point, require new generation capacity to supply that increased demand. I think it is helpful, however, to explain exactly how this happens. That knowledge will be especially important in my next post discussing ways to mitigate the price effects of increased demand. To do this, I am going to dust off one of my not very good old graphics—THE CURVE—which I haven’t used in a while.

To start the discussion, I first am going to present THE CURVE in its most basic form.

This version of THE CURVE illustrates how the amount of demand for electricity on the grid changes during the day and night in the summer. The curved red line shows the amount of demand at different times of the day. As you can see, demand is low at night when most people are asleep. But it is not zero, because some people are up, your refrigerator and AC are running, streetlights are on, etc. Demand increases as people wake up, go to work or school, and increases further as the temperature gets hotter. Demand peaks in the afternoon when the temperature is at its highest, and then declines as the temperature drops, people go home, have dinner and eventually go to bed. (The shape of the demand curve varies among different electric systems and changes in the winter, but this does not affect the point I am making here)

The next version of THE CURVE shows why the shape of the demand curve has important implications for the grid.

As you can see, I have inserted into THE CURVE the different types of generation needed to supply demand throughout the day. The bottom row consists of large generators, known as “baseload” generators. that operate all day and night. Baseload generators are the only generators needed when demand is low at night, but as demand increases throughout the day additional generators also need to be turned on. The second row consists of somewhat smaller “intermediate” generators that run for much of the day but typically not that much at night. The third row includes even smaller generators and the fourth row the smallest generators, known as “peakers.” The peakers are used in the relatively few hours in the middle of the day when the electric system’s demand is at its highest; known as the peak (I could have put some wind and solar generators in the graphic but that adds a layer of complexity not needed for the point I am making here).

The final version of THE CURVE shows the potential problem raised by increased data center demand.

As you can see, data center demand is relatively constant all day, so it increases total system demand by about the same amount each hour of the day. For most of the day, there is more than enough generation capacity to serve all demand even when the data center demand is included. It is only when, as shown in the graphic, data center demand causes the peak system demand to exceed an electric system’s entire complement of generation that there is a problem. And this problem lasts for only a short time, at most a few hours, during the time of peak demand.

Typically, data center demand will not exceed an electric system’s total generation capacity, even during peak demand hours. All electric systems are required to acquire more generation capacity than they project will be needed during their highest peak demand hours. This difference between an electric system’s total generation capacity and its expected peak demand is known as its capacity reserve margin.

Capacity reserve margins shrink as demand increases. To keep the margins at appropriate levels, new generation capacity must be added to offset the increase in demand. And sufficient new generation capacity has not been constructed. Studies performed by the North American Electric Reliability Corporation—the entity designated by Congress to ensure the grid’s reliability—and by the Department of Energy have identified potential generation capacity shortages in the near future. To date, the forecasted potential blackouts would occur only in the event of extraordinary circumstances, such as very high temperatures and greater than normal generator outages in the summer, or very cold weather that forces generators offline in the winter. As of yet, no one has forecast blackouts under typical operating conditions.

In saying this, I am not trying to make light of the problem. Several regions in the US have suffered from major heat waves and/or severe winter storms in the past few years. It appears likely that we will continue to suffer from extreme summer heat and winter cold in the future. Without sufficient new generation capacity additions, it is inevitable that large regions of the country will suffer from persistent rolling blackouts in the near future.

This need for new generation capacity has already had an effect on generation capacity prices. For example, the price for capacity currently being charged by the PJM Regional Transmission Organization (RTO), established in an auction conducted in 2024, reflects the need for new generation capacity to serve rapidly increasing data center demand for electricity. That price, $270 per megawatt/day,[2] is about 10 times higher than the capacity price for the previous year, and about $100 higher than the highest price prior to the current year. The PJM capacity prices in PJM’s following two auctions, which will take effect in June of 2026 and June of 2027 respectively, are about $60 higher than the current capacity prices.[3] Despite this high price, the most recent PJM capacity auction failed to acquire the amount of generation capacity PJM determined was necessary to ensure the reliability of its system.[4] Similarly, capacity prices have gone up in the last few years’ auctions in the three other RTOs that conduct capacity auctions (ISO-New England, New York ISO, and the Midcontinent ISO).

It is important to note that the capacity prices established in the RTO capacity markets are paid to all generators receiving capacity contracts. And all utilities located in an RTO that sell electricity to retail customers must acquire capacity from the RTO at the capacity auction market price. Consequently, high capacity prices resulting from increased data center demand for electricity increase the costs incurred to provide electric service to all retail customers in these RTOs. Outside of the RTOs, electric utilities also need to construct or purchase new capacity to maintain their capacity reserve margins, the cost of which also must be recovered in their prices for electricity.

Cost of New Transmission and Distribution Facilities

In addition to requiring new generation capacity, upgrades to transmission and distribution facilities may be required to deliver electricity to data centers. These upgrades can be expensive, with costs in the hundreds of millions of dollars. The need to construct these expensive upgrades to serve data centers is frequently cited by opponents who argue that data centers have caused increases in electricity prices.

As I noted near the beginning of this post, there has been record investment in the grid’s transmission and distribution systems for 13 years in a row, and this investment represents a significant factor in electricity price increases. This investment, however, was not driven by increased data center demand but instead has primarily been directed towards hardening the grid against natural disasters and interconnecting the grid with wind and solar generation facilities in remote locations. The increase in demand associated with data centers has been too recent to have had much of an impact on transmission and distribution upgrade costs to date.

Data center demand certainly is driving the need for future transmission and distribution upgrades. However, any future upgrade cost increases resulting from data center demand for electricity likely will be largely offset by the revenues from data centers’ increased use of the existing transmission and distribution system. As we saw on the various versions of THE CURVE graphic, the grid operates at levels well below its maximum capacity for most of the day, especially in the night hours. This means there is more than enough transmission and distribution capacity in most, if not all, hours of the day to deliver the increased amount of electricity needed by data centers on a 24/7 basis. As the Lawrence Berkeley National Laboratory/Brattle Group joint study showed, higher demand reduces the per unit price of electricity needed to cover the costs of facilities. When upgrade costs are considered in conjunction with the revenues resulting from increased transmission and distribution revenues, the effect of increased data center demand on transmission and distribution system costs should not be significant.

Wholesale Cost of Electricity

A not very well known effect of increased data center demand for electricity is that the higher demand should increase the wholesale cost of electricity. I will return to my frequently-used graphic, THE STACK, to explain why this is so.[5] This discussion likely will be repetitive for those of you who have met THE STACK before, but hopefully it will be useful to my new subscribers.

If you have already read any of my previous posts where THE STACK made an appearance, then you know how this works. To recap, all offers to sell electricity in an RTO wholesale market are stacked, from lowest cost offer to highest cost offer. The RTO then determines how much customer demand exists, and then compares the total amount of customer demand with THE STACK. The highest priced offer needed to serve the last megawatt of customer demand—known as the marginal offer or the market-clearing offer—and all lower priced offers, receive awards to sell electricity. And every winning generator is paid the market-clearing price even if it offered to sell at a lower price. If you want to know why everyone gets that price, read Explaining the Grid Part Four—RTO Energy Markets.

If you look at THE STACK, you can probably figure out quickly how an increase in demand can increase the price of electricity. If an increase in demand results from data centers’ need for large amounts of electricity, the total demand can easily move up the stack to the next highest cost offer to sell electricity, as shown by the following version of THE STACK:

Demand has increased by 150 MW, from 850 MW to 1000 MW. This increase in demand has moved the price of the marginal market-clearing price from $50 to $75. And that increase in the marginal market-clearing price means the entire 1,000 MW of electricity is sold at a price of $75/MW, a $25/MW increase in price caused by increased data center demand for electricity.

The actual setting of prices in RTO markets is much more complicated because there are hundreds, if not thousands, of offers submitted into those markets, and in the energy markets prices are calculated as often as every five minutes. And because there are so many more offers being submitted, the difference in prices between offers is usually much less than shown in my STACK graphic. But the general principle is the same. An increase in demand for electricity can cause an increase in the wholesale price of electricity, at least in the short run.

This does not necessarily mean that the increased demand for electricity caused by data centers will lead to higher prices in the RTO wholesale markets in the longer term. As the demand for electricity in an RTO increases, so will the amount of generation capacity (hopefully). And if the new generation capacity is able to offer its electricity into the market at a relatively low price, that lower price offer could offset the price effects of the increase in demand, or even result in a decrease in the RTO market price. This is illustrated by the third and last of my STACK graphics. In this graphic, the 150 MW increase in demand is accompanied by an offer from a newly constructed low cost generation facility (shown in green on THE STACK) at a price of $25. This new offer moves the higher priced offers up in the stack so that the marginal cost offer once again comes from the facility offering to sell at $50. Because the 150 MW increase in demand was offset by the new low cost generation facility, the increase in demand had no effect on the market-clearing offer price, returning the market price to $50, where it was before data centers caused the demand for electricity to increase.

It is impossible to know today whether and when newly constructed generation facilities might offset increased data center demand. But for the near future at least, it appears likely that demand will increase faster than supply. If that happens, the likely result will be an increase in the wholesale cost of electricity.

Recovery of Increased Costs in the Retail Price of Electricity

Up to now, we have discussed how increased data center demand for electricity has affected the costs incurred by utilities that sell electricity at retail to consumers. But that is not the end of the story. Utilities develop rates governing the price of the electricity they sell, based on the costs they incur. Regulators, typically a utility’s state public utility commission, then must approve those rates. Regulators are required to allow utilities to recover all of their operating costs in their electricity rates. Consequently, they cannot simply bar utilities from including in their rates the costs associated with increased data center demand.

But regulators do have the ability to establish different rates for different classes of customers. And general regulatory principles provide that customers should be charged the costs incurred to provide their service. This means that regulators have the ability to establish a separate customer class for data centers. They can then allocate to this separate customer class all of the additional costs associated with data center demands for electricity. In this way, regulators can protect their other customers from having to pay higher electricity prices based on the costs caused by data center demand for electricity.

It would be relatively simple to determine the cost of upgrades incurred to deliver electricity to data centers served by a utility and to allocate all of those costs to the rates charged to those data centers. These costs are calculated before an upgrade is constructed and are easy to identify. But it is not necessarily an easy thing to identify and assign to data centers all of the other costs associated with their demand for electricity. For example, it would be much more difficult to determine how much of any particular RTO-wide capacity or energy market price is attributable to a data center or group of data centers. Nevertheless, regulators have the authority to set utility retail rates in such a manner as to protect utility retail customers, at least to some extent, from price increases caused by data centers’ increasing demand for electricity.

I have not done a survey of how state utility commissions have set the prices of the electricity their utilities charge to data centers, but I do not think that many states to date have attempted to allocate greater costs to data center rates to compensate for the increased costs data centers impose on utilities. To the contrary, I understand that, in some states, data centers have been offered relatively low electric rates as an incentive to locate data centers in that state. The states do, of course, have the prerogative to offer this type of incentive to data centers if they believe the benefits outweigh the costs. But doing so requires that other customers pay more for their electricity.

***

I hope you enjoyed this post. I enjoy writing them and will never charge for subscriptions or ask for donations. I only ask that, if you did enjoy it, you press the “like” button below. And please consider sharing my posts with other people you think might be interested. Doing so will help me evaluate interest in the book I am writing on grid operations. Of course, if you have a reaction to, or question about, this post, please leave a comment and I will be happy to respond.

[1] I know I wrote a post last November on whether data center demand causes electricity price increases (How Does the Explosive Growth of Ai’s Electric Demand Affect the Price of Electricity?), but I have some additional thoughts based on developments since then and more thinking about the issue.

[2] To translate a megawatt/day price into an annual charge for capacity, it is necessary to multiply the price by the 365 days of the year. For example, with the current $270 capacity price, the capacity price paid for each megawatt of a generation facility’s capacity would be 365 time $270 which equals $98,550. The capacity price paid for a not very large 100 megawatt facility would be $9,855,000, and for a large 1,000 megawatt facility $98,550.000.

[3] If you are interested, I provided a more detailed post about PJM’s recent capacity auctions in my January 18, 2026 post, PJM Failed to Acquire the Capacity it Needs to Ensure Reliability. What Does it Mean?

[4] I discussed this auction and its implications in more detail in my January 18, 2026 post, Pjm Failed to Acquire the Capacity it Needs to Ensure Reliability. What Does it Mean?

[5] THE STACK made its debut in Explaining the Grid Part Three—Market Products, and then made a repeat appearance in Explaining the Grid Part Four—RTO Energy Markets.

What is the real problem with renewable wind and solar generation? Well, it’s not that these generation facilities are a fraud or a scam, allegations I frequently hear. And its not that, as one person told me, “[t]here’s no development [of renewable resources] being ‘fostered.’ There’s just graft, profiteering, and fraud.” The fact is that solar and wind generators do work as designed. And they generate large amounts of electricity. According to the most recent Electric Power Monthly report published by the U.S. Energy Information Administration, solar and wind produced close to 20 percent of all electricity generated in the US. That’s not a scam.

It is true that federal and state subsidies have encouraged investments in wind and solar facilities and made them more profitable. But that’s what subsidies are intended to do. We can debate whether these subsidies represent good or bad policy. But the subsidies for renewables do not make them a scam any more than the subsidies currently provided by the federal government and some states to nuclear generators make nuclear power a scam. The state and federal subsidies for renewable capacity are all structured so that wind and solar generators mut produce electricity in order for them to be profitable. No one can get rich investing millions of dollars in a wind or solar facility that does not generate electricity.

Nor is it correct that production of electricity from renewables is unpredictable. We can predict to the very second when the sun will rise and when it will set. And we are also reasonably good at predicting daily temperatures and cloud cover. From this, we can derive reasonably good predictions of the output of solar facilities. Similarly, it is possible to predict with reasonable accuracy when the wind will be blowing and how hard it will blow.

I recently visited the headquarters of the PJM Regional Transmission Organization (RTO) as part of my work in writing my book about the grid. While there, I was told by the PJM operators that they routinely perform six-day forecasts of both solar and wind production. The operators told me that their forecasts are reasonably accurate for the day the forecast is made, although they are less accurate for each following day of the forecast. We also know there is a reasonable correlation between the output of solar facilities and the high temperatures that cause the demand for electricity to peak in the warm summer months. This makes solar power especially helpful in meeting summer demands for electricity.

Furthermore, once wind and solar generation facilities have been constructed, they do not require fuel to generate electricity, which means they can produce electricity almost for free. It is not inevitable, as some claim, that adding new renewable generation capacity will cause the price of electricity to increase.

But even though solar and wind generators are not scams, it is also true that the intermittent nature of solar and wind generation places strict limitations on these facilities’ ability to support the reliability of the grid. And these limitations have very important consequences for grid operations. Let me explain.

First, although it is possible to predict wind and solar output with reasonable accuracy on at least a same day basis, sometimes those predictions show that there will be relatively little wind and solar electricity produced. When that happens, there is nothing that can be done to increase these renewal facilities’ production of electricity. For example, the same forecasts that precisely identify when the sun will rise and solar panels start to generate electricity also show precisely when the sun will set and solar panels stop producing electricity. Similarly, sometimes weather forecasts show little or no wind is expected, which means little or no wind energy will be produced.

The New England Independent System Operator (ISO-New England) publishes reports on its web site that show the output of various types of generation located in New England on a daily basis. Using this data, I created the two graphs below, based on this data, to help illustrate the discussion. Each graph shows what happened in New England this past January of 2026, when New England was hit by frigid weather from Winter Storm Fern during the last ten days of the month.

This first graph shows the daily output of New England’s solar and wind generators in January, both individually and combined. As you can see, these daily outputs were quite volatile. The average daily solar output for the month was close to 5,000 megawatt-hours a day, but on three different days the solar generators produced less than 1,000 megawatt-hours, and on January 26 they produced less than 500 megawatt-hours, only about 10 percent of the average. Wind was similarly volatile, with the lowest day’s production less than 25 percent of the average.

An electric system must be able to serve its customers’ full demand for electricity on a 24/7 basis. If it cannot do so at any point in time, the lights go out. As a consequence, an electric system’s reliability is only as good as its worst day. This means that, in looking at these graphs, it is the minimum electric output of solar and wind facilities, not their average daily output, that is relevant in considering their contribution to the grid’s reliability. [1] The same is true for every generation technology employed on the grid.

When viewed in this way, the limitations of solar and wind generation facilities are apparent. The average daily solar output in New England in January 2026 was a little under 5,000 megawatt-hours. But if New England had relied on solar generators to produce 5,000 megawatt-hours every day of the month, then the lights would have gone out on January 26, when only 488 megawatt-hours of solar electricity was generated. The lights also would have gone out on the 15 other days of the month, including each of the last seven days of the month, when solar production was under the 5,000 megawatt-hour average output.

The story is a little better when looking at the combined solar and wind production because wind production was somewhat less volatile. But the combined wind and solar output was still volatile. There were 17 days in January of 2026 when the combined wind and solar output was less than the average daily output of just under 23,000 megawatt-hours. And there two days when the combined wind and solar output was under 8,000 megawatt-hours, which is only about one third of the average daily output in that month. Consequently, even though the combined solar/wind electric output in New England was as high as 34,000 megawatt-hours on January 12, 2026, solar and wind could not have been relied on to consistently produce even 8,000 megawatt-hours in a day.

Batteries or other storage technologies can help to smooth out the volatility of the wind and solar output. But it is important to note that batteries only store electricity, they do not produce it. This means that an electric system has to produce more electricity than needed to serve demand in some hours in order to use electricity stored in batteries in other hours. That can be difficult to do when an electric system is under stress, as the New England system was during Winter Storm Fern. And batteries typically are used to store electricity only for a number of hours, not days. To my knowledge, batteries could not have been used to store the relatively larger amount of wind and solar electricity that was generated in the first half of January to be used in the second half of the month. Perhaps the story might be better in the future if larger amounts of battery and other storage facilities are placed in service, but today there is only about 2,500 megawatts of storage capacity in New England, which is not enough to have a meaningful ability to smooth out the daily volatility in renewable energy production. Storage capacity in the US as a whole is similarly limited.[2]

The bottom-line conclusion to be reached from this data is that New England’s solar and wind capacity did assist ISO-New England in meeting the increased demand this past January caused by the cold weather. But ISO-New England was able to keep the lights on in January only because it also had adequate conventional generation capacity available to replace the solar and wind capacity when those facilities were unable to generate much electricity. Ironically, this backup capacity consisted primarily of oil-fired generation, which is the most expensive and most polluting type of generation capacity. But without that expensive, dirty, oil-fired capacity, the lights in New England would have gone out many days this past January.[3]

This second graph is easier to explain. It compares the January 2026 daily solar output in New England with New England’s average solar output for the month of July, 2025. New England’s peak demand for electricity in 2025 was in July, and it seemed to me that the average solar output in that month represents a reasonable proxy for how much electricity could be produced by New England’s solar generation facilities under optimal conditions. This comparison between solar output in January and July illustrates an important aspect of solar power that does not receive much attention.

As my graph shows, New England’s daily solar output in January never reached even half of the average July output. And on two days the daily output was only two percent of the average July output. On average, the amount of electricity solar generators produced each day in January was only 20 percent of what they produced in July. The important conclusion to be reached from this graph is that, while solar power can be useful in the summer, it is not a good resource for meeting the grid’s demand for electricity in the winter. It would have taken 50 times more solar generation capacity in New England on January 18 and 25 for solar generators to have produced the same amount of electricity that they produced on average every day last July.

New England is not unique in this regard. All other regions in the US see similar reductions in solar output in the winter. This is illustrated by the following two graphs, based on data from the Energy Information Administration’s Electricity Browser.

The first graph shows the monthly variation in solar output in the US since 2017, when solar electricity production first started to reach significant levels. Since 2020, the country’s solar output in the summer has been two times or more higher than in the following winter.

The second graph charts the same data for California, which has a relatively warm climate and is one of the country’s leading producers of solar electricity. It shows that California follows the same general trend as in the country as a whole. Every year, California’s monthly winter solar electricity production is approximately half, or less than half, of its summer production. It thus is clear that, regardless of location, solar energy does not make a reliable contribution to the grid’s ability to meet all demand for electricity in the winter.

***

My discussion of the important limitations on the ability of wind and solar generators to reliably supply electricity does not mean that I am going back on my statement at the beginning of this post that renewable solar and wind generation is not a fraud or a scam and that these generators provide a significant amount of valuable low-cost electricity. The real problem with renewables is not with solar or wind technology. Nor is it with the push to construct increasing amounts of wind and solar generation facilities.

Rather, at the same time there was a strong push over the last 15 years to boost the construction of renewable energy facilities, there also was a push to reduce the grid’s fossil generation capacity. This was largely done through state action because the states have exclusive jurisdiction over the construction and retirement of generation facilities, but federal regulation and public interest group litigation also has played a role.

The first target was coal-fired generation. Burning coal has severe environmental consequences. In addition to large amounts of carbon, coal can also emit sulfur, mercury, lead and other heavy metals, nitrous oxide, and ash.[4] Despite recent claims, coal is neither beautiful nor clean. Consequently, some states have required coal plants to retire before the end of their service lives, environmental groups have initiated litigation designed to force those plants to shut down, and the federal government has issued regulations requiring the installation of expensive pollution controls that can render coal plants uneconomic. The result has been a significant reduction in the country’s coal-fired capacity since 2014. As shown in the following graph, the amount of coal-fired capacity in the US declined from about 300,000 megawatts in 2014 to about 170,000 megawatts in 2025. This is a reduction of 130,000 megawatts (43 percent) in just 12 years. In addition, over the same period, over 14,000 megawatts of oil-fired generation also have retired, again for cost and environmental reasons.[5]

It’s hard for me to argue that, when viewed in isolation, retiring old coal plants is a bad thing. The harmful environmental effects of these plants cannot be contested. Further, these plants are uneconomic when compared to almost all other forms of generation. Oil-fired generators are even dirtier and more expensive. I can almost hear myself saying good riddance.

But unfortunately, we can’t view the loss of this capacity in isolation. However beneficial the reduction of coal-fired generation capacity might be for the environment, we also have to consider the effect on the reliability of the grid of removing 144,000 megawatts of coal and oil-fired capacity. It is indisputable that this now retired capacity played an important role in ensuring that the grid was always able to meet all demand for electricity, in both hot weather and cold. Before we can say good riddance to coal and oil, we need to make sure this 144,000 megawatts of retired coal and oil-fired capacity has been replaced by other types of generation.

Among conventional generation technologies, natural gas is the logical choice to pick up much of the slack. There are abundant reserves of natural gas in the US, priced at levels that make gas-fired generation less expensive than coal. Natural gas combustion turbines are relatively inexpensive to construct as well. And burning natural gas emits much less carbon than coal and virtually none of the harmful emissions that come from burning coal. So, it is no surprise that natural gas-fired generation capacity increased at the same time coal-fired capacity declined.

But the increase in natural gas capacity did not fully replace the decline in coal and oil-fired capacity. Instead, natural gas capacity increased only by about 82,000 megawatts. All other types of conventional generation capacity, primarily nuclear and hydroelectric, stayed close to the same. The changes in total US capacity between 2014 and 2025 for the different types of conventional generation are shown in the chart below. As this chart shows, the total amount of conventional generation in the US has declined by about 62,000 megawatts since 2015, about a 7 percent reduction.[6]

This failure to match the decline in oil and coal-fired capacity with an increase in natural gas-fired generation capacity is due primarily to policies discouraging the construction of additional natural gas-fired capacity. Some states have refused to allow new gas-fired generation and mandated the retirement of all existing gas-fired generators in the next 15-20 years. These and other states have also effectively opposed the construction of new natural gas pipeline capacity needed to supply gas to new generation facilities. States and environmental groups also litigated the granting of approvals for new pipelines, delaying their construction, increasing the cost, and in some instances causing the abandonment of proposed new pipeline projects.

The reduction in conventional generation capacity between 2014 and 2025 represents only half of the equation, however. It also is important to know how demand has changed. As to that, it should not be surprising to anyone who has paid attention to the grid that we have had an explosion in demand in the last few years, due to the need to supply data centers. According to Energy Information Administration data, demand has increased by about 300 million megawatt-hours since 2014, an increase of almost 8 percent.[7] The complete equation determining the grid’s ability to continue providing reliable service thus includes both about a seven percent reduction in conventional generation capacity and about an 8 percent increase in demand.

Of course, a large amount of solar and wind generation capacity also has been constructed over the same period. As of the end of 2025, there is about 158,000 megawatts of wind generation capacity and about 151,000 megawatts of solar generation capacity.[8] Combined, the total amount of operational US renewable generation capacity is about 310,000 megawatts. The total solar and wind capacity operational today is roughly five times greater than the reduction in fossil capacity since 2014.

But is 310,000 megawatts of solar and wind capacity enough to fill the gap caused by the reduction in conventional generation capacity and the increase in demand? As we have seen, the limitations on the availability of solar and wind power means we can’t simply look at the 310,000 megawatt maximum capacity of these facilities in determining their contribution to the grid’s reliability. But what is the reliable capacity value of the country’s solar and wind generators? Figuring out how to make a valid comparison between the capacity value of wind and solar generation as opposed to conventional generation is complicated and, not surprisingly, controversial.

This is not a post about how to calculate the capacity value of renewables, so I am simply going to apply the Electric Load Carrying Capacity (“ELCC”) values used by PJM for its capacity auctions for purposes of this discussion. As I have explained in prior posts, PJM calculates separate ELCC values for different types of renewables (offshore wind, onshore wind, fixed solar, etc.), based on the attributes of the different technologies and how they contribute to PJM’s reliability. Don’t read this as an endorsement of PJM’s ELCC values or its application to all solar and wind generation facilities in the US. But the methodology PJM uses for its ELCC calculations appears to me to be reasonable in theory, and application of those values to the whole country provides a good enough estimate of solar and wind’s contribution to the grid’s reliability for purposes of this post.

Applying PJM’s ELCC values to the various types of renewable technologies currently operating in the US, I have calculated that the 310,000 megawatts of maximum solar and wind capacity in the US provides about 63,000 megawatts of reliable grid capacity. Coincidently, this is only slightly over the 62,000 megawatts of fossil generation capacity that has been retired since 2014. I didn’t know this would be the result of my use of the PJM ELCC values, I promise.

I know some of you will not accept a calculation showing that the solar and wind capacity in the country has offset the amount of fossil capacity retired since 2014. But I did this calculation not for that purpose, but rather to get a feel for whether US renewable capacity can both replace the loss of fossil generation capacity since 2014 and at the same time satisfy the 8 percent growth in the demand for electricity. And my calculation shows this is something that solar and wind capacity cannot do.

It is incontrovertible that the grid currently is in a precarious reliability position. The North American Electric Reliability Corporation (NERC) is the entity statutorily assigned the task of monitoring and ensuring the reliability of the US electric grid. NERC’s 2025 Summer Reliability Report and its 2025-26 Winter Reliability Report each identified several US regions that faced an elevated risk of blackouts this past summer and winter because of the potential for generation shortages.[9] More recently, NERC issued its 2026 Long Term Reliability Assessment, which concludes that large segments of the country face either a high or an elevated risk of blackouts by 2030. NERC’s conclusions are similar to those reached in a report issued by the Department of Energy last summer, which concluded that the risk of blackouts in 2030 would increase by a factor of 100 if no steps are taken to increase the grid’s generation capacity.[10] And at its last capacity auction, PJM was unable to acquire all of the capacity it determined it needed to satisfy its reliability requirements.[11]

We can’t blame these reliability problems on solar and wind generation technology. The limitations on these facilities’ ability to provide significant reliability benefits to the grid have always been well known and understood. Every generation technology has its advantages and disadvantages, which is why electric systems have always needed to employ a variety of generation technologies to provide reliable electric service in an efficient, low-cost manner. Rather, the current problems result from asking solar and wind generation to play a role in ensuring electric system reliability that they were never intended to be able to assume. And that is the real problem with solar and wind energy.

***

I hope you enjoyed this post. I enjoy writing them and will never charge for subscriptions or ask for donations. I only ask that, if you did enjoy it, you press the “like” button below. And please consider sharing my posts with other people you think might be interested. Doing so will help me evaluate interest in the book I am writing on grid operations. Of course, if you have a reaction to, or question about, this post, please leave a comment and I will be happy to respond.

[1] I should note that, just as there is a wide range of daily outputs by different generation technologies, there is a wide range of hourly outputs each day. To get a more accurate view of the minimum output of wind and solar facilities, it is necessary to look at hourly data instead of daily data. The ISO-New England reports, however, don’t include hourly output data. But the variety in the daily output data in New England nevertheless can be used to illustrate the problem caused by the volatile nature of solar and wind facilities.

[2] Energy Information Administration, Electric Power Monthly December 2025, Table 6.2.A.

[3] To be fair, New England’s biggest problem was not low renewable energy production, but rather a significant reduction in the production of electricity by natural gas-fired generation facilities due to a shortage in natural gas pipeline capacity.

[4] Energy Information Administration, Coal explained; Coal and the environment, Coal and the environment - U.S. Energy Information Administration (EIA).

[5] Data for this graph was taken from Table 4.2.A of the Energy Information Administration’s Electric Power Annual 2024 (data for years 2014-2024), and Table 6.1 of the Energy Information Administration’s Electric Power Monthly December 2025 (data for year 2025).

[6] Data for this chart was taken from Table 4.2.A of the Energy Information Administration’s Electric Power Annual 2024 (data for years 2014-2024), and Table 6.1 of the Energy Information Administration’s Electric Power Monthly December 2025 (data for year 2025).

[7] Energy Information Administration, Electric Power Monthly December 2025, Table 5.1.

[8] Energy Information Administration, Electric Power Monthly December 2025, Table 6.1.

[9] These reports are discussed in my post What Can We Learn from NERC’S New Winter Reliability Assessment? (December 17, 2025).

[10] DOE, Evaluating the Reliability and Security of the United States Electric Grid (July 7, 2025).

[11] I described this auction result in my post PJM Failed to Acquire the Capacity it Needs to Ensure Reliability. What Does it Mean? (January 18, 2026).

Since my last post, my wife and I went on a cruise off the Norwegian coast up above the Arctic Circle. The name of the cruise was “In search of Northern Lights,” and boy did we find them! We saw lights every night (and up there, night started around 3 pm), and also did some north of the Arctic Circle kind of things, like going dog sledding and reindeer sledding, spending the night in a yurt in a Sami camp, and visiting an Ice Hotel. All in all a very nice trip.

While we were gone, the Trump Administration[1] reached an agreement (Agreement) with 13 governors (PJM Governors) on a “Statement of Principles Regarding PJM.”[2] PJM is a Regional Transmission Organization that operates an electric system supplying electricity in a 13-state region spanning from New Jersey, Pennsylvania, Delaware, Maryland, and Virginia, and westward as far as Illinois. The governors who signed the Agreement are the governors of the 13 PJM states. These include both Democrats and Republicans, so the Agreement represents a bipartisan agreement of sorts.

In its first paragraph, the Agreement states that the Trump Administration and the PJM Governors “urge PJM expeditiously to file tariff changes at FERC that reflect” six principles set out in the Agreement. The PJM Governors also separately agreed on a seventh principle applicable only to themselves that did not involve any action by the federal government.

I have started, and then abandoned, writing this post several times. My problem has been figuring how to describe each of the Agreement’s seven individual principles in a coherent way. Ultimately, I realized that the seven principles do not represent disparate agreements on seven different subjects, but instead fit together to form a single deal between the Trump Administration and the PJM Governors. I have no inside information about the negotiations, so I suppose I could be wrong. But whether or not I am right, I think the best way to understand the Agreement is to view it as a single unifying deal.

In order to understand this deal, it is first necessary to briefly understand where each side was coming from. From day 1, the Trump Administration has pursued the goal of trying to get more conventional generation capacity built. One of the 26 Executive Orders the President signed on his first day in office was an order declaring an energy emergency and requiring, among other things, the government to act to facilitate the generation of electricity.[3] Since then, the Administration has taken a number of steps to advance this goal, including issuing orders preventing the retirement of old coal and natural gas units and taking steps to try to jump start the construction of new nuclear power plants.[4] As I wrote in my last post, PJM failed to acquire all of the capacity it determined is needed to ensure the reliability of its system in its most recent capacity auction conducted this past December.[5] The Trump Administration undoubtedly believes that drastic action is needed to remedy PJM’s capacity deficiency.

At around the same time, rising retail electricity rates started to become a major political issue in many of the PJM states. Most notably, the price of electricity was a major issue in last year’s gubernatorial race in New Jersey. It also was an issue in the Virginia gubernatorial race. And Pennsylvania Governor Shapiro has made a big issue out of PJM’s electric rates in the past year.

The deal made through the Agreement’s seven principles involves a simple trade involving these two issues. The Trump Administration got the agreement of every governor in PJM not to oppose giving incentives for the construction of new generation facilities in PJM. In exchange, the PJM Governors got the Trump Administration’s agreement not to oppose provisions carefully circumscribing these incentives designed to protect residential customers. To be fair, the Trump Administration no doubt is sensitive to the issue of increases in the price of electricity, and the PJM Governors surely recognize the need to quickly construct new generation capacity to bolster the reliability of the grid. But I am confident I have accurately described the primary motivating factor for each side.

This deal is an important one for PJM. But is also could have repercussions that extend beyond PJM’s borders. The problems caused by the high electricity demands of new data centers is not limited to PJM, but extends to virtually every part of the US. If the Agreement’s principles can be translated to concrete action that actually works to address these problems, they could be used as a blueprint for the rest of the country.

Incentives for New Generation Capacity

The first principle in the Agreement (“Provide Revenue Certainty to New Generation”) represents the primary incentive included in the Agreement for the construction of new generation capacity. It calls for providing “15-year revenue certainty for new capacity resources . . . by holding a Reliability Backstop Auction for new capacity resources commencing no later than September 2026.” (I am going to call this auction the Emergency Auction, which is what DOE called it in its press release about the Agreement). I had to smile when I saw this principle. As I have been saying for some time now, long term revenue certainty is essential to encourage the construction of the expensive new generation required to meet growing demand.[6] Investments in the construction of new generation facilities are long-term investments. Very few generation companies or lenders are going to be willing to commit the large amounts of capital needed for new generation capacity when they have no idea what revenues that capacity will earn after the first year of operations. I am glad to see that the Trump Administration agrees.

Although not as important as the Revenue Certainty principle, one other principle included in the Agreement also supports the goal of facilitating the construction of new generation capacity. The fifth principle (“Accelerate Ongoing Generator Interconnection Studies”) addresses PJM’s process for interconnecting generation facilities to the grid. As I explained in more detail in a prior post,[7] the little known and little understood requirement for generation facilities to obtain approvals before interconnecting with the grid represents a significant barrier to bringing new generation capacity online. The interconnection study process can take several years and interconnections can cost generation owners hundreds of millions of dollars.

The Generator Interconnection principle addresses this problem by providing that PJM should give interconnection priority to proposed new generation facilities that clear the Emergency Auction. Hopefully, by prioritizing the interconnection process for new generation facilities selected in the Emergency Auction, those facilities will be able to interconnect with the grid and deliver their electricity when their construction has been completed. I think this is an important step that needs to be taken to bring needed new capacity online on a timely basis. And as is the case with 15 year revenue certainty, granting interconnection priority to new capacity constructed for resource adequacy purposes is something I have previously proposed.[8]

Protecting Residential Customers From Price Increases

Several of the Agreement’s principles, including the first principle, are intended to protect residential customers from price increases. A bit of background is necessary to explain how this is accomplished. Starting in 2004, PJM has conducted annual capacity auctions in which it acquires one-year capacity contracts, with delivery starting three years after the auction. These auctions are called “Base Residual Auctions.” I’ve described how the auctions work in some detail in my previous posts, including in my most recent post,[9] so will not repeat that description. The important point here is that those auctions are open both to existing generation facilities and to proposed new facilities not yet in operation.

The simplest way to incorporate long-term revenue certainty into the PJM capacity auctions would have been to revise the Base Residual Auctions to provide for the necessary revenue certainty. But the Revenue Certainty principle of the Agreement does not do this. Instead, it calls for PJM to conduct a single separate Emergency Auction, no later than this September, to provide revenue certainty for new generation facilities. It is because the revenue certainty principle applies only to this one separate Emergency Auction that the other principles included in the Agreement are able to limit the residential price effect of the incentives provided to encourage new construction.

The first such limitation appears in the Agreement’s second principle (“Protect Residential Customers from Price Increases”). This principle provides that PJM’s existing capacity price caps would continue to apply to the next two Base Residual Auctions. The first of these is scheduled to be conducted this May and a second auction is scheduled for December of 2026. Importantly, the Protect Residential Customers principle applies only to those two Base Residual Auctions; it would not apply the price cap to the separate Emergency Auction. Clearly, the expectation is that not only will the Emergency Auction provide winning generation facilities with price certainty, but also that the capacity price established in that auction would exceed the price cap.

After protecting residential customers from price increases from the next two Base Residual Auctions, the Agreement’s third principle (“Allocate Costs to Data Centers”) then works to also prevent the presumably higher capacity prices established in the Emergency Auction from increasing residential customer rates. That principle attempts to achieve this by calling for all capacity costs resulting from the Emergency Auction to be allocated to the utilities serving new data centers. This is done because these new data centers presumably are responsible for the increase in demand necessitating the Emergency Auction. Any remaining costs are to be allocated to utilities based on “their remaining net short positions.”

Finally, the seventh principle (“Use [State] Authorities to Allocate Costs to Data Centers and Protect Residential Customers” ) calls for the states to require their utilities to allocate the cost of the capacity procured in the Emergency Auction to new data centers. This principle does not involve the federal government but rather represents an agreement among the governors of the various PJM states. I’m not sure why there is a need for the PJM Governors to agree among themselves to protect their residential customers this way. Allocating increased costs to data centers instead of residential customers presumably would be politically popular. Perhaps there was a concern that some states would try to attract new data centers by insulating them from the costs of the Emergency Auction. If so, I am not sure how this principle would stop that from happening. The principle is not legally binding, and no state has the ability to contest how another state allocates costs among different customer classes.

Evaluation of the Deal

So that’s the deal. To incentivize the construction of new generation capacity, the Trump administration gets a single uncapped capacity auction that provides the long-term revenue certainty needed to support construction financing. To protect residential customers, the PJM Governors get a guarantee that the current price caps will be applied to the next two Base Residual Auctions and also get protection for residential customers from the uncapped Emergency Auction. How well does the Agreement implement this deal? I have a few observations.

First of all, it is important to keep in mind that neither the Trump Administration nor the PJM Governors have any direct authority to make any of it happen. Instead, PJM must first make the necessary changes to its tariffs, and then those tariffs must be approved as just and reasonable by the Federal Energy Regulatory Commission (FERC). I could fill several posts describing how difficult and time-consuming this could be. Instead, let’s just say that the agreement between the Trump Administration and the PJM Governors to “urge PJM expeditiously to file tariff changes at FERC” to implement the principles in the Agreement does not mean that those principles will be implemented exactly as written, if at all. I know, however, that the “urging” part of the Agreement has already had an effect. PJM quickly started the stakeholder process needed to change its tariffs one week after the Agreement was made public. But there is a long way to go before any tariff changes are finalized and FERC approval is obtained.

Putting aside that minor boring technical issue, I am not sure that the end result of the Agreement will be exactly what the parties intend even if PJM and FERC cooperate. Let’s start with the question of how well the residential customer protections will work. I can see a few potential problems. These problems stem from the possibility that any new generation capacity currently being contemplated by developers would be offered into the Emergency Auction instead of the May 2026 Base Residual Auction (May Auction). This could happen because developers of new capacity might prefer to offer their capacity in an Emergency Auction that not only is expected to yield higher prices but also will provide 15 years of revenue certainty. If that happens, it could have two unanticipated consequences for the May Auction.

First, basic economics tells us that lowering the amount of supply in a market translates into a higher market price. If owners of proposed new generation facilities skip the May Auction and instead offer into the Emergency Auction, the potential result of this reduction in the amount of capacity supplied in the May Auction could be a higher capacity price than would have resulted in the absence of the Emergency Auction. It is true that the price cap applicable to the May Auction could very well limit any such price increase. But there is a non-zero possibility that the May Auction capacity price would fall below the cap if all contemplated new generation facilities were offered into that auction instead of in the Emergency Auction. And those higher capacity prices would affect the price of electricity paid by residential customers.

Second, if developers of new capacity decide to offer their capacity into the Emergency Auction instead of the May Auction, the likely result is that the May Auction would also fall short of PJM’s target, perhaps by even more than the capacity shortfall in PJM’s December auction last year. If so, conducting the Emergency Auction could have the counterintuitive result of increasing the amount of generation capacity PJM needs to acquire in that auction.

The extent to which the Emergency Auction will help to advance the Trump Administration’s goal of incentivizing the construction of enough new capacity is more difficult to assess. This is because there are ambiguities in the Agreement that leave important unanswered questions about that auction.

Let’s start with perhaps the most basic ambiguity. How much generation capacity is the Emergency Auction supposed to acquire? The Agreement is silent on this point. Is it the 6,000 megawatt shortfall that PJM experienced in last December’s Base Residual Auction? If so, will any additional shortfall experienced in the upcoming May Auction be included? Or will the amount procured be determined in an entirely different manner? This is an important question because the only auction contemplated by the Agreement to provide for long-term rate certainty is the Emergency Auction.

A related ambiguity is how far in the future is PJM supposed to look for the new capacity to come online? A problem with PJM’s Base Residual Auctions is that capacity selected in those auctions is supposed be available when the delivery period commences in three years.[10] It is possible for a generation facility to receive an extension if it is not ready to operate, but to participate in an auction it must demonstrate a reasonable schedule that shows commercial operations commencing in three years. This is barely a sufficient amount of time for relatively small and inefficient natural gas combustion turbines to be constructed and placed in service.

But PJM needs larger new baseload units, in addition to combustion turbines, to meet the increasing demands for electricity of data centers and to replace aging generation facilities slated for retirement. And these larger units can take considerably longer than three years to construct. As PJM’s Interim President and CEO noted at the first stakeholder meeting conducted on January 22 to begin the process of implementing the Agreement, “new generation procured in any type of auction probably would not be operational until 2032 or 2033”[11]

These longer construction times would not be a problem if the Emergency Auction looks beyond the current three-year window to predict when and how much new capacity will be required over a longer period, perhaps 10 years. This would allow developers of larger generation facilities to participate in the auction and come online at a time consistent with reasonable construction times. The ability of larger-sized facilities to participate in the Emergency Auction would be especially valuable to the owners of those facilities. The 15 years of revenue certainty provided by that auction—as opposed to one year of revenue certainty currently provided in the Base Residual Auction—is more important for obtaining financing for these more expensive large facilities.

Unfortunately, basing the Emergency Auction on such an extended forecast of demand potentially could conflict with the Agreement’s fourth principle (“Improve Load Forecasting”). This principle states that PJM should work to improve its electric demand forecasting to ensure that capacity is acquired only “for real and verified demand.” This would be done by only including demands in the forecast that can demonstrate “a significant financial commitment.” Application of this requirement to the Emergency Auction would severely hamstring any effort to use that auction to procure capacity provided by large generation facilities. Construction of those facilities could not be completed until well after any data center would be willing to enter into the kind of significant financial commitment contemplated by the Agreement. Perhaps, however, the Load Forecasting principle should be read to apply only to the Base Residual Auctions, where requiring the use of significant financial commitments to forecast the demand to be served in the auction would make more practical sense.

What’s Next?

Let me end this post by discussing the Agreement’s sixth principle (“Return PJM to Market Fundamentals”). This principle provides that PJM should revise its Base Residual Auction for capacity “to ensure long-term viability and prevent consumers from bearing excessive ongoing costs.” The “expectation” is that the new capacity auction rules would be in place for next year’s Base Residual Auction, scheduled for May of 2027. I read this principle as having two purposes. The first is apparent from the text. Neither the Trump Administration nor the PJM Governors are happy with PJM’s current rules for the Base Residual Auctions, and both agree that PJM should make major changes. The second purpose can be gleaned from the fact that the principle provides nothing other than vague general guidance as to how the Base Residual Auctions should be revised. I take this to mean that the Agreement represents nothing more than a proposed temporary solution to the problem raised by the recent rapid and unexpected growth in demand to serve data centers. For now, at least, there is no agreement between the Trump Administration and the PJM Governors on a long-term solution to the problem.

We live in interesting times. And it will be interesting to see how both the Agreement and the attempts to arrive at a permanent solution to the problem of meeting increased data center demand will play out, not only in PJM, but also in the rest of the country.

***

I hope you enjoyed this post. I enjoy writing them and will never charge for subscriptions or ask for donations. I only ask that, if you did enjoy it, you press the “like” button below. And please consider sharing my posts with other people you think might be interested. Doing so will help me evaluate interest in the book I am writing on grid operations. Of course, if you have a reaction to, or question about, this post, please leave a comment and I will be happy to respond.

[1] The Trump Administration was acting through the White House National Energy Dominance Council.

[2] The Agreement can be accessed at PJM SoP 011526 Secretary Wright and Burgum Signed (2026-01-15 1453 ET)-DOI to DOE 9 Signed.pdf.

[3] Executive Order Declaring a National Energy Emergency (Jan. 21, 2025); Declaring a National Energy Emergency – The White House.

[4] I discussed these initiatives in two of my posts: Is Bringing Generation Facilities Back From the Grave a Good Approach to Grid Reliability? (June 8, 2025); and Will the New Nuclear Executive Orders Jumpstart the Construction of New Nuclear Reactors?(June 1, 2025).

[5] PJM Failed to Acquire the Capacity it Needs to Ensure Reliability. What Does it Mean? (January 18, 2026).

[6] See for example, my post A Closer Examination of the Effects of the Explosion in the Demand For Electricity Caused by Data Centers (Part Three), (Sept. 1, 2025).

[7] The Little-Known Bottleneck Slowing Down the Addition of Generation Capacity to the Grid, (April 15, 2025).

[8] A Closer Examination of the Effects of the Explosion in the Demand For Electricity Caused by Data Centers (Part Three), (Sept. 1, 2025).

[9] PJM Failed to Acquire the Capacity it Needs to Ensure Reliability. What Does it Mean? (January 18, 2026).

[10] This period has been even shorter in recent auctions as PJM was required to make up for the four year gap between 2018 and 2022 in which FERC failed to authorize new auctions.

[11] PJM Begins Discussion on Creating a Backstop Auction, (Jan. 28, 2026); PJM Begins Discussion on Creating a Backstop Auction | PJM Inside Lines.

This past year, something happened that has never happened before. Well, actually many things happened last year that had never happened before. But I am talking about an important unprecedented event on the grid. Specifically, the capacity auction for the PJM Regional Transmission Organization (RTO) failed to acquire the amount of capacity PJM determined was necessary to meet its one-event-in-10 year reliability standard.[1] PJM’s capacity market is the world’s largest electric generation capacity market, supplying capacity across a 13 state region spanning from New Jersey, Pennsylvania, Delaware, Maryland, and Virginia, and westward as far as Illinois. Consequently, events that affect PJM can have wide-ranging effects.

Never before has PJM, or any other RTO that conducts capacity auctions, failed to acquire enough capacity in its auction to meet its reliability standard. Let me see if I can explain what this means, as well as its implications. Along the way, you can learn something about how the RTO capacity auctions work.

Let’s start out with a quick review of the capacity product that PJM purchases in its auction.[2] Capacity is not a physical product, but rather a contractual obligation to offer to sell electricity into the PJM energy market. For those of you familiar with financial markets, the Supreme Court has characterized capacity as being like an options contract, which gives PJM the option, but not the obligation, to purchase electricity from a generation owner in PJM’s energy market. The energy market is the physical market in which all electricity generated in the 13-state PJM region is bought and sold. And it is from this market that all electricity consumed in that 13-state region is supplied.[3] PJM attempts to acquire enough capacity in its annual auctions to ensure there is always enough electricity offered into its energy market to serve all customers, even when electricity demand is at its highest. The rolling blackouts in Texas during winter storm Uri, and later during summer heat waves, show what can happen when you don’t have a capacity market, which the RTO in Texas does not have.

The purpose of the capacity auctions that PJM and other RTOs use to acquire capacity is two-fold. First, the auction is intended to send price signals to generation owners. When there is an oversupply of generation, prices are low, signaling that older inefficient generation facilities should be retired. When supplies tighten, prices increase, signaling that there is a need to construct new generation facilities.

Second, the auction is intended to provide the “missing money” to keep needed existing generation facilities in operation. Initially, the RTO wholesale markets relied solely on energy markets to provide the price signals for new generation to be constructed and for existing generation to either continue operating or retire. However, it soon appeared to many that the energy markets were not providing the revenues necessary to support an adequate amount of generation capacity. The solution was to add a new market in which RTOs could acquire capacity. This was somewhat controversial because it deviates from pure economic theory. Nevertheless, today all RTOs, except for the one in Texas, either have capacity markets or require utilities serving customers to own their own generation capacity or else enter into contracts to acquire capacity outside of the RTO market.

In its auction, PJM acquires contracts to supply one year of capacity. The auction is supposed to be conducted three years in advance of the year the commitment commences, which allows time for new generation facilities to be constructed after signing a capacity contract. However, due to a gigantic screw up at FERC that’s not really relevant to this post (I was working there at the time but was conflicted out of participation so don’t blame me), PJM fell behind schedule in conducting its auctions.[4] The most recent one was conducted about 1.5 years before the capacity commitment commences on June 1, 2027.

Let me return to three versions of my old, frequently used but not very good graphic that I created, the “stack,” to try to explain how PJM’s capacity auction works. This first one will be familiar to those of my long-time readers, but I will explain it to those of you who haven’t been subjected to it before.

As you can see, the RTO has concluded it needs to acquire 1,000 MW of capacity in its auction to ensure there will be enough capacity to always provide reliable service. Offers to provide capacity have been submitted by five generation owners, each at a different price. The RTO stacks these offers, placing the lowest offer ($20 from the yellow generator) on the bottom, then the next lowest ($25 from the green generator), and so on, with the highest offer ($75 from the blue generator) on top. The RTO awards capacity contracts, starting with the lowest offer, and so on up the stack to the last offer needed to supply 1,000 MW of capacity, which is considered to be the marginal market clearing offer. Here, this is the $50 offer from the red generator. Each generation owner is paid the $50 price offered by the marginal market clearing offer.[5]

One concern that people have had with the RTO capacity markets is that generation owners might have market power that would allow them to cause auction prices to be set above competitive levels. How this could be accomplished is a complicated subject I won’t go into today. But one way to limit market power is to impose a cap on the price that the RTO will pay for capacity. These caps are based on the cost that would be incurred to construct a hypothetical new generation facility. People who work with the grid love acronyms, and they say this price cap is set at the “cost of new entry,” or “CONE.” Don’t worry, I have declared this Substack to be a low acronym zone that uses only a few acronyms common in the electric industry, like FERC or RTO. I am not going to use CONE again.

Determination of this hypothetical new generation facility requires the use of many somewhat controversial decisions that can have a huge impact on the price cap. The most important is the type of new generation facility that is assumed to be constructed. The cost of a new large nuclear reactor—the most recent of which cost about $15 billion—is much higher than the cost of a small combustion turbine. If the cost of a new nuclear reactor were chosen for the cost of new entry calculation, the resulting price cap would be very high. For most of their history, however, RTOs have chosen inexpensive combustion turbines and applied other assumptions that tended to lower the price cap. These decisions were controversial and heavily litigated.

Ultimately, however, it didn’t really matter what type of generation facility was assumed because the marginal market clearing price in the RTO capacity auctions has almost always been well below the price cap. The effect, or rather lack of effect, of a price cap well above the market clearing price can be shown on a second version of the stack, shown below.

As you can see, the price cap of $75 is well above the marginal market clearing offer of $50, and therefore has no material effect on the auction price, although perhaps the price cap had a dampening effect on the offers that some of the generation owners submitted.

What happens when the marginal market-clearing price is over the price cap? This third version of the stack illustrates the effects.

As you can see, the amount of capacity the RTO is seeking to acquire (1,000 MW) has not changed, nor have the offers into the capacity auction. But now, the RTO has applied a $30 price cap. Because the offers have not changed, the marginal market-clearing offer ($50) has not changed. But because the price cap is $30, the price paid for capacity is only $30. This means that the owner submitting the $50 offer does not receive a capacity award because that owner’s offer is above the capacity price established in the auction. Consequently, the RTO is able to acquire only 750 MW of capacity instead of the 1,000 MW it determined was needed to ensure reliable service.

This is what happened in PJM’s December capacity auction. The marginal market clearing offer price was higher than the price cap, which was $333 per megawatt-day.[6] And PJM acquired 6,623 MW less capacity than it determined was needed. I see five important points to be made about the consequences of this result.

1. PJM’s Failure to Acquire All the Capacity it Determined is Needed Will Not Inevitably Result in Blackouts

First of all, people who live in the PJM region don’t need to rush out to buy home generators. Although PJM was unable to acquire all of the capacity that it said it needed to ensure reliability, this does not mean PJM will inevitably be subjected to blackouts. PJM was able to acquire significantly more capacity than it anticipates will be necessary to serve its maximum demand for the year. Instead, the shortfall affects PJM’s reserve margin, which is the amount of capacity PJM acquires above its projected peak demand. The reserve margin allows PJM to supply the peak demand even if some capacity is unavailable due to problems with equipment or for needed maintenance, and/or if demand is higher than expected.

PJM wanted to acquire enough capacity to achieve a 20 percent reserve margin. Although this did not happen, PJM still acquired enough capacity to have a 14.8 percent reserve margin. This is a healthy margin, and close to PJM’s target reserve margin in many previous auctions. I know in the past PJM has been criticized as using overly conservative assumptions for determining its needed reserve margin. And even if a 20 percent margin is needed to meet its one-event-in-10 year reliability standard, there is only a 10 percent chance that once in 10 years circumstances will occur in the year in which PJM failed to acquire enough capacity to achieve a 20 percent reserve margin.

2. Higher Capacity Auction Prices Should Result in Higher Electricity Prices in the PJM Region

Another consequence of PJM’s capacity auction is that the price established in that auction will cause an increase in electricity prices charged to retail customers in the PJM area. PJM allocates the cost of the capacity it acquires in the auction on a pro rata basis to each utility or other entity in the region that sells electricity at retail. For example, if a utility supplied electricity to 10 percent of the retail customers in PJM, it would be allocated 10 percent of the capacity costs incurred by PJM. Although state utility commissions set the utilities’ retail rates, the utility commissions are required by law to allow the utilities to pass through to their customers the costs of the capacity allocated to them by PJM. Consequently, the cost of the capacity purchased at PJM’s capacity auctions is ultimately reflected in customer retail bills in the year that the capacity is delivered.

The following graph (drawn automatically by Excel so it is nicer than the ones I do myself) shows the capacity prices established in all of PJM’s capacity auctions since the first auction in 2004. The dates shown on the graph are the dates the auctions were conducted. If more than one auction was held in a year then the first auction held that year is listed on the table as the A auction, and the second auction of the year is listed as the B auction. For example the first auction conducted in 2025 is listed as the 2025A auction and the second auction is listed as the 2025B auction (the 2025B auction is the subject of this post).

As you can see, capacity prices have increased dramatically in the last three auctions, and there also was a big jump between the 2024A and 2024B auctions. According to PJM, the total cost of capacity acquired in the 2024B auction is $14.7 billion, almost ten times as much as the cost of the capacity acquired in the 2024A auction (although, to be fair, the 2024A auction price is one of the lowest on record). The total cost of the 2025A auction is $16.1 billion, and the total cost of capacity in PJM’s most recent auction is $16.4 billion.

As I mentioned above, PJM’s capacity auctions establish prices for capacity delivered in future years. As a result, PJM currently is paying the $270 capacity price established in the 2024B auction. That price is almost $100 higher than the previous highest PJM capacity price and constitutes an important factor underlying the increased retail electricity prices currently being experienced in PJM. Starting in June of 2026 the capacity price will increase to $329, and in June of 2027 it will increase again to $344, which is the price established in the most recent auction. The inevitable result is that the current high retail electricity prices will continue to go up for at least the next few years.

3. The Price Cap Had the Unintended Consequence of Reducing the Amount of Capacity Acquired by PJM

The obvious purpose of price caps in RTO auctions is to limit the capacity prices determined in the auction and, consequently, limit the price of electricity paid by consumers. The current price cap applicable to PJM’s capacity auction was reduced last year from over $500 to $333 as a result of a settlement of a complaint brought by Pennsylvania Governor Shapiro. That settlement was hailed as potentially saving consumers billions of dollars. And in fact, this year the settlement price cap did save consumers billions of dollars; PJM calculated that without any price cap the marginal market clearing capacity price would have been $529, about 60 percent higher.

Keeping electricity prices as low as possible is certainly an important goal. But what no one mentioned, and what I have never seen in any discussion of RTO capacity auction price caps, is that the caps also have the effect of reducing the amount of capacity acquired in the auction below the amount the RTO considers to be necessary to ensure the reliability of its system. This was illustrated in the third version my stack graphic shown above, where the $30 price cap reduced the amount of capacity obtained in the auction from the 1,000 MW the RTO needed to 750 MW.

Just as the $30 price cap reduced the amount of capacity acquired in the third version of my hypothetical stack, the application of the $333 price cap in PJM’s auction reduced the amount of capacity acquired by PJM. PJM’s calculation of what the marginal capacity clearing price would have been absent the price cap also showed that PJM would have acquired almost 800 MW more capacity without the price cap. Perhaps that was a good result in this case. PJM’s acquisition of 800 MW of additional capacity in the auction would have raised the total cost of capacity acquired by PJM by almost 60%, which is close to $10 billion, a very high price to pay for a relatively small amount of additional capacity. But if capacity reserve margins continue to decline in the future, which is a distinct possibility, it is important to keep in mind that capacity price caps could prevent PJM (and other RTOs) from acquiring the capacity they need to provide reliable service.

4. AI and Green Energy Policies Both Share Responsibility

Some people place the blame for the current problems faced by the grid on artificial intelligence’s (AI) explosive demand for electricity. Others claim that the grid’s problems result from green energy policies favoring solar and wind generation facilities. Both theories are correct. AI and green energy policies have worked in tandem to create a problem much worse than either would have caused absent the other.

AI’s role is relatively straightforward and easy to understand. As I have noted in many of my recent Substacks, the grid’s reliability has been seriously threatened by the increase in demand for electricity to serve data centers that support AI. PJM has echoed this concern in its press release about the most recent capacity auction: “this auction leaves no doubt that data centers’ demand for electricity continues to far outstrip new supply, and the solution will require concerted action involving PJM, its stakeholders, state and federal partners, and the data center industry itself.” Without a doubt, PJM is correct. 5,100 MW of the projected 5,250 MW increase in demand in PJM is expected to come from new demand by data centers. Without the rapid increase in AI’s demand for electricity, there would be more than enough capacity to supply all of the grid’s remaining demand.

But why has AI’s demand for electricity far outstripped new supply and why isn’t there enough generation capacity available? Here we have to look at green energy policy over the last 10-15 years. Its role requires a bit more explanation.

To start off with, solar and renewable generation facilities do not harm grid reliability or cause price increases in and of themselves. These renewable generators can, depending on the circumstances, produce valuable low cost electricity in those peak demand hours when every megawatt-hour of electricity is important. The problem doesn’t result from those times when wind and solar do produce electricity, but instead from those hours when they do not produce electricity. Wind and solar generators are dependent on weather conditions and on sunlight, and these conditions are too uncertain for system operators to depend on the electricity they produce without any backup. This is especially true in the winter months when cold weather can cause sharp increases in demand in the early morning hours when little or no solar capacity can produce electricity and the wind may or may not be blowing with any strength.

The intermittent nature of wind and solar production do not disqualify these facilities from providing capacity. But the nature of wind and solar facilities does require that they be carefully integrated into electric systems, and backed up by conventional resources or by energy storage (battery) facilities that, to date, have not been constructed in great numbers in most of the RTOs, including PJM. Without any backup, the amount of capacity that wind and solar can provide is heavily discounted by PJM and other RTOs when acquiring the capacity they need to ensure system reliability. Otherwise, there is a risk that, when solar and wind facilities are needed to serve demand, they will not be able to do so.

For many years, green energy policy has ignored these limitations on the capacity value of wind and solar, and supported the construction of as many wind and solar facilities as possible regardless of the reliability benefits they provide. This would not be a problem if green energy policy had not, at the same time, chipped away at the conventional generation capacity that can be turned on and off when so directed and thus able to produce electricity whenever needed to meet demand. This was done through a trifecta of green energy policy positions: (1) opposition to the construction of new conventional generation capacity, primarily natural gas-fired facilities; (2) opposition to the construction of new natural gas pipelines needed to bring natural gas to new generation facilities; (3) support for the premature retirement of an ever-increasing number of conventional generators, mostly large coal facilities, but also some older natural gas facilities and nuclear power plants.

Green energy policy also has affected the economic viability of conventional generation capacity through subsidies granted to solar and wind facilities. The federal production tax credits and state renewable energy credits given to these facilities allow them to offer to sell into the energy markets at very low, even negative, prices. Similarly, state subsidies fixing prices that utilities must pay for wind and solar electricity allow those facilities to offer to sell capacity at low prices. In each case, the subsidies have artificially lowered the RTO energy and capacity market prices, thereby masking the price signals those market should have sent. These artificially low market prices have failed to provide the missing money needed by existing conventional generation facilities to avoid premature retirement and have also slowed the construction of new conventional generation facilities. This would not have been a problem if there had been relatively little demand growth. But thanks to AI, demand unexpectedly exploded.

Economic theory tells us what inevitably happens when you combine increased demand with little or no increase in supply. The result was a PJM capacity auction that yielded the highest capacity price in its history and at the same time failed to acquire enough capacity to satisfy PJM’s reliability standard. AI’s insatiable demand for electricity and green energy policy each share responsibility for this outcome.

5. The PJM Auction Result Highlights the Grid’s Emerging Reliability Crisis

A big problem in our society today is that we spend too much time trying to place blame for the problems we face rather than working to fix them. If you wake up at 2 am and find that your house is engulfed in flames, you shouldn’t spend time arguing over who is at fault; you need to get everyone out of the house and call the fire department. Once the fire is out, then you can try to figure out why the fire started.

Unfortunately, the recent PJM auction results represent one more sign that the grid could soon be engulfed in flames. As I have written in previous posts, in the past year we have already seen the following: (1) the North American Electric Reliability Corporation (NERC), the entity statutorily designated to ensure the reliability of the grid, predicted last May that there was an elevated risk of blackouts this past summer in large parts of the country because of a lack of generation capacity; (2) the Department of Energy issued a report in July finding that the risk of blackouts in 2030 would increase by a factor of 100 if no steps are taken to increase the grid’s generation capacity; and (3) in November NERC again found an elevated risk of blackouts in large regions of the country, this time in the winter, again because of potential capacity shortages.

Now PJM has failed to acquire the amount of capacity it determined is needed for reliability purposes. This failure may, or may not, result in blackouts. Nevertheless, the 6,623 MW shortfall in generation capacity acquired by PJM in its auction is a very concerning event. Perhaps even more concerning is the fact that, even without a price cap, PJM’s auction would have resulted in the acquisition of less than 800 MW of the 6,623 MW needed to ensure reliability.

The failure of PJM to acquire sufficient capacity in its auction, and the other events I have described, are sirens warning that danger is imminent. Hopefully, the high prices in the PJM capacity markets will send the price signals needed to forestall premature retirements and incentivize new construction. I have described other actions in detail in my September 2 post that also would address the need for new capacity, A Closer Examination of the Effects of the Explosion in the Demand For Electricity Caused by Data Centers (Part Three). Let’s stop pointing fingers at AI and renewables. It’s time to start putting the fire out.

[Note] Just as I was finishing writing this post, President Trump and a group of Northeastern governors announced that they had reached an agreement in principle as to how to address the need to meet AI’s demand for electricity in PJM. In my quick review of these principles, I see more questions than answers. I will discuss this agreement in my next post, which should drop in the middle of February.

***

I hope you enjoyed this post. I enjoy writing them and will never charge for subscriptions or ask for donations. I only ask that, if you did enjoy it, you press the “like” button below. And please consider sharing my posts with other people you think might be interested. Doing so will help me evaluate interest in the book I am writing on grid operations. Of course, if you have a reaction to, or question about, this post, please leave a comment and I will be happy to respond.

[1] PJM issued a news release that provides a description of its capacity auction results: “PJM Auction Procures 134,479 MW of Generation Resources,” as well as a much more detailed report; “2027/2028 Base Residual Auction Report.” The news release and detailed report can be found on PJM’s website, PJM.com.

[2] . If you want a more detailed explanation, you can look at one of my older posts: Explaining the Grid Part Five: RTO Capacity Markets plus a bit of history.

[3] Technically, not all electricity consumed in PJM comes from its energy market. Electricity generated by consumers who generate electricity for their own consumption, electricity generated from rooftop solar panels for example, does not have to be sold in PJM’s energy market.

[4] I described the delay in the auction in more detail in my February 9, 2025 post What’s Going on in the PJM Capacity Market?

[5] If you want to know why each owner gets $50, read my really old post from September of 2024: In Defense of RTOs, Part Three.

[6] A price per megawatt-day means the generation owner is paid that price for each day of the contract period. Since the PJM capacity contracts are for one year, this means that the generation owner is paid the price per megawatt-day times the 365 days of the year. So, for example, if the price per megawatt-day was $10, the generation owner would be paid $10 * 365, or $3,650 for each megawatt of capacity sold. The most recent PJM auction price of $333 translates to an annual payment of $33,344 for each megawatt of capacity sold.

The North American Electric Reliability Corporation (NERC) is the entity tasked by Congress with ensuring the reliability of the grid. NERC’s summer reliability assessment, issued last May, showed that four large regions, covering all or part of 23 states, were facing an “elevated” risk of blackouts during “extreme conditions” due to inadequate generation capacity.[1] This was big news, and was widely discussed in the media, including by me in my Substack. Luckily, however, the extreme conditions specified by NERC did not materialize and there were no widespread blackouts.

This past November NERC issued its reliability assessment for the upcoming winter.[2] This winter assessment again shows that some regions face an elevated risk of blackouts. But NERC’s conclusion that blackouts could be possible this winter has not been featured as prominently as NERC’s assessment for the summer. Perhaps it’s because, despite NERC’s warning, there were no blackouts this past summer, making NERC’s winter assessment seem less urgent. Whatever the reason, there are some interesting details in the winter assessment I think are worth sharing in this post.

First, I want to point out what I think is the most important and troubling point of the analysis. NERC notes that the total amount of the grid’s generation capacity has increased by 9.4 gigawatts since last winter. This sounds encouraging, given that the grid needs new capacity. However, NERC also notes that, at the same time, the peak demand for electricity has risen by 20 gigawatts.[3] In other words, demand for electricity in the winter has grown over twice as much as the supply of generation capacity. This trend is not sustainable. NERC’s forecast for this coming winter is that blackouts are possible only under extreme conditions. But if demand keeps rising twice as high as supply, then blackouts will be possible, or even likely, under normal conditions.

What does NERC mean by an “elevated risk” of blackouts? As NERC explains, when there is an elevated risk of a blackout in a region, that region should have adequate capacity under normal conditions. However, “resources will not be sufficient to meet operating reserves under extreme peak-day demand with normal resource scenarios (i.e., typical or expected outage and derates)”; or “under normal peak-day demand with reduced resources (i.e., extreme outage and derate scenarios).”[4] In other words, there is a potential for blackouts either if cold weather causes the demand for electricity to be much higher than expected or if the weather causes much higher than expected outages of generation facilities.

I should note that, at the time NERC issued its assessment, projections were that this winter would be slightly warmer than last year.[5] Today, however, the country is in the middle of a cold spell with lower than normal temperatures projected to last as long as a month. The extreme weather conditions that NERC believes could cause blackouts may not be as unlikely as NERC thought they were at the time it issued its assessment.

As I mentioned, NERC found an elevated risk of blackouts this winter for several regions of the US. However, the potential for blackouts affects fewer states and mostly different regions than the ones FERC determined were at an elevated risk last summer. The following map from NERC’s assessment shows where NERC thinks problems might arise this winter.

Unlike this past summer, NERC does not predict an elevated risk of blackouts in the MISO or SPP Regional Transmission Organizations (RTOs) this winter, which cover a significant swath of the middle of the country. Instead, the regions NERC believes could potentially be affected are located in the Southeast, in the Pacific Northwest and inland West, and in Texas. Of these regions, only Texas was identified as having an elevated risk of blackouts this past summer. In addition, NERC shows a potential problem in New England resulting from a “Risk to Natural Gas Pipeline Capacity.” The reasons NERC provided for concluding there is an elevated risk of blackouts vary from region to region. I think these reasons provide some useful insights into grid operations, so I will go through them briefly.

Texas-ERCOT. Even though the highest (peak) demand in Texas occurs in the summer, high demand from new data centers and other industrial customers is also increasing the demand for electricity in the winter. NERC notes that the winter peak demand periods in Texas occur before sunrise and after sunset. This is a problem because there is significant solar generation in Texas, but solar panels produce no electricity during these peak hours when the sun is down. NERC also notes that batteries have been constructed in Texas that help store electricity produced by solar generators during the day for use at night. However, the around-the-clock operations of data centers extend the length of the peak demand periods and, as a result, “maintaining sufficient battery state of charge will become increasingly challenging for extended periods of high loads, such as a severe multi-day storm like Winter Storm Uri.”[6]

New England. NERC determined that New England has an adequate amount of generation capacity for this winter. New England’s problem instead is a potential shortage of the fuel needed to supply that capacity. New England relies heavily on natural gas-fired generation and, as NERC notes, there is only limited natural gas pipeline capacity available to transport gas into New England. On extremely cold winter days, much of the pipeline capacity is needed to transport gas used for residential heating purposes. On these days, sometimes there is not enough pipeline capacity left to supply all the gas needed to fuel natural gas-fired generation facilities in the region.

To address this problem, many natural gas generation facilities in New England are equipped to use oil as fuel when natural gas is not available. New England’s reliance on oil-fired generation to meet its winter peak demand is ironic. Oil is much dirtier than natural gas, emitting significantly more carbon dioxide, sulfur dioxide, and nitrogen dioxide. Nevertheless, large amounts of oil must be used to generate electricity in the winter in New England because environmental opposition has killed all proposed new gas pipelines in the region.

NERC is focused on reliability, not the environment. Consequently, in its winter assessment, NERC found that the use of oil as a backup fuel should be adequate to prevent blackouts this coming winter. NERC notes, however, that on very cold days when demand for electricity is highest, the amount of oil stored by generation facilities can be rapidly depleted.[7] This could be a potential problem when there is an extended cold snap, and it is not possible to refill the facilities’ storage tanks with more oil.

SERC East (South Carolina and portions of neighboring states). NERC explains that the potential problem in SERC East is similar to that in Texas. As in Texas, there is no solar power output to meet high demand for electricity in SERC East in the early morning hours.[8]

Remaining Regions. The remaining regions identified by NERC as having an elevated risk of blackouts all have a similar potential problem. NERC’s analysis shows that each of these regions might not have sufficient capacity to supply extremely high winter demands on cold days. In that event, they would be dependent on imports of electricity from neighboring regions. There may not be sufficient amounts of electricity available in the neighboring regions, however, if those regions likewise are facing extreme cold weather conditions.

There are two other aspects of NERC’s winter assessment that I found interesting. First, NERC noted that ERCOT and MISO have implemented new methods for determining the capacity contributions of wind turbines, resulting in a significant reduction (14.2 gigawatts) in the amount of capacity provided by wind turbines in the winter.[9] Wind turbines can provide significant amounts of low cost electricity, but it is much more difficult to forecast when wind generators can produce electricity than it is for solar power. This reduction in the calculation of wind capacity highlights the problem of relying on wind to supply electricity during peak demand periods when wind turbines are not paired with batteries.

Second, of the 9.4 gigawatts of new generation capacity NERC determined was added to the grid’s generation capacity since last winter, only 3.2 gigawatts was thermal or hydroelectric capacity; the remainder was solar power, batteries, and demand response.[10] It is not my intention to turn this post into a discussion of the relative merits of different kinds of generation facilities, but in my view the NERC winter assessment shows that these other types of capacity are not as well-suited as conventional generation capacity for supplying peak demand in the winter:

(1) Solar power is useful for meeting peak demand for electricity in the summer months because solar panels produce the most electricity on the hot days where demand is highest. However, NERC found that winter peak demand occurs when the sun is down and solar panels produce no electricity in those hours;

(2) Demand response, in which consumers agree to cut their use of electricity when so directed, is less likely to be effective in extreme cold conditions. Some industrial consumers may be willing to cut their use of electricity during these conditions by shutting down their operations, but residential consumers may be reluctant to do so; and

(3) Batteries are effective, but as NERC noted, it may be difficult to keep batteries charged during extended cold snaps when all electricity produced is needed to supply consumer demand and there is little excess electricity production that could be used to charge batteries.

I think that the grid needs more conventional thermal, hydroelectric and nuclear capacity, and cannot rely solely on renewables, batteries, and demand response to meet the explosion in demand the grid currently faces, especially in the winter.

What is the main conclusion to be learned from NERC’s Winter Reliability Assessment? I think it is that, even if it turns out all demand for electricity is met this winter, that would not mean there is no winter resource adequacy problem for the future. The analysis provided by NERC shows both that the grid will be living on a knife’s edge this winter and that increases in demand for electricity continues to outstrip the increase in generation capacity. NERC’s assessment confirms the urgent need to rapidly expand the grid’s generation capacity.

In accordance with my current monthly publication schedule while I work on my book, my next post will come out some time in the middle of January. I hope that, in the interim, you all have a great Christmas and a good new year.

***

I hope you enjoyed this post. I enjoy writing them and will never charge for subscriptions or ask for donations. I only ask that, if you did enjoy it, you press the “like” button below. And please consider sharing my posts with other people you think might be interested. Doing so will help me evaluate interest in the book I am writing on grid operations. Of course, if you have a reaction to, or question about, this post, please leave a comment and I will be happy to respond.

[1] NERC, 2025 Summer Reliability Assessment (May 2025).

[2] NERC, 2025-2026 Winter Reliability Assessment (November 2025).

[3] Winter Assessment at 5.

[4] Winter Assessment at 11.

[5] Winter Assessment at 10.

[6] Winter Assessment at 6.

[7] Winter Assessment at 5.

[8] Winter Assessment at 6.

[9] Winter Assessment at 8.

[10] Winter Assessment at 8.

Those of you who have subscribed to or follow Explaining the Grid for at least the last few months know I have written several articles about how the explosive growth in the demand for electricity to serve AI has affected the grid. Up to now I have focused on the effect of AI on the reliability of the grid. As I have explained, if significant new conventional generation capacity is not constructed in the near future, the grid will not have enough generation capacity to supply all demand for electricity on hot summer days. The result will be persistent rolling blackouts on those days.

In this post I am going to address a different AI issue: the effect of AI on the price of electricity. In my view, this issue is not as important as the issue of whether the grid will be able to meet demand, but it is receiving more national attention. Perhaps this is because prices for electricity have been rapidly increasing since 2024, and these price increases became a major political issue in the recent New Jersey governor’s election. Many theories have been suggested for the cause of the price increases, and the confluence of electricity price increases and AI’s exploding demand has caused some people to place the blame for the price increases on AI’s doorstep.

Recently, a joint study by the Lawrence Berkeley National Laboratory and the Brattle Group (Factors Influencing Recent Trends in Retail Electricity Prices in the United States: What do we know? Where are the gaps?) provided a detailed analysis of the causes of the price increases. Accompanying the study was a set of slides that summarized this analysis.[1] The study did not specifically address the effect of AI on electric rates. However, National Lab/Brattle did analyze the effects of increased demand growth on rates in general. This analysis implicitly addressed the effects of increased demand associated with AI since otherwise there would have been no compelling reason to do the analysis.

The results of this analysis are summarized on slide 35 of the National Lab/Brattle slides. It shows that, contrary to the claim that increased demand from AI is causing electric price increases, over the last five years the states with the highest demand growth actually saw a decline in their electric prices. By contrast, states whose demand declined often saw increases in the price of electricity.

On its face, this is counterintuitive. Why did it happen? Think of it this way. Suppose you agree with two other families to share a beach house and split the cost of the house equally. You find a house everyone likes. It rents for $1,200. You notice, however, that there are more rooms than you need, and you could fit in a fourth family. Even though more people would be staying in the house, the rental fee does not change. As a result, adding a fourth family would reduce the price each family has to pay from $400 ($1,200/3) to $300 ($1,200/4). As this example illustrates, sharing fixed costs among more people means, as a matter of math, each person’s share of the costs is less.

Similarly, up to now, the grid has been able to supply electricity to the AI data centers without needing to construct much new infrastructure. As a result, just as it was possible to add a fourth family into our beach house and reduce the cost each family pays for the house, increasing the demand for electricity allows the fixed costs of existing infrastructure to be shared among a greater amount of electricity sales. And as a matter of math, sharing a set amount of fixed costs over a greater amount of electricity sales reduces the price per megawatt of the electricity being sold.

Importantly, the National Lab/Brattle study is based simply on an analysis of electricity prices from 2019-2024. The study’s conclusion about the effect of demand growth on electricity prices is not meant to describe a universal relationship between demand and electricity prices that has existed and will continue to exist in the future. Instead, slide 35 explicitly states that “this relationship need not always exist.” Demand growth can cause electricity prices to increase in the future “if new supply and delivery infrastructure is constrained and costly.”

What does this mean? Let’s go back to our beach house for an illustration. The four families had such a great time together at the beach that they decide to invite another family to join them the next year. However, the old beach house cannot accommodate all five families. They decide to rent two houses next to each other, and because costs have gone up, the combined rent of the two houses is $2,500. As a result, even though the total rental cost is being divided among more families, the total cost has also increased. This means that each family’s share has increased from $300 to $500 ($2,500/5). Similarly, if the grid is required to construct expensive new generation and transmission facilities to serve increased AI demand for electricity, that could increase the price of electricity if the increase in demand does not fully offset the increase in cost.

And it seems likely that expensive new infrastructure will be required to meet AI’s rapidly increasing demand in the future. Today, the capacity supply situation in many regions of the US is very tight. In fact, the North American Electric Reliability Corporation (NERC) concluded in May of 2025 that four large regions, covering all or part of 23 states, were facing an “elevated” risk of blackouts during “extreme conditions.” [2] These conditions did not occur and, as a result, there were no blackouts resulting from a shortage of generation capacity. But there is a serious question whether the grid will be able to continue to supply all demand for electricity in the future using only the grid’s existing infrastructure and currently planned additions. In July, the Department of Energy (DOE) issued a report analyzing the potential for blackouts in 2030 under current plans. DOE found that, given the increase in demand to serve AI, the risk of blackouts in 2030 would increase by a factor of 100 if no steps are taken to increase the grid’s generation capacity.[3]

This means, as I have written in several previous articles, that extensive new investment in infrastructure will be needed to meet all demands for electricity as those demands increase in the future. Much of this will need to be investment in new generation facilities, but there will also need to be investment in transmission facilities. It is not clear whether the resulting increase in the fixed costs of the grid’s infrastructure will be offset by the increased demand for electricity. But it appears to me it is likely the price increases resulting from the need to recover the fixed costs of new facilities will out strip the price decreases resulting from sharing those costs among a greater amount of sales. This is especially the case given that increases in the cost of transmission facilities are greatly outpacing inflation, as shown on slide 29 of the National Lab/Brattle slides.

There also is another way increased demand could cause the price of electricity to increase in the future that was not mentioned in the National Lab/Brattle slides. To illustrate this, I am going back to my favorite graphic (which is my favorite not only because I created it but because it helps to explain a lot about electric markets). I call this graphic THE STACK, which made its debut in Explaining the Grid Part Three—Market Products, and then made a repeat appearance in Explaining the Grid Part Four—RTO Energy Markets. You can pass over this brief discussion if you are familiar with THE STACK. But Substack is not like a book where everyone has read all the previous chapters, and I am going to do this again for those of my newer readers who haven’t read those relatively old posts (but if you are interested in how the grid works, you should read those posts).

If you already read either of my previous posts where THE STACK made an appearance, then you know how this works. To recap, all offers to sell electricity in an RTO wholesale market are stacked, from lowest cost offer to highest cost offer. The RTO then determines how much customer demand exists—either in real time for energy markets, or projected for the year in capacity markets, and then compares the total amount of customer demand with THE STACK. The highest priced offer needed to serve the last megawatt of customer demand—known as the marginal offer or the market-clearing offer—and all lower priced offers, receive awards to sell electricity. And every winning generator is paid the market-clearing price even if it offered to sell at a lower price. If you want to know why everyone gets that price, read Explaining the Grid Part Four—RTO Energy Markets.

If you look at THE STACK, you can probably figure out quickly how an increase in demand can increase the price of electricity. If an increase in demand results from AI’s need for large amounts of electricity, the total demand can easily move up the stack to the next highest cost offer to sell electricity, as shown by the following version of THE STACK:

In this version of THE STACK, demand has increased by 150 MW, which has moved the price of the marginal market-clearing price from $50 to $75. And this means the entire 1,000 MW of electricity is sold at a price of $75/MW, a $25/MW increase in price caused by AI’s demand for electricity.

The actual setting of prices in RTO markets is much more complicated because there are hundreds, if not thousands, of offers submitted into those markets, and in the energy markets prices are calculated as often as every five minutes. And because there are so many more offers being submitted, the difference in prices between offers is usually much less than shown in my STACK graphic. But the general principle is the same. An increase in demand for electricity will cause an increase in the wholesale price of electricity, at least in the short run. And because retail electricity prices are based on the wholesale price (plus additional costs), an increase in wholesale prices will cause a corresponding increase in the retail price of electricity paid by consumers.

This does not necessarily mean that the increased demand for electricity caused by AI will cause prices in the RTO wholesale markets to increase in the longer term. As demand for electricity in an RTO increases, so will the amount of generation capacity (hopefully). And if the new generation capacity is able to offer its electricity into the market at a relatively low price, that lower price offer could offset the price effects of the increase in demand, or even allow a decrease in the RTO market price. This is illustrated by the third and last of my STACK graphics. In this graphic, the 150 MW increase in demand is accompanied by an offer from a newly constructed low cost generation facility (shown in green on THE STACK) at a price of $25. This new offer moves the higher priced offers up in the stack so that the marginal offer once again comes from the facility offering to sell at $50. As a result, the 150 MW increase in demand was offset by the new low cost generation facility, and the increase in demand had no effect on the market-clearing offer price, returning the market price to $50, where it was before AI caused the demand for electricity to increase.

It is impossible to know today whether and when newly constructed generation facilities might offset increased AI demand. But for the near future at least, it appears likely that demand will increase faster than supply. If that happens, the likely result will be an increase in the price of electricity.

In accordance with my new publishing schedule intended to give me more time to work on my book, my next Substack post should come out some time in the middle of December. In the meantime, I hope that all of you have a great Thanksgiving.

***

I hope you enjoyed this post. I enjoy writing them and will never charge for subscriptions or ask for donations. I only ask that, if you did enjoy it, you press the “like” button below. And please consider sharing my posts with other people you think might be interested. Doing so will help me evaluate interest in the book I am writing on grid operations. Of course, if you have a reaction to, or question about, this post, please leave a comment and I will be happy to respond.

[1] The study, the slides, and other related documents can be found at: https://eta-publications.lbl.gov/publications/factors-influencing-recent-trends.

[2] NERC, 2025 Summer Reliability Assessment (May 14, 2025).

[3] DOE, Evaluating the Reliability and Security of the United States Electric Grid (July 7, 2025).

I recently realized that, on my new approximate monthly publication schedule, my next Substack should have been posted around the beginning of October. For some reason, I was thinking it should be around the beginning of November. Sorry, my bad. I will try to post the next one a bit before Thanksgiving, which should be easier for me to remember.

In my last post I presented data on prices charged for electricity in different states. This data suggested that a high renewable energy penetration in a state did not inevitably lead to high electricity prices to consumers. I was inspired to do that post by repeatedly reading assertions that “of course renewables cause high rates for electricity. Just look at California.”

Although my last post showed that high renewable penetration does not invariably lead to high electric prices, there is no denying the fact that California does have higher electricity prices than any other state. This can be seen on the chart below, based on data in EIA’s energy data browser. This chart shows where California’s rates for electricity stood relative to the other three highest priced states in 2001-2003 and in 2022-2024. To make it easier to see California’s electricity prices, I have listed those prices on the chart above the bars representing the price of electricity in California. As always, please forgive my less than expert graphics ability.

Although renewables don’t inevitably cause high electricity prices, that does not mean that they can never cause high prices. The fact California’s prices are so high made me wonder whether renewables have caused those high prices even though that doesn’t happen in all other states. I decided that for this post I would try to answer this question by taking a deeper look at California to see what causes its rates to be so high.

Let me start by observing that the question of what determines prices for electricity paid by consumers in California (or in any state) can be divided into two distinct questions: (1) what does it cost to generate electricity; and (2) what goes into the prices paid by customers in addition to the cost to generate electricity? In order to understand why electricity prices are high or low in California (or any state), you need to look at these two questions separately.

I look at wholesale prices of electricity to evaluate this first question. Wholesale prices reflect the purest (but not totally pure) indication of the cost of electricity. This is especially true in the Regional Transmission Organization (RTO) markets where most of the wholesale electricity prices are established in competitive markets. EIA publishes (on-line) a report called the Electric Monthly Update. Among other things, this report shows a comparison of wholesale electric prices in the RTOs and other locations for the month covered by the report, and the previous 12 months. Below is a screenshot of the most recent edition of the report, which shows wholesale price comparisons for July of 2025 and for the 12 months of August 2024 through July 2025.

What this comparison shows is that wholesale prices in California are about the same, or less, than wholesale prices in the RTOs and other locations. There is no reason to believe, based on this data, that the cost of generating electricity in California explains why California’s electric rates to consumers are the highest in the country. Similarly, the comparison shows that wholesale prices in New England, where all of the other high cost states currently are located, is not out of line with many other regions. This confirms my belief that most of the difference among states in the cost of electricity results from the non-electricity costs that are added on to the bills paid by consumers.

I looked at a number of sources in trying to figure out what these non-electricity costs are in California. The most helpful of these (and the one I am going to cite in this post) is a January 2025 report by California’s Legislative Analyst’s Office entitled “Assessing California’s Climate Policies—Residential Electricity Rates in California.” The report goes into detail as to the various elements of the prices charged by different utilities in California. It also describes the elements that have led to high electricity prices and attempts to quantify the effects of each element on electricity prices.

As this report explains, an important reason why California’s electricity prices have risen dramatically in the last few years has been the costs incurred by California’s three investor-owned utilities—Pacific Gas & Electric (PG&E), Southern California Edison (SCE), and San Diego Gas & Electric (SDG&E) is the costs incurred by those utilities from the large wildfires that have raged in California the last few years. California has a unique system that has assigned the cost responsibility for these wildfires to these utilities (I could write a long post on this but am refraining for now). This means the utilities’ electric rates must be high enough to recover not only some of the damages they have been forced to pay, but also: (1) the cost of steps they have been required to take to mitigate future wildfire damage; and (2) payments for wildfire insurance. According to the report, wildfire costs account for 7 to 13 percent of these utilities’ electric bills. These wildfire costs, which put California’s electric rates at the top, have nothing to do with renewable resources.

But that doesn’t mean renewables don’t also bear some responsibility for California’s high electricity prices. Unlike other states, including many of the New England states, California does not include requirements that utilities purchase electricity from renewables at fixed above-market prices that are then passed on to their customers. But California does have a Renewable Portfolio Standard (RPS) designed to ensure that a significant portion of the state’s electricity comes from renewable resources. One way this is implemented is to impose a requirement that utilities purchase renewable energy credits (RECs), which consist of a set payment for each megawatt-hour of electricity the renewable facility generates. These REC payments are separate from the payments the facility also receives for the electricity it sells. For example, a renewable facility can sell its electricity to one utility for a price of $30 per megawatt-hour and also sell a REC for the very same electricity to a different utility for a price of $20 per megawatt-hour.

The costs of a utility’s REC payments are included in its rates in California. This approach is not unique to California, but is similar to REC programs in several other states. According to the Assessing California’s Climate Policies report, RPS costs have increased the investor-owned utilities’ rates by about 5 percent. An additional 4 percent rate increase comes from other state mandated climate-related support included in the investor-owned utilities’ rates.

In addition to these REC and other payments increasing the price of electricity, another factor in California’s high electricity prices is what the Assessing California’s Climate Policies report calls “solar cost shifts.” This cost shift results from customers who install solar panels that produce more electricity than they use in their homes or other buildings. These customers sell their excess electricity back to their utility at the retail price of electricity, instead of the lower wholesale rate that utilities otherwise pay for electricity. The higher price for electricity the utilities pay to these solar customers is reflected in higher electricity rates for customers without solar power. For many years the solar cost shift was relatively minor, but the Assessing California’s Climate Policies report concludes that the shift has doubled since 2020. This is because California’s fixed costs are increasing (due to, among other things, the wildfire related costs) while at the same time more customers are installing solar panels, which leaves fewer customers to pay for the fixed costs. The report estimates that the cost shift in 2023 increased customers’ bills by 11 percent to 20 percent.

What is the total effect of these various non-electricity cost additions to the price of electricity in California? It is hard for me to quantify the costs very precisely, but the following chart from the Assessing California’s Climate Policies report provides one way to assess the price effects of these additions.

The chart shows the electric rates charged by various utilities in California. Most of the costs I have discussed above are paid only by the three investor-owned utilities, and the chart shows this effect; the three investor-owned utilities have by far the highest rates for electricity. For example, the rates of the Sacramento Municipal Utility District (SMUD) are only about half as high as the rates of PG&E, which serves cities in neighboring locations. Similarly, the rates of the Los Angeles Department of Water and Power (LADWP) are only about 70% of the rates charged by SCE, which serves communities outside of Los Angeles. There are many reasons why the non investor-owned utilities have lower rates, including that they own fewer generation and transmission facilities and operate on a non-profit basis, but clearly the special costs borne by the investor-owned utilities play an important factor.

Finally, I would note that my table shows that California had relatively high rates in the 2001-2003 period, before the introduction of renewable resources. One reason for this was that customers in California tended to use less electricity, due to California’s relatively mild climate. This meant that the costs of the assets used to serve customers were spread out among fewer sales of electricity. Think of it as like renting a beach house. If five people agree to share the costs of the house, the amount they pay per person will be higher than if only three people share the costs, even though the cost of the house rental is the same.

California continues to have relatively low electricity usage rates today, not only because of its climate but also because it has the highest penetration of solar panels in the country. These solar customers rely on their panels to supply most or, if they have batteries, all of their electricity, leaving fewer customers to pay the fixed costs required to provide them with their electricity. The end result is that non-solar customers in California pay higher rates for electricity.

With this explanation, let me return to the question that started me on this post, which is whether California’s high electricity rates can be blamed on its high renewable generation facility penetration. It seems obvious that California’s rank as the state with the very highest rates is due to the damage done by California’s series of devastating wildfires, and not to renewable generators. But even without the wildfires, California’s electricity prices would be near, but not at, the top of all states. And California’s electricity prices were relatively high in 2014-2018, when California was a leading renewable energy producer but did not have significant wildfire costs. Certainly, the very significant costs associated with renewable energy identified in the Assessing California’s Climate Policies report represent a significant element of the electricity prices charged by California’s investor-owned utilities.

But it also is true that many other factors causing California’s high electricity rates are unrelated, or at least not directly related, to renewable generators. Further, it is important to consider that, according to EIA, renewable generators produced approximately 80,000 megawatt-hours of electricity in California in 2024. Without those renewable generators, California would have needed to produce those 80,000 megawatt-hours of electricity in some other way. And there would have been a cost to that production.

Consequently, I am comfortable saying that, most likely, California’s subsidies of renewables have raised the price of electricity to some extent, at least to the customers of the three investor-owned utilities. But I think that is all I can say.

***

I hope you enjoyed this post. I enjoy writing them and will never charge for subscriptions or ask for donations. I only ask that, if you did enjoy it, you press the “like” button below. And please consider sharing my posts with other people you think might be interested. Doing so will help me evaluate interest in the book I am writing on grid operations. Of course, if you have a reaction to, or question about, this post, please leave a comment and I will be happy to respond.

“Yes, renewable resources cause rates for electricity to increase. Just look at California. It has a lot of wind and solar, and it has the highest electric rates in the country.” I know that many of you have read statements like this, and some of you have even written something like this to me in your comments to my posts. Similarly, some people have cited to high electric rates in California and New England to support their contention that Regional Transmission Organizations (RTOs) cause rates for electricity to increase. I have been wanting to write a post about this issue for a while now, and in the month I have had under my new publication schedule, I have been able to figure out what I want to say.

Let me start by going off on a tangent. I have been a huge fan of the Washington Nationals baseball team ever since their first season in Washington in 2006. I try to follow at least part of every one of their 162 games (plus playoffs in some years), either on TV, radio, or the internet. And I have many superstitions about my part in ensuring their success. One of these is my selection of coffee cups in the morning. When the Nats win a game after I used a particular cup, then I keep using that cup every day until they lose. I know deep inside that just because the Nats won a game when I used a particular coffee cup does not mean they will win every time I use that cup. Frequently, especially in the last few years when the Nats haven’t been very good, it takes only one day to prove that a particular cup did not cause the Nats to win.

In the same way, we need to look at only one state to know that reliance on renewable generators or participation in an RTO does not inexorably lead to high electricity rates. Texas has by far the most renewable generation in the country. In 2024, renewable generators located in Texas generated over twice as much electricity as renewable generators in California. Yet rates for electricity in Texas are below the average for the US as a whole, and lower than all but 12 other states. And yes, Texas has formed its own RTO. What is more, although Texas’ rates for electricity are lower than rates in most states today, they were the 14^th highest in the country and above the national average in 2001, when Texas had virtually no renewables and wasn’t in an RTO.

California and Texas represent only two data points. What can we learn by looking at all states? The Energy Information Administration’s (EIA) Electricity Data Browser, https://www.eia.gov/electricity/data/browser, has data on many different subjects going back to 2001, including data on prices to consumers. I used that browser to get the data on Texas’s renewable generation and rates for electricity, and I also used that browser to get data on all 50 states plus the District of Columbia to develop the following three tables. (I excluded Hawaii and Alaska because their rates are affected by their remote and isolated locations). Here is the first table.

On the left side of this table, I have listed in order the top 10 states in terms of their production of electricity by renewable generators in 2024 (other than hydroelectric generation). In the column next to these states, I have listed their rank in terms of their rates for electricity. To make it easier to consider this information (hopefully not harder), I color coded each entry. Yellow shows the one state—California—that is ranked in the top ten in both renewable generation and rates. Green shows the states whose rates for electricity are in the top half of all states. And the reddish color shows the states whose rates for electricity are in the bottom half of all states. This table shows , as I described above, that Texas produced the most renewable electricity in 2024, but had only the 39th highest electricity prices.

This table suggests that the correlation between high renewable electricity production and high rates for electricity is about as good as the correlation between my choice of a coffee cup for breakfast and a victory by the Washington Nationals. Half of the states that produce the most renewable electricity have relatively high electricity prices that are in the top half of all states, and an equal number have relatively low prices in the bottom half of all states. I would note, however, that California is the only one of these high renewable production states whose rates are in the top ten, while three of the high renewable generation states—Iowa, Oklahoma, and New Mexico—have rates in the bottom ten. As result, if anything, the table suggests that states with high amounts of renewable energy are more likely to have lower electricity prices.

My next table looks at the correlation between renewable generation and rates for electricity from the opposite perspective.

This table lists the ten states with the highest electricity prices in 2024 and then shows where those states stand in terms of the production of electricity by renewable generators. The point of this table is to see whether high electricity rates are associated with the production of large amounts of electricity from renewable generators.

The table shows that there does not appear to be an obvious correlation between high rates and high penetration of renewable generation. Similar to my first table, I have color coded the table to (hopefully) help you understand what it shows. I highlighted in yellow the only state (California) that is in the top ten in both rates for electricity and amount of electricity produced by renewable generators. I also highlighted in green the only state (New York) that is both in the top ten in electricity rates and in the top half in production of electricity from renewable generators. The other eight states that are in the top ten in terms of electricity prices are highlighted in a reddish color, indicating that they are in the bottom half of all states in production of electricity from renewable generation.[1] Again, if this table shows anything, the fact that eight high price states have low renewable electricity production suggests that lower amounts of renewable generation are more likely to result in high electricity prices.

Those of you who have some familiarity with the US electric industry may have noticed something else about my table above listing the ten states with the highest prices for electricity. Every one of these states is located in an RTO. Does this mean that RTO participation leads to higher rates for electricity? Not necessarily. Let me explain with my third table based on data in EIA’s energy data browser.

This table shows the states with the ten highest electricity prices in the years 2001-2003 and 2022-2024. I prepared this table to see if there was any change in which states had high prices at a time when there were few renewable energy generators and relatively little RTO participation and which states have relatively high prices today. I included three years in each period to remove the possibility that an anomaly in one year might skew the results.[2]

What is striking about this table is that there is almost no change in the high price states between these two periods. The District of Columbia appears only in recent years, while Pennsylvania and Nevada took turns appearing at the number ten spot in the earlier period. But otherwise, the same states that had the highest electricity prices in the earlier period have the highest electricity prices today, with only a modest shuffling of their rankings. And the same is generally true for the states not shown on this table. For example, eight of the ten lowest electric price states in 2001 remained one of the ten lowest electric price states in at least one of the most recent three years.

What this table suggests is that many of the factors causing a state to have relatively high or low rates for electricity predate the advent of RTOs (and also the use of renewable resources). This is not surprising to me. The desire to get a handle on high rates was one of the prime motivating factors that led states to require, or at least permit, their utilities to join RTOs. The three regions that resisted RTO participation—the Southeast, Southwest, and Pacific Northwest—all enjoyed relatively low electricity prices due their generation mix. The Southeast and Southwest had ample relatively low cost baseload coal and nuclear generators, while the Pacific Northwest enjoyed access to large amounts of inexpensive hydropower. Utilities in these regions did not need to join an RTO to address concerns about high electricity prices. Twenty-five years later, states whose utilities joined RTOs continue to have high electricity prices relative to prices in lower rate states. But that can be largely attributed to the same low cost generation advantages low rate states enjoyed 20-25 years ago, and not necessarily to participation in RTOs or a decision to favor renewable generation.

Further, it is not the case that all states located in RTOs have high prices for electricity. Although high rates led many utilities to join RTOs, different operational considerations led other utilities with relatively low electricity rates to join, particularly the Midcontinent Independent System Operator (MISO) and the Southwest Power Pool (SPP). And these utilities’ rates have remained relatively low despite their RTO membership. Seven of the ten lowest electricity price states in 2024 are located in RTOs, and another ten states located in RTOs are ranked in the bottom half of prices for electricity.

I want to emphasize I am not asserting that renewable generation and RTO participation have no effect on electricity rates. Each of these factors can play an important role in determining the rates for electricity paid by consumers. My point instead is that it is not possible to reach conclusions about the price effects of renewable generation and RTO membership based on a simplistic look at electricity rates in California and a few other states. My analysis suggests that a different, more complicated, picture emerges when you look at the electricity prices, renewable electricity production, and RTO participation in every state. This analysis shows it is necessary to understand each state’s unique circumstances in order to understand why electricity prices in that state are relatively high or low.

California, the poster child for those who claim that renewable resources lead to high electricity prices, is the perfect example. There are many factors contributing to the high electricity prices in California that have nothing to do with renewables. One such factor is that electric utilities in California have been permitted to recover through rates a portion of the billions of dollars they have been forced to pay as damages resulting from the major fires that have hit California over the past few years. There are many other factors as well causing California to have, by far, the highest electricity prices in the US. Perhaps this would make a good subject for my next post.

***

I hope you enjoyed this post. I enjoy writing them and will never charge for subscriptions or ask for donations. I only ask that, if you did enjoy it, you press the “like” button below. Doing so will help me evaluate interest in the book I am writing on grid operations. Of course, if you have a reaction to, or question about, this post, please leave a comment and I will be happy to respond.

[1] Some of you might question whether this chart is misleading in that smaller states and District of Columbia don’t produce large amounts of electricity from any source, and that they might be producing a relatively large amount of electricity from renewable generation relative to the total amount of electricity produced. I did a quick check to see if there were any meaningful differences if I ranked the states by the percentage of total electricity produced from renewable generation facilities. As far as I could tell, there wasn’t any real difference except for the District of Columbia.

[2] I used a slightly different version of this table in one of my earliest posts, “What Can We Learn About RTOs From Data on Consumer Rates?” (August 5, 2024).

My previous post earlier today was my 57th post since my first one about 13 months ago on August 5, 2024. In total, those posts contain about 105,000 words. In that same period, I have written only about 30,000 words of my book. I think that doing the Explaining the Grid Substack has been extremely helpful to me in thinking about how to write my book, and I also have about 600 subscribers/followers, which is not a huge group but also not insubstantial.

But now I need to devote more time to working on my book. Over the last 13 months I have tried to post a new article about once a week. And although I have not always been able to do so, I have averaged about one post a week for the last year. Now I am going to change my schedule to about one post a month. Hopefully they will continue to be interesting to you, and fun for me to write. Look for my next post sometime in early October.

And now for the last of my three-part series of posts on dealing with the increased demand for electricity caused by AI. In my last post I examined the steps that can be taken quickly to help ensure that existing conventional generation will not be retired. This time I am looking further out into the future, into the intermediate and longer term future time frames. These are the time frames in which new generation facilities can be constructed. And new construction will be needed. Even if no existing generation facilities are retired, the grid’s existing generation fleet will not be able to keep up with the growth in demand.

Even though I am talking about adding generation a number of years into the future, that doesn’t mean that action to make this happen can be put off into future years. It is pretty much impossible to plan, finance, and construct any type of new generation facility in less than three years. And the time it will take to construct larger baseload plants, such as nuclear power plants, is considerably longer, even under optimistic assumptions. This means action needs to be taken now to ensure that the necessary new generation capacity is constructed and placed into service in the future.

There are two aspects to the question of what should be done to promote the construction of the new generation. The first is what generation should be constructed. The second is what can be done to facilitate this construction. Let’s take these in order.

What Generation Should Be Constructed?

Generation For the Intermediate Term

In thinking about what kinds of generation facilities should be constructed, it’s important to understand the projected shortage in generation capacity faced by the grid. To do that, let’s start by looking at the North American Reliability Corporation’s (NERC) reliability assessment for the summer of 2025. NERC is statutorily tasked with the responsibility for ensuring the reliability of the grid, so its assessment deserves some weight. We have looked at this assessment in a couple of previous posts, but for those who don’t want to look for my old posts, here is a copy of a slide that summarizes NERC’s findings.

As you can see, NERC determined that four Regional Transmission Organizations (RTOs) faced a risk of having insufficient operating reserves, but only in “Extreme Conditions.” And these conditions are; (1) a very hot day, resulting in a very high demand for electricity; (2) “historical high” outages of generation facilities; and (3) low production from “variable energy resources,” i.e. wind and solar facilities. I am going to use some of my clunky graphics to try to illustrate what this means.

First, above is a graphic I have used before, showing the cycle of demand for electricity. The graphic is illustrative only, but the basic shape applies in every electric system. As the graph shows, demand is relatively low at midnight, then starts to rise as people get up, eat breakfast, and go to work/school. Demand also increases as the temperature rises, peaking around 3 pm, and then declining as people go home, have dinner, watch TV, and go to bed, until it drops to approximately the same level as it was the previous midnight. This cycle is intuitively obvious, but many people overlook its importance in understanding how the grid works.

This is also a graphic I have used before. It shows a typical generation mix used to satisfy the demand for electricity at different times in the daily demand cycle. The large coal and nuclear units on the bottom row are baseload units that operate all day. The next row shows what are what are called intermediate units, which I have depicted as natural gas combined-cycle units. They run during the day, but not much at night. The next row include smaller natural gas combustion turbines, solar panels, and wind turbines. These units are less expensive to construct, and operate mostly in the middle of the day. The last row consists of small solar and wind units. Their operations center around the time of the peak demand. Again, this graphic is for illustrative purposes only; the actual generation mixes employed by electric systems are more complex.

Now we get to see a brand new graphic that I made just for this post. It shows what happens when hot temperatures drive demand up to historic levels at the same time that generator outages increase and wind and solar output declines. The blank spots under the demand curve represent those times when there is not enough operating generation to supply all demand. The important point illustrated by this graph is that the blank spots appear mostly for a few hours in the middle of the day around the time of peak demand.

Typically, electric systems plan in advance for how they will meet the demand for electricity in the future and assemble a mix of different types of generation that best meets that demand efficiently and reliably. However, the explosion in demand that the grid faces today was unexpected. Only limited types of generation facilities can be constructed in time to meet increased demand in the intermediate term (3-5 years), Large baseload units can’t be permitted, financed, and constructed in that time frame.

Luckily, as our illustrative graphic shows, new baseload units are not absolutely required in the intermediate future. Instead, the primary need for the summer of 2025 is for generation facilities that operate only a few hours a day, and even then only during extreme conditions. Demand will increase and a certain amount of generation will be retired in the next few years, but the basic problem should not change in the intermediate term. Electric systems have sufficient capacity that should permit them to meet the demand for electricity in most hours, even though they potentially could struggle to meet peak demand on hot summer days or cold winter days.

This means that the primary type of new generation capacity required to be constructed in the intermediate term is “peaking capacity,” which is designed to operate at most only a few hours of the day when demand is at its highest. Luckily, peaking capacity tends to be the cheapest type of generation capacity. It also can be constructed more quickly and therefore be placed in service in time to address potential capacity shortfalls that otherwise might occur over the next few years. Unfortunately, peaking capacity also tends to be less efficient, and therefore more expensive, to operate. Which is why “peakers,” as they are sometimes called, typically are only called on to operate a few hours a day during peak demand conditions.

Most peaking generators currently being constructed today are natural gas combustion turbines. This is a very flexible generation technology in that combustion turbines can be very small, with capacities below 10 megawatts, or relatively large, with capacities over 100 megawatts. Combustion turbines represent the principal type of conventional generators currently being planned or under construction, and it should be possible to add additional combustion turbine capacity to meet increasing demand in the intermediate term.

You will note that this graphic, and my previous graphics, all show some amount of solar and wind generation being used to meet electric demand. I don’t intend for this post to address the issue of whether we can rely on solar and wind capacity to meet increased demand for electricity, and the inclusion of wind and solar facilities in my graphic certainly does not constitute a prediction of how much of these types of generation capacity will be used. Let me just say, as I have written in previous posts, that I do think renewables can provide at least some amount of capacity in an integrated electric system. At the same time, however, if you look at the NERC 2025 Summer Assessment slide that I inserted above, you will see that part of NERC’s definition of “Extreme summer conditions” includes low output from variable resources (i.e. renewables). I think it is fair to say that NERC is concerned that some RTOs may be placing too much reliance on solar and wind generation. That being the case, I believe that most of the new generation capacity required in the intermediate term will need to consist of natural gas turbine generators.

Generation for the Longer Term

At some point, hopefully not in the next few years, the increase in the demand for electricity to serve AI will affect not only the peak demand for electricity but demand in all hours of the day. The effect of AI’s rapacious thirst for electricity potentially could be magnified by an increase in demand for electricity to serve electric vehicles if the current backlash against EVs does not take permanent roots. Further, even if it is possible to stave off the retirement of old coal and natural gas steam plants for a few more years, many of those plants aren’t just old, they are very old and cannot run forever. The following graphic shows how this increase in demand affects the ability of our hypothetical electric system to meet the increased demand.

As you can see, the higher demand curve leaves holes in just about every hour of the day, even before considering that, in my graphic, two of the coal-fired baseload units have been retired. With those retirements, the graphic shows a major shortage of generation capacity. Adding peaking capacity alone will not be a viable option. Instead, new baseload generation capacity will also be required.

What kind of baseload capacity are we talking about? Today, there are five different kinds of generation facilities in operation that can be used to provide baseload service: (1) nuclear; (2) coal; (3) natural gas steam; (4) natural gas combined cycle; and (5) hydroelectric. Which of these should be pursued for new construction?

Of these five, natural gas steam units are the only ones that can easily be eliminated from consideration. Natural gas steam units produce electricity by using natural gas to heat up water in boilers to produce steam that rotates turbines to produce electricity. This is an efficient way to generate electricity. What is more, natural gas steam units are very inflexible operationally. It takes time to start them up, and they cannot “ramp” their output up or down very quickly.

The second type of generation technology—natural gas combined cycle units—represents a much better choice for generating electricity from natural gas. Combined cycle units employ one or, more frequently two or more, combustion turbines. Combustion turbines alone are not particularly efficient, but a combined cycle unit employs the waste heat from the combustion turbines to heat water to produce steam to rotate turbines and generate electricity the same way as natural gas steam units. Combined cycle units are the most efficient of all fossil generators, and also are very flexible operationally. For these reasons, I think that natural gas combined cycle units are the only one of the five types of generation technologies that definitely will be employed in constructing new baseload generation facilities. I should note, however, that an essential requirement for a baseload combined cycle unit is a supply of natural gas. There is a shortage of natural gas pipeline capacity in much of the country, so constructing new gas pipelines is part and parcel of adding significant new natural gas combined cycle capacity.

Nuclear power is the third potential candidate for future baseload generation. And I think that constructing multiple new nuclear power plants would be the best possible approach. Nuclear power plants have very high capacity factors which means that they can produce much more electricity than any other kind of generator of the same size. This means it is possible to satisfy future growth in demand by constructing a relatively few new nuclear generation facilities. And nuclear power plants have no emissions of any kind. As I have written, however, to date nuclear power is fiendishly expensive, takes a long time to construct, and any new plants will face strong local opposition. Nevertheless, as I also have written, there has been a strong push recently to develop new nuclear plants. For example, as I explained in my post “Will the New Nuclear Executive Orders Jumpstart the Construction of New Nuclear Reactors?” on May 23, the President signed four executive orders with the goal of having 10 large reactors under construction in the United States by 2030 and placing 300 gigawatts (GW) of new nuclear energy capacity in service by 2050 (currently there is about 100 gigawatts of nuclear energy capacity in operation). I hope that these goals are realized but, for the reasons I described in my Nuclear Executive Orders post, I am somewhat skeptical that it will happen. As a result, it is not clear to me whether very many, if any, new nuclear power plants will be constructed in the longer term to meet the growth in demand for electricity,

The fourth generation technology that potentially could be employed for use as baseload generation is hydropower. In many respects, hydropower has the same benefits and disadvantages as nuclear power. Like nuclear power, hydropower has no emissions but, also like nuclear power, large hydropower facilities are very expensive to construct and the permitting process can take a long time. Hydropower also can be quite flexible in its operations. Unlike nuclear power, however, there is no current push to construct major new hydropower facilities. To the contrary, the major push has been by environmentalists who want to remove hydroelectric dams because of the environmental harm they do, particularly to fish and other wildlife. Not only are there no planned major new hydroelectric projects on the horizon, but four large hydroelectric dams were recently removed from the Klamath River in northern California. As a result, it seems unlikely that significant amounts of new hydroelectric capacity will be constructed over the longer term to add to the grid’s baseload generation capacity.

This leaves coal-fired generation. Coal is the dirtiest of the fossil fuels currently used for baseload generation (oil is dirtier, but is not used for baseload generation and mostly is used for emergency peaking power). Coal capacity also is very expensive, with a recent cost estimate coming in at over $1billion for a new baseload coal plant. No new coal plants have been built in over 10 years and no new coal plants are currently planned. It appears, however, that the current administration may be trying to facilitate a comeback for new coal plants. Consequently, although I think it unlikely that any new coal plants will be constructed, I can’t count it out with the same certainty that I can for natural gas steam plants.

Some of you may notice that I haven’t mentioned renewable generation facilities as candidates for new baseload capacity. The problem is that baseload capacity has to be capable of running all 24 hours of the day, and renewables cannot do that. It might be possible to pair renewables with storage to provide baseload service. But the amount of excess renewable capacity and storage required makes providing baseload capacity with renewables a very expensive proposition at the present time. Perhaps baseload service can be provided by renewables in the distant future. However, I don’t see significant amounts of baseload service being provided by renewables in the same timeframes as other conventional generation technologies.

Facilitating the Construction of New Generation Capacity

Identifying the new generation facilities we need is not that hard, Getting anyone to build them will be much harder. Constructing even the smallest, least expensive combustion turbine can cost several million dollars and larger ones can cost $100 million or more. It will cost much more than that to build new large baseload generators. New conventional baseload generators cost around a billion dollars or more. The last two large nuclear reactors that were constructed—Southern Company’s Vogtle 3 and 4—had a combined cost of $34 billion.

Not surprisingly, no company is going to invest this kind of money in a generation facility without having a reasonable expectation that the facility will earn revenues that cover all of its costs. This includes the cost of paying off the facility’s debt, plus an acceptable profit. (Just what constitutes an acceptable profit depends on a company’s other profit opportunities) In order to facilitate the construction of the needed new generation capacity, it therefore is necessary to have mechanisms in place that can provide assurances that the new generation facility is likely to earn the revenues needed to make this profit..

For most of the grid’s history, there were two ways to provide the necessary assurances that the needed revenues could be earned. The most common way was for a utility to construct, own, and operate the new generators it needed to serve its customers. The utility’s rates were set by the local public utility commission at a level that allowed the utility to recover the projected operating and capital costs of its generation facilities, plus a return on investment. This approach does not represent an absolute guarantee of profit, because the rates might not be set high enough to recover actual costs and the utility commission is able to change the amount of the utility’s profit included in its rates. But the approach provided sufficient assurance of revenues that it supported the construction of all generation necessary to meet the demand for electricity.

Starting in the 1980s, however, an independent power producer category of generation owners, i.e. owners whose rates are not regulated by local public utility commissions—started to emerge. The utility rate model doesn’t work for independent power producers, because their rates are not regulated as cost-based rates. Instead, independent power producers have largely financed their generation facilities by entering into long-term fixed price contracts with customers; mostly utilities but also some large retail customers. The fixed prices are calculated to produce revenues that should cover the generators’ costs plus earn a profit.

Today, most state public utility commissions require electric utilities to engage in what is called “Integrated Resource Planning.” This involves projecting the demand for electricity in future years. If the projections reveal that the utility will need more generation capacity in the future, then the Integrated Resource Plan also identifies how that increased demand will be met. Typically, utilities propose to either construct new generation facilities or enter into power purchase contracts with independent power producers. The utility commission can either approve this plan or require that changes be met before the plan can be approved. Once approved, however, the utility moves forward to implement the plan and either constructs new generation facilities or enters into power purchase contracts with independent power producers.

Thus, state public utility commissions will play a crucial role in facilitating the construction of new generation facilities in states where public utility commissions continue to have jurisdiction over their utility’s acquisition of generation capacity. These commissions must recognize that existing generation capacity is inadequate to meet increased demand in the future. And state utility commissions must be willing to approve the construction of the new generation capacity needed to meet future demand, as well as the recovery of the costs of such construction in utility rates. This likely will be difficult politically as it will result in an increase in the rates for electricity. But it must be done.

Providing the necessary financial assurances for new generation facilities is more complicated, and more difficult, in the seven Regional Transmission Organizations (RTOs). These RTOs operate competitive markets that supply about two-thirds of all electricity consumed in the US. As a result, those markets will play a vital role in supporting the construction of new generation capacity required to meet increasing demand for electricity.

I have written extensively in other posts about the RTO competitive markets and how they work,[1] and will not repeat that discussion here. All RTOs have somewhat different markets, but they all have energy markets, in which electricity is bought and sold for short durations ranging from one hour at the longest to five minutes at the shortest. The energy markets are designed to respond to the level of demand for electricity, where prices are relatively high when demand is high and relatively low when demand is low. The belief when the RTOs were first being formed in the late 1990s and early 2000s was that high prices in the energy markets when demand is high would provide sufficient encouragement for the construction of the new generation needed to meet the increased demand.

In practice, this has turned out not to be the case, especially for large baseload generation facilities. There are at least three reasons for this. First, companies understandably are reluctant to commit hundreds of millions of dollars, or more, when they have no assurance about the future price of electricity. There is just too great a risk that energy prices in an RTO market will be too low to support a proposed generation project.

Second, that risk is magnified by the fact that the RTO energy markets are self-correcting. The high prices that exist in these markets when capacity is short decline once new capacity has been constructed and there no longer is a shortage. The same dynamics that drive prices up in a shortage drive prices back down when more capacity is on the market. As a result, if a new generator relieves the shortage of capacity causing high prices, it may never see those high prices after construction is completed.

Third, after RTOs started operating with energy markets as their only market, it became clear that these markets were not providing generators with sufficient revenues. As a result, six of the seven RTOs—the Electric Council of Texas (ERCOT) being the sole exception—have determined it is necessary to establish an additional “capacity” mechanism to provide additional revenues to generation owners. Capacity is not a physical product, but rather a contractual obligation to be available to produce electricity when required to meet the demand for electricity. Payments for capacity were intended to provide what was called the “missing money” not coming from the RTOs’ energy markets that was needed to support the continued operations of generators. ERCOT decided that capacity payments were not needed to support Texas generators and continued to rely solely on its energy market.

Two of the RTOs—the California ISO and the Southwest Power Pool (SPP)—opted not to implement a capacity market, but instead require that the electric utilities in their regions acquire generation capacity either through direct ownership or by contracting with an independent power producer. These utilities will all require approval from their state utility commissions to construct or contract for new generation capacity. Consequently, the analysis I have already described above applies to these RTOs, i.e. the public utility commissions must approve the recovery of the costs of needed new facilities in retail rates.

With a few exceptions, state utility commissions do not approve the prices paid for capacity in three RTOs: (1) ISO-New England; (2) New York ISO; and (3) PJM. These three RTOs cover the entire Eastern seaboard from Maine to Virginia, and also extend as far west as Chicago. These RTOs instead have established competitive capacity markets. Those markets are structured similarly to the RTO energy markets. The markets set high prices when there is less generation capacity, and lower prices when there is adequate capacity to meet the anticipated demand. In theory, high capacity market prices encourage the construction of new generation capacity when needed to meet the demand for electricity.

The RTO capacity markets worked reasonably well during the past 20 years or so; a period when demand for electricity has been reasonably flat and relatively little new generation capacity has been needed. However, those markets are not structured in a way that will likely encourage the large build out of new generation capacity that will be needed to meet the increased demand for electricity in the future. The problems are similar to the problems with the energy markets I have described above.

First, although the terms of the capacity purchased in the RTO capacity markets is longer than the five minute to one hour terms for sales of energy, those terms are still too short to encourage much construction of generation capacity, especially large baseload generation facilities. The ISO-New England and PJM capacity markets acquire capacity every year for a term of one year. Companies generally are unwilling to commit the capital required for the construction of a new generation facility if they do not know the price they will receive for their capacity more than one year after they go into service. The terms of the capacity purchased in the New York ISO’s capacity market are even shorter; less than one year.

Second, like the RTO energy markets, the RTO capacity markets are self-correcting, in that high prices caused by a shortage of capacity decline in future auctions after new capacity is purchased. There is a reasonable likelihood that the high capacity prices which encouraged a company to offer to construct a new generation facility will not be the capacity prices that actually prevail when that facility has been constructed and placed into service. This self-correcting nature of the RTO capacity markets only increases the risk associated with the short terms of the capacity sold in the RTO markets.

It order for the RTO capacity markets to support the construction of the large amounts of new capacity needed, those markets should be modified to establish a new capacity product. This product could be structured in many different ways. But the essence is that the new capacity product should have a term much longer than just a single year. Setting a price for capacity for a much longer term—say 10-15 years to match the typical terms of loan agreements—could provide the revenue assurance needed to finance large new generation facilities.

There are many technical details that would need to be ironed out for this new capacity product to work. Unfortunately, RTOs are not structured to develop these kinds of mechanisms quickly. There are literally hundreds, if not thousands, of “stakeholders” for each RTO who are entitled to have some input into the process. And even if an RTO is able to develop a proposal, that proposal then must be approved by the Federal Energy Regulatory Commission (FERC), where it almost certainly would face strong opposition. But I think that the RTOs must take some kind of action to modify their capacity markets soon or they will not attract the construction of new generation capacity needed to meet the demand for electricity.

There is a second approach that also could support the construction of at least some new generation capacity. Electric utilities in RTOs could enter into what are called “contracts for differences.” These contracts don’t require a generator to physically deliver energy to the utility, but rather are a form of contract known as a financial contract. I won’t go into the somewhat complex details of how contracts for differences work, but in brief they require the generator to pay the utility when RTO energy market prices are above a certain level and it requires the utility to pay the generator when energy market prices are below a certain level. The practical effect of these offsetting payments is that the generator receives the contract price for all of the energy it sells into the RTO energy market no matter what the actual market price is, and the utility pays the same price for all the electricity it purchases no matter what the actual market price.

Contracts for differences can be beneficial for both the generators and the utilities. The generator has locked in a price for the electricity it produces for the term of the contract, and the utility has protected itself against having to pay high market prices for energy during that same term of the contract. With the price fixed for the term of the contract, a company building a new generation facility has the assurance about the revenues it will receive once the facility is placed in service. Of course, it may not be possible for a generation company to reach a deal with a utility on a contract for differences at a price and a term that will support the financing of a new generator. And it almost certainly is the case that not enough contracts for differences can be executed to support the construction of all the new generation that will be needed to meet the increased demand for electricity caused by AI. But contracts for differences may support the construction of at least some new generations facilities that otherwise might be built.

You may have noticed that, so far, I have only mentioned six RTOs. The seventh RTO, the Midcontinent ISO (MISO), is kind of a hybrid of the other RTOs. Most of the utilities in MISO own their generation facilities, the rates of which are regulated by their local state utility commissions. Those commissions do authorize the construction of new generation facilities and the recovery of the costs of that construction through retail rates. Not all utilities own generators, however. And some utilities that do own generation facilities do not have enough generation capacity to serve all of their customers. Consequently, MISO conducts capacity auctions to provide for the sale of capacity to utilities that need it. Because MISO is a hybrid of the two types of RTOs I have already discussed, the same issues and solutions I have discussed for each of these two types of RTOs applies to MISO.

Other Impediments to the Construction of New Generation Capacity

Although I think that the most important way to support the construction of new generators is to facilitate their financing, there are other impediments that also should be addressed. These impediments do not necessarily bar new construction. They do, however, affect the timing and the cost of construction. I briefly discuss two of the most significant of these here.

The first impediment involves the need to interconnect new generation facilities to the transmission grid. I discussed this issue in detail in my April 15 post, “The Little-Known Bottleneck Slowing Down the Addition of Generation Capacity to the Grid.” As I explained there, before any new generation facility can be interconnected to the grid, that proposed interconnection must be studied to determine whether the transmission system can accommodate the interconnection. If the study shows the transmission system would need to be upgraded, the owner of the generation facility must pay the costs of all necessary upgrades. Depending on the nature of the upgrade that is needed, the cost could be in the hundreds of millions of dollars.

But the primary problem with interconnection process is not the potential cost of upgrades. There currently are thousands of interconnection requests across the country waiting to be studied, and these requests have overwhelmed the ability of transmission owners to do the necessary studies. This means it can take years before a requested interconnected study can be completed. As a result, even if the construction of new generation facilities could be sped up, they might not be able to deliver their electricity to the grid because of the time needed to get an interconnection approval.

The second impediment is the more general issue of the permitting of new generation and transmission projects, including environmental reviews. Opponents of proposed new projects have become adept at delaying proposed projects. Their goal is not so much to ensure compliance with all permitting requirements, but rather to delay the process and impose additional conditions for approval until the project becomes too expensive and is abandoned.

It is, of course, important for proposed projects to be adequately reviewed for compliance with applicable legal and environmental requirements. But the permitting process has become burdened with lengthy and unnecessary procedures that serve to lengthen the time for approval without providing any real benefit in the review of proposed projects. The current administration has attempted to streamline the permitting process in several ways, some of which are controversial. Without commenting on any of these specific permitting changes, I do agree it is essential to streamline the permitting process so that the necessary new generation capacity can be placed in service in time to meet the rapidly increasing demand for electricity.

***

I hope you enjoyed this post. I enjoy writing them and will never charge for subscriptions or ask for donations. I only ask that, if you did enjoy it, you press the “like” button below. Doing so will help me evaluate interest in the book I am writing on grid operations. Of course, if you have a reaction to, or question about, this post, please leave a comment and I will be happy to respond.

[1] The most detailed descriptions appeared in Explaining the Grid, Parts Three-Five, which were posted in October and November of 2024; Subsidies, How Do They Affect RTO Markets?, posted on November 24, 2024; and Renewable Generation Has an Outsized Effect on RTO Energy Market Prices, posted on July 14, 2025.

Sorry for the delay between posts. Unfortunately, life happens. Will try to do better next time.

As I said in my last post, the solution to the problem created by the current explosion in demand for electricity caused by AI has four time dimensions: (1) the right now; (2) the near future; (3) the intermediate future; and (4) the longer term future. These four dimensions in turn can be divided into two categories. The focus of the right now and near future dimensions is to retain as much existing conventional generation as possible while the focus of the intermediate future and longer term future dimensions is to facilitate the construction of new generation capacity needed to satisfy the increases in demand for electricity caused by AI. In this post I will discuss the first two dimensions. The next post will discuss the third and fourth dimensions.

Right Now

As the name suggests, the right now dimension describes actions that need to be taken right now, in 2025. It is not possible to instantly add new generation facilities or change electric markets, so there are only a few available options.

One option is for the Department of Energy (DOE) to continue to exercise its emergency authority under section 202(c) of the Federal Power Act to require generators slated for retirement to continue to produce electricity on a temporary basis. I wrote about this authority in more detail in my June 8 post entitled “Is Bringing Generation Facilities Back from the Grave a Good Approach to Grid Reliability?” DOE has used this authority to issue orders five times since May of 2025. Two of those orders required generation plants slated for retirement to remain in service.[1] Three other orders permitted generation facilities to temporarily exceed limits on their operations imposed by environmental regulations to the extent needed to prevent blackouts.[2]

State public utility commissions also have some authority over generation facilities owned by regulated utilities. Like DOE, state public utility commissions might, depending on their statutory authority, be able to prohibit or delay planned retirements of existing generation facilities owned by electric utilities if that generation is needed to serve projected peak loads. Unlike DOE, however, state commissions have no authority to authorize operations at levels above those permitted by federal environmental regulations.

Other than exercising these—and similar—temporary emergency powers, there is not much that can be done right now to ensure that the lights stay on.

Near Future

In contrast to right now, where we are stuck with the grid as it currently exists, it is possible to make changes that can help address the growth in demand in the near future. In this post I am going to focus on actions that can be taken that will help retain existing generation in the near future, before significant amounts of new generation capacity can be constructed. Actions also need to be taken in the near future to facilitate the construction of new generation capacity, but I will address those actions in my next post.

As an initial matter, the authority provided to DOE by Federal Power Act section 202(c) is an emergency authority that can have only a temporary effect. And DOE’s orders apply only to generation facilities that otherwise would be retired. DOE can continue to issue similar orders in the near future, but those orders represent nothing more than a band-aid that temporarily stops the bleeding but does not address the root cause of the bleeding.[3]

And just what is that root cause of the bleeding? The ultimate decisions about whether to construct new generators or to keep them in operation are economic decisions made by entities with access to the necessary capital to fund construction or continued operations of generation facilities. And the primary reason that owners are retiring their generation facilities is because they no longer are profitable enough to justify the incurrence of the costs (maintenance, fuel, capital improvements, etc.) required to keep them operating. In fact, many generation facilities that have been retired in recent years were operating at a loss for several years.

Consequently, the most important step that can be taken to prevent existing generation facilities from retiring is to ensure that the revenues they earn are sufficient to keep them in operation. The way this can be done varies depending on whether or not a generator is located in a Regional Transmission Organization (RTO) and, if so, what type of wholesale markets are operated by that RTO.

Competitive RTO Capacity Markets

Theoretically, the competitive RTO markets should be self-correcting. Prices in these markets should increase as demand increases until the price is high enough to attract the amount of generation needed to supply the full demand for electricity. Prices in the RTO energy markets have been dampened in recent years, however, by the very low prices offered into those markets by solar and wind generators who can submit these low offers because they have no fuel costs. This would not necessarily be a problem, except that solar and wind facilities provide very little capacity value due to the intermittent nature of their generation of electricity. The low prices in the energy markets therefore threaten the economic viability of conventional generators, which have a much higher capacity value, that are needed for system reliability.

For this reason, it is important that existing conventional generators in RTO markets also have, in addition to revenues from energy markets, access to adequate capacity payments to provide the necessary revenues to cover their costs and earn the returns on investment necessary for their owners to keep them in operations. This appears to be happening in those RTOs with competitive capacity markets, where prices have increased after a period of low capacity market prices.

For example, after the PJM capacity market yielded historically low prices in the three capacity auctions conducted from 2022 to early 2024, the capacity prices in the last two PJM capacity market auctions have been much higher—the two highest prices that market has ever yielded. The price for capacity delivered in the summer season for the MISO RTO was even higher in its most recent capacity auction. High capacity prices inevitably lead to a public backlash, particularly after a long period of low prices. But it is absolutely essential that the RTO capacity markets continue to yield the higher prices necessary to provide the revenues needed for existing generation to be kept in service.

Capacity Acquired Outside of RTO Capacity Markets

Not all RTOs have capacity markets. Some RTOs instead require utilities to acquire capacity by constructing it themselves or contracting directly with generation owners.[4] And outside of the RTOs there are no central wholesale electric markets. In these locations, local public utility commissions for the most part control the prices for capacity paid by utilities serving retail customers. But the essential imperative is the same—public utility commissions need to ensure that the owners of existing conventional generators have access to sufficient revenues to keep operating their generation facilities. This means that public utility commissions must allow utilities to enter into capacity purchase contracts with existing generators covering the next few years at prices that will allow the owner of those generators to continue to operate.

Public utility commissions must also set rates for sales of electricity by public utilities at levels that will encourage the continued operation of their existing generation facilities. Utility rates are set by establishing a return on investment for the utility and setting retail rates that allow the utility to recover all of its operating costs, plus earn that rate of return on all of its assets. If these rates do not allow a utility to earn an adequate return on its generation facilities, the utility will have an incentive to retire those facilities.

Unfortunately, the standard form of cost-based utility ratemaking can create an incentive for a utility to retire its older generation facilities no matter high a rate of return is allowed by the utility’s public utility commission. This can be illustrated by the graph below that was prepared by the Energy Bad Boys and that I discussed in my recent post, “Data Do Not Lie, But You Have to Listen Carefully to What They Say.”

This graph shows a comparison of the costs of different types of generation facilities owned by large electric utilities in the MISO RTO. I am not going to comment on what this graph says about the relative costs of renewable v. conventional generation; that was the topic of my Data Do Not Lie post. But this graph is also useful for illustrating how cost-based ratemaking works and the problems in incentivizing the continued operation of existing generators in a cost-based ratemaking context. I need to provide a bit of background to explain the problem, but my explanation is not too long, so bear with me.

If you look at the bottom of the graph, it shows the various different categories of costs that are considered in setting cost-based rates. Three of these—fuel, variable operations and maintenance, and fixed operations and maintenance—are operating expenses that cost-based ratemaking allows to be recovered. But utilities don’t make any money from this recovery. It’s like being paid for the cost of the gasoline you purchased on a trip. You are reimbursed for the cost of the gasoline you purchased, but you don’t make any money from that reimbursement.

This leaves two more cost categories shown on the graph: capital costs, and returns. The capital cost category of costs is simply the recovery over time in rates of the cost of constructing the generation facility (plus subsequent improvements to the facility). For example, if you paid $100 to construct a generator and were allowed to recover those costs over ten years, you could recover $10 of the costs every year for ten years. This recovery of capital costs, referred to as depreciation, is in some ways like one of the operating costs. It simply represents the recovery of a cost you previously incurred. The return category represents the profit a utility earns from its generator. The return earned on a generation facility is equal to the return authorized by utility commission times the remaining capital investment. So if the capital investment was $100 and the allowed return was 10%, you would get to include in your rates $10 in profit (10% of $100).

The problem of incentivizing the continued operation of old facilities results from the fact that the amount of depreciation each year is subtracted from the capital cost amount. So, in my example, in year one you recovered $10 of the $100 of the generator’s capital costs. The next year there are only $90 of capital costs. And that means instead of earning $10 of profit the next year, your 10% return equals only $9 (10% of $90) The following year the return will be $8, and then $7 the next year, and so on. If you don’t like numbers, think of it this way. Every year a utility earns less profit from its generation facility because the capital cost associated with that generator has been reduced. Old existing generation facilities are almost fully depreciated. This means that they have little or no capital costs, And, because return is calculated based on the amount of a generation facility’s capital costs, little or no capital costs means little or no profit for the facility’s owners.

This effect is illustrated by the chart prepared by the Energy Bad Boys, which I pasted above. Unlike my hypotheticals, this chart is based on actual cost data of large midwestern electric utilities. The renewable generators shown on the right side of the chart are relatively new and, as a result, have relatively high capital and return costs. This means that new wind facilities subject to cost-based ratemaking would earn almost $20 of profit for every megawatt-hour of electricity they produce, and new solar facilities would earn over $30. By contrast, the existing nuclear and coal plants—which tend to be the largest plants owned by a utility, generating the most electricity—have completely depreciated and their owners get no return whatsoever from their operations. The other types of conventional gas-fired generation have some undepreciated capital costs and therefore earn some returns for their owners, but these returns are much lower than the renewable generators’ cost-based returns.

Older generation facilities produce significant amounts of electricity and are important to maintaining the reliability of the grid. But their owners are rational profit-maximizing companies. The owners’ incentive is to retire these facilities when they no longer operate profitably. Developing rate mechanisms that give utilities an incentive to keep operating them would be helpful to ensure they remain in service until enough new generation capacity can be constructed.

It is important to note that this problem exists only for existing generators operating under a cost-based rate regime. The revenues earned by existing generators participating in RTO markets do not depend on the amount of remaining undepreciated capital costs, but instead simply on the energy and capacity prices established in the RTO markets. This means that they can operate profitably when the RTO market prices are high enough regardless of whether or not they are fully depreciated.

I said at the outset that economics represents a major factor in the retirement of existing conventional generation facilities. But economics is not the only factor. Some states have enacted laws mandating that fossil units be retired and replaced by renewable generators. This perhaps made sense when demand levels were flat; or perhaps not. But it no longer makes sense today, when the demand for electricity is exploding because of the need to supply electricity for AI. It is imperative that the states with such policies re-evaluate them in light of this increased demand.

In saying that we need to retain existing conventional generation, I do not suggest that the transition to a low carbon grid be terminated and reversed. But the pace of that transition needs to be re-evaluated. If there are blackouts because fossil generators have been replaced by wind and solar generation facilities too quickly, then there will be significant political repercussions. And as a result, the transition may never happen.

***

I hope you enjoyed this post. I enjoy writing them and will never charge for subscriptions or ask for donations. I only ask that, if you did enjoy it, you press the “like” button below. Doing so will help me evaluate interest in the book I am writing on grid operations. Of course, if you have a reaction to, or question about, this post, please leave a comment and I will be happy to respond.

[1] These orders can be found at Midcontinent Independent System Operator (MISO) 202(c) Order_1.pdf ; Federal Power Act Section 202(c) PJM Interconnection.pdf.

[2] These orders can be found at PREPA 202(c) Emergency Measures Generation.pdf; Duke Energy Carolinas 202(c) Order_1.pdf; PJM's 202(c) Order No. 202-25-6.pdf

[3] As I explained in my Bringing Generation Facilities Back from the Grave post, Executive Order 14262 instructs DOE to use its Federal Power Act section 202(c) authority to prohibit the retirement of any generation facility in a region with an inadequate capacity reserve that has a capacity of 50 megawatts or more.

[4] These RTOs are SPP (Southwest Power Pool) and the California ISO. In addition, ERCOT—the RTO operating in Texas—does not impose any capacity obligation at all.

Last night I attended a performance at Wolftrap, which is the open air performing arts center near Washington DC, where I live. We listened to the National Symphony Orchestra (NSO) playing the musical score from the 1985 movie, Back to the Future, which was displayed on several large screens as the NSO was performing. Back to the Future is a fun movie starring Michael J. Fox, which I highly recommend if you have not seen it. I am not going to describe the plot, except to note that it involves time travel in a Delorean sports car, which requires 1.21 gigawatts of electricity to break the time barrier. At the end of the movie, the car returns to 1985 from 30 years in the future (2015), where it has been outfitted with a “Mr. Fusion Home Energy Reactor" from that time, capable of converting a banana peel, a beer can, and some coffee grounds into the necessary 1.21 gigawatts of electricity required for the next jump in time.

Too bad that no Mr. Fusion Home Energy Reactor has yet been invented. It would provide the perfect solution to the most pressing issue faced by the grid today; the rapid increase in the demand for electricity caused by the need to supply energy-hungry data centers that support AI and crypto mining. I have already devoted one post to this issue, entitled “How Explosive Load Growth Will Affect the Grid.” But this is an important issue that deserves a closer look and, so, I have decided to write a series of posts discussing various aspects of the problem. This post will go into more depth to explore exactly what the problem is. In subsequent posts I will explore ways of dealing with the problem.

Source of the Problem

As I explained in How Explosive Load Growth Will Affect the Grid, the recent rise in demand for electricity comes after several decades of an almost flat demand for electricity. My post included a copy of the following chart, which was prepared by GridStrategies, showing demand growth rates from the 1950s to 2024.[1] As this chart shows, after very large increases in demand for electricity in the 1950s and 1960s, the increase in demand for electricity started to decline in the 1970s and has been relatively flat for the last three decades.

Because demand has been fairly stable, electric systems have been able to provide reliable service without constructing significant amounts of new generation capacity. The almost flat demand for electricity also facilitated the push towards replacing existing fossil generators with new wind and solar facilities, starting in earnest around 2010.

The relatively flat demand for electricity also has influenced the competitive Regional Transmission Organization (RTO) wholesale electric markets. These markets supply about two-thirds of all electricity sold in the US. The RTO markets provide for the sale of relatively short term products; no longer than one year. Short term prices can—depending on the price level—support continued operation of existing generation facilities. Consequently, the RTO markets have supported the retention of enough generation capacity to satisfy the flat demand levels that have existed for much of the history of those markets. The large investments required to construct new facilities, however, frequently require multi-year price certainty. And today, the RTO markets do not provide such price certainty.

Moreover, the prices in the RTO energy markets have been suppressed by two independent factors. First, the price of natural gas started to decline dramatically starting around 2009, as fracking technology provided access to plentiful amounts of gas from shale formations located in various regions of the country. Second, wind and solar generators have been able to offer their electricity into the energy markets at low prices because they do not incur any costs to purchase fuel. With federal tax credits, these generators are even able to profitably offer into the markets at negative prices (e.g. they pay to sell their electricity). The submission of these very low offers into the energy markets can have the effect of reducing the prices resulting from those markets.

Given these circumstances, it is not surprising that the total amount of capacity of conventional generation in the US (coal, natural gas, oil, nuclear, and hydro) has declined by about 7.5% since 2010. Hardest hit has been coal generation capacity, whose economics worsened significantly after the decline in natural gas prices. The total amount of coal generation capacity has declined by about 45% since 2010. Natural gas generation capacity has increased by about 25% in the same time period, but most of that increase took place in the first few years. There has been only a small increase in natural gas generation capacity since 2022.[2]

This trend of gradual reductions in conventional generation capacity, coupled with large increases in renewable generation capacity, might not have caused reliability concerns had the demand for electricity remained flat. But with data centers’ increased use of electricity, demand has not remained flat. It has skyrocketed. The following chart, also prepared by GridStrategies, shows that the demand for electricity is projected to increase by 15% over the next five years.[3]

The unexpected rapid growth in demand, coupled with a reduction in conventional generation capacity, is already causing a strain on the grid. This is illustrated by the following graphic, prepared by the North American Reliability Corporation (NERC) that was included in the FERC Staff’s presentation of its 2025 Summer Energy Market and Electric Reliability Assessment.[4] NERC is the entity statutorily responsible for ensuring the reliability of the US grid. This graph shows there are four areas in the US in which NERC has determined there is an elevated risk of supply shortages for stressed summer conditions this summer due to insufficient operating reserves.

These areas are: (1) NPCC-New England, which covers the ISO-New England RTO; (2) MISO, which covers the northern part of the MISO RTO; (3) MRO-SPP, which covers the SPP RTO; and (4) Texas RE, which covers the ERCOT RTO. It is important to note that NERC’s assessment applies only to the summer of 2025, and not to future years.

The Problem

It’s not hard to see the problem. The declining conventional generation capacity combined with a sharp increase in demand has created the potential for the grid’s electric systems to be unable to supply all demands for electricity. And supply failures could happen as soon as this summer. So far, there have not been any major blackouts caused by a shortage of operational generation capacity. All electric demands have been met even through the recent extended periods of very high heat throughout much of the country. This includes hot temperatures in the four areas highlighted by NERC as having an elevated risk of supply shortages. And to be clear, the elevated threat of supply shortages applies only to extreme circumstances when: (a) the demand for electricity is very high; (b) there are “historical high generator outage rates;” and (c) there is low output from “variable energy resource[s],” i.e. wind and solar generators. No one is predicting blackouts anywhere in the country under ordinary summer conditions, at least not in the summer of 2025.

But summer is not yet over, and the demand for electricity to supply data centers is projected to increase significantly every year for the next few years. Even though there have not been any supply shortages to date does not mean there won’t be any in the future. And if something is not done, those shortages may not be limited to the extreme circumstances that could lead to blackouts this year.

The Solution

If we want to assure an adequate supply of electricity, we need to ensure that we have an adequate supply of conventional generation capacity. We can, and do, argue over whether wind and solar generators provide any capacity value. But no one can reasonably argue that wind and solar can provide the same level of reliability as conventional generators, at least today or in the foreseeable future.

This task will be complicated by a couple of factors. First, there is no central entity with the authority to determine the scope of the problem or what steps must be taken. The federal government has no general authority over the construction of new generation facilities. The government does have authority to issue certain types of permits, such as permits authorizing the construction of nuclear reactors and permits authorizing the location of generation and transmission facilities on federal lands. But that authority does not extend to deciding what types of generation facilities should be constructed or, indeed, if any new generation facilities should be constructed at all. Instead, that authority is vested in the states. And each state has its own authority to decide what, if any, new facilities should be constructed to address the problem and, indeed, whether there is a problem at all. It will be very difficult to get all states to agree to take consistent action to address potential supply shortages.

Second, not even the states have the authority to order any particular entity to commit the significant funds required to construct new generation facilities. Yes, federal and state subsidies can be granted to encourage the construction of certain types of new facilities. And each state can authorize the recovery of the costs of certain facilities through electric rates. But at the end of the day, private entities must decide to make the necessary investments to construct new generation facilities based on the expected economics of those facilities. It is economics that ultimately explain why retirements of conventional generation facilities have outstripped the construction of new facilities in the past, and it is economics that will ultimately determine whether adequate conventional generation capacity is constructed and remains in operation in the future.

As I see it, the solution to the problem created by the current explosion in demand has four time dimensions: (1) the right now, i.e., what must be done immediately to ensure the adequacy of the supply of electricity this year; (2) the near future, i.e., what must be done to ensure the adequacy of the supply of electricity in the next couple of years before significant new generation capacity can be constructed; (3) the intermediate future, i.e., what can be done to quickly increase the construction of new generation capacity as demand is rapidly increasing; and (4) the longer term future, i.e., what makes the most sense in the long term for supplying all demands for electricity, including data centers. I will address these dimensions in future posts.

***

I hope you enjoyed this post. I enjoy writing them and will never charge for subscriptions or ask for donations. I only ask that, if you did enjoy it, you press the “like” button below. Doing so will help me evaluate interest in the book I am writing on grid operations. Of course, if you have a reaction to, or question about, this post, please leave a comment and I will be happy to respond.

[1] GridStrategies, Strategic Industries Surging: Driving US Power Demand, at p 5 (December 2024), Strategic Industries Surging: Driving US Power Demand.

[2] This data on generation capacity was taken from Table 4.2.A of the Energy Information Administration’s Electric Power Annual.

[3] GridStrategies, Strategic Industries Surging: Driving US Power Demand, at p 3 (December 2024), Strategic Industries Surging: Driving US Power Demand. In an update in February of 2025, GridStrategies revised the projected load growth slightly downward from 128 GW to 116 GW, The Era of Flat Power Demand is Over.

[4] 2025 Summer Energy Market and Electric Reliability Assessment, Slide 8 (May 15, 2025). Summer Assessment 2025 | Federal Energy Regulatory Commission. The actual report presenting the 2025 summer assessment can be found at FERC Releases 2025 Summer Assessment | Federal Energy Regulatory Commission.

I’m finally fulfilling my promise from a couple of posts ago to write about something other than renewables. Instead, I am going to write about the deals that data center owners have been entering into with generation owners to purchase electricity from specific generation facilities. The deals primarily have involved purchases of electricity from nuclear power plants, but I also have seen one or two announced deals between data center owners and natural gas-fired generators.

Every announcement I have seen explains the deals as being motivated by the desire of data center owners to secure a more reliable source of electricity. On its face, this explanation makes perfect sense. Reliability is extremely important for data centers; an interruption in power supply could have catastrophic consequences. But when I think about it, I can’t make sense of this explanation.

We read that data center owners want to contract with nuclear power plants because nuclear plants are the most reliable of all types of generators. They can continuously operate at a high level of output for long periods of time. This rationale is correct as far as it goes. It is true that nuclear reactors have by far the highest capacity factors of any type of generation technology. This is why they make excellent baseload generators. There are less than 100 nuclear plants operating in the US today, yet they supply about 20 percent of the nation’s electricity.

But the capacity factors of nuclear plants, although high, are not 100 percent. Nuclear reactors have unexpected outages on occasion, and all existing reactors have to be taken offline—sometimes for over a month—to be refueled. If the goal is to have an electricity supply that is never interrupted, this cannot be accomplished by contracting with a single nuclear power plant. And contracting with individual natural gas plants—with much lower capacity factors—makes even less sense.

Instead, if reliability is the concern, it makes much more sense to interconnect with an electric system that includes multiple generation facilities. For example, there are thousands of generators that participate in the PJM Regional Transmission Organization (RTO) energy markets. From these generators, PJM is able to provide a supply of electricity that is almost never interrupted. PJM acquires more generation capacity than its expected peak demand, so it is able to supply electricity at all times, even when demand is high and some generators are offline and unavailable. Other RTOs and utility systems have fewer numbers of generators. But every RTO and utility system has access to more generation capacity than it anticipates it will need to serve its peak demand. These systems likewise can supply electricity more reliably than can be provided by a single generation plant.

Why, then, are these deals being done? Perhaps it is because the deals are motivated more by a desire to acquire a large supply of electricity from non-emitting generators than they are to improve reliability through direct interconnections with a specific generation facility. Many of the owners of data centers have publicly expressed the goal of relying solely on electricity provided by generators with no carbon emissions. Originally, this meant acquiring electricity from renewable solar and wind generators. However, the recent explosion of the data centers’ demand for electricity fueled by AI cannot be fully satisfied by renewable generators. Consequently, data center owners have started to turn to nuclear generators.

The data owners appear to have decided to achieve their no emissions goals by entering into contracts that are similar in some respects to so-called “green energy” contracts offered by some retail electricity suppliers. These retail suppliers purchase electricity from wind and solar generators who deliver their electricity into an electric system. The green energy suppliers then arrange with the system operator to have that electricity delivered to their customers at different locations on the system. The electricity physically delivered to the retail customers most likely is not entirely, or even mostly, produced by renewable generators. The bottom line, however, is that the amount of renewable electricity delivered into the electric system equals the amount of electricity delivered to the customers of the green energy supplier.

Similarly, data centers can interconnect with an electric system operated by a large RTO or electric utility. The data center owners then contract to purchase electricity from a nuclear power plant that is interconnected with the same system. Under these contracts, the nuclear power plant delivers at least the same amount of electricity into the system that the data center takes off. Not all of the electricity physically delivered to the data center will have been generated by the nuclear power plant, but as a contractual matter that plant will have delivered its electricity to the data center. And if, for some reason, the nuclear power plant is unable to deliver the amount of electricity required under its contract, the data center can rely on the other power plants in the same electric system to provide it with electricity. In this way, the data center owner can satisfy its goal of purchasing emission free electricity while also securing a very reliable power supply. And by agreeing on a fixed price for electricity, a data center owner also can protect itself from future increases in the price of electricity.

A good example of this type of transaction is the power purchase agreement between Amazon and Talen Energy announced recently this June. Under this contract, Talen will supply Amazon with 1,900 megawatts of capacity from its Pennsylvania-based Susquehanna nuclear power plant. Electricity generated by the Susquehanna plant will be delivered into the PJM transmission system in amounts equal to the electricity delivered from the PJM system to multiple Amazon data centers in Pennsylvania. As a contractual matter, all of the electricity delivered to the data centers will have been generated by the Susquehanna plant, even though as a physical matter much of that electricity will have been generated by other generators, including fossil facilities.

The desire to rely solely on electricity generated by non-emitting sources does not explain the deals announced to purchase electricity from specified natural gas generators. Quite frankly, I don’t understand those deals. Natural gas generators are not as reliable as nuclear generators, and I am not aware of any reason why a data center owner would want to ensure that its electricity is produced from natural gas.

Further, I should note that some data centers have announced deals that involve purchases from two or more small modular reactors (SMRs), which are nuclear reactors with a capacity of about 500 megawatts (as opposed to large nuclear reactors which frequently have capacities of about 1,000 megawatts). Contracting with multiple SMRs should increase the reliability of the supply of electricity because the SMRs are able to back each other up. But it would be even more reliable to purchase electricity from a system that is made up of thousands, or hundreds, or even tens of generation facilities.

Additionally, contracting with multiple SMRs would be an extremely expensive approach to ensure reliability. Yes, the total cost of an SMR is less than that of a new large reactor. Based on cost estimates for SMRs I have seen, however, I don’t think an SMR’s cost per megawatt-hour of electricity could be much less expensive than the cost per megawatt-hour of electricity produced by a new large nuclear reactor. Indeed, the cost of electricity produced by a new SMR almost certainly would be more expensive than the cost of electricity produced from an existing large nuclear plant whose capital costs have been largely depreciated. And because SMRs should operate at a high capacity factor, most of the time the SMR providing backup service will not be needed to serve the data center. Paying for a second SMR simply to provide backup service that is rarely needed seems like quite a high price to pay. Purchasing electricity from an electric system would be both more reliable and less expensive.

In the last two paragraphs, I have admitted that I can’t explain the rationale for at least two of the types of deals that data centers enter into with generation owners. Sorry, but I just can’t figure out how the natural gas and multiple SMR approaches make sense. There are, however, two conclusions I believe you can legitimately take away from this post. First, contracting to directly purchase all of a data center’s electricity requirements from a single generation facility does not make sense as a strategy to improve the reliability of supply, even when the purchase is from a nuclear power plant. Second, at least some of the announced data center supply deals can be better understood as motivated by a desire to acquire a supply of electricity from non-emitting nuclear generators than by a desire to improve the reliability of supply.

***

I hope you enjoyed this post. I enjoy writing them and will never charge for subscriptions or ask for donations. I only ask that, if you did enjoy it, you press the “like” button below. Doing so will help me evaluate interest in the book I am writing on grid operations. Of course, if you have a reaction to, or question about, this post, please leave a comment and I will be happy to respond.

This past Saturday (July 19), the Energy Bad Boys (Bad Boys) posted an article entitled “MISO's Existing Nuclear, Natural Gas, and Coal Plants are Way Cheaper than New Wind and Solar.” As is usual, the Bad Boys have presented an in-depth analysis based on detailed objective data. This time, the data they analyze relates to the costs and other information about the ten largest investor-owned utilities in the Midcontinent Independent System Operator (MISO) region. In their article, the Bad Boys say that “the data don’t lie.” And while it is true that their data aren’t lying, the story being told by those data is not as significant as the Bad Boys suggest. Instead, the data tell a very unsurprising story that has only limited relevance to what’s going on in the grid today.

As the title of the Bad Boys’ article suggests, they have presented a comparison of the costs of existing nuclear, natural gas, and coal plants with new wind and solar plants. The results of that analysis are summarized in a chart they included in their article. I have copied this chart below so that you can understand what I am going to be writing in my article.

As you can see, the Bad Boys’ chart shows average cost calculations for each of several different types of generation facilities, broken down into different categories of fixed and variable costs: capital costs, fuel, variable operating costs, fixed operating costs, and return on investment. The chart purports to demonstrate that new solar is more expensive than all existing generators, and new wind is more expensive than all existing generators except for combustion turbines.

It is Not Surprising that the Costs of New Generation Plants Are Higher than the Costs of Older Existing Plants

Why is this not surprising? Those of you who have been reading my posts for a while know that I like to use analogies to explain my points and I am going to use another one here. I own a very nice Audi convertible that I bought back in 2013. It looks great and I get lots of compliments about it. Since I have had my Audi for almost 12 years it is fully paid off and my chief operating expense is for gas.

Let’s compare the cost of owning my Audi today to what I would have to pay if I bought a new Nissan Versa, one of the least expensive cars on the market, well below what I paid for my Audi 12 years ago. The Versa also has considerably better gas mileage than my Audi. Despite the fact that the cost of purchasing a Versa is less than the price I paid for my Audi and the Versa gets better gas mileage, it is not hard to see that keeping my Audi is the least cost option. If I buy the Versa, I will have to make a down payment and pay hundreds of dollars a month in loan or lease payments, something I can avoid by keeping my Audi. Yes, I would have to pay less for gas if I buy the Versa, but those cost savings would be overwhelmed by the much larger amounts of the Versa’s loan or lease payments.

To be clear, the fact that the new Versa will cost me more to own than my older Audi doesn’t mean that the Versa isn’t cheaper to buy or less expensive to operate. It simply means that I have not yet paid the cost of purchasing the new Versa but have already paid the cost of purchasing my older Audi.

Essentially the same thing is going on in the Bad Boys’ cost comparison. As they acknowledge, “[t]he existing nuclear, coal, and natural gas plants are incredibly affordable because they have already paid off almost all of their upfront capital costs, and utilities no longer earn much in the way of a rate of return on these depreciated assets.” And you can see this on their chart. There are no capital cost or return costs (return costs are related to the amount of capital costs) shown for coal or nuclear plants and relatively low capital and return costs are shown for the other existing plants.

By contrast, capital and return costs represent the overwhelming share of the costs for new solar and wind facilities shown on the Bad Boys’ chart. About 85% of the total cost of new solar on the chart consists of capital and return costs, as are about 74% of the costs of new wind. That the total costs of new wind and new solar are higher than the costs of existing plants that are mostly or completely depreciated says nothing about which type of generation is less expensive to construct or operate. As a result, that the existing plants cost less when capital costs and return are considered shouldn’t be any more surprising than the fact that it would cost me more to buy a new Versa than to continue to own my fully paid off Audi.

The Bad Boys’ Analysis is Largely Irrelevant to Grid Operations

Let me illustrate my next point by adding my wife’s car to my analogy. The exact model, year, and cost of her car is not important here, only that our family owns two cars. When we are deciding which of our two cars to drive for our next vacation, we make the decision based only on the advantages and disadvantages of our two cars, including perhaps which would be cheaper to drive. But whatever we do consider, what we don’t consider is what it would cost to own a new Versa. If we are not looking to buy a new car, the costs of the Versa are completely irrelevant.

In the same way, the costs of new renewable solar and wind facilities are not relevant to what choices system operators make today about which generators to dispatch. System operators must choose between the options available to them today in operating their systems today. The cost of new solar and wind facilities is not relevant because those facilities have not yet been added. The analysis in the Bad Boys’ post says nothing about the decisions by system operators as to which generators to dispatch.

OK, now in my analogy (but not in my real life) our daughter has been bugging my wife and me to buy a car for her to drive. Today, we have to drive her to school and to her numerous after school activities, which is getting to be a pain, so we agree buying a new car is a good idea. In considering what car to buy for her, we consider a number of factors, the most important of which are safety, reliability, and cost. What we do not consider is how the cost of buying and owning a car for our daughter compares with the cost of the cars we already own. That simply has no bearing on which new car we should buy.

Moving away from my analogy, the most pressing issue for the grid today is how to meet the exploding demand for electricity caused by data centers. It appears that demand is on the verge of outstripping current supply, and new generation capacity is going to be needed. Just as my wife and I consider a number of factors in deciding what new car to buy for our daughter, a number of factors need to be considered in deciding what kinds of new generation facilities should be built to satisfy the rapidly increasing demand for electricity. The relative costs of different types of facilities is important to this decision, but other relevant factors include the time to construct, and the safety and reliability of the new generation. Another important factor, at least to some, is the emissions that will be released by the new capacity. What will not be considered—because it has no relevance—is whether the costs of existing capacity are less or greater than the costs of new capacity. Again, the Bad Boys’ analysis is of no help in answering the question of what type of capacity should be added to the grid.

The Data Are Telling Us the Truth About Other Issues

Some other truths that were not mentioned in the Bad Boys’ article can be gleaned from their data. First, the data illustrates the point I have made in many of my articles that solar and wind generators can produce electricity more cheaply than conventional generators. The Bad Boys’ graph shows that solar and wind have no fuel costs and variable operating costs, which are the two categories of variable costs incurred to generate electricity. By contrast, all other types of generation shown on the chart, even nuclear generators, incur significant variable costs in generating electricity.

I think this is an important point, so I have made a chart comparing the combined fuel and variable operating costs for each type of generation shown on the Bad Boys’ chart (I do not show the costs for wind and solar on my chart because wind and solar have no fuel or variable operating costs). The exact level of fuel and variable operating costs are not shown on the Bad Boys’ charts, so I had to estimate those costs by eyeballing the graph. As a result, the amounts shown on my chart are approximate, but I believe they are reasonably close to what the Bad Boys data shows.

As this chart shows, the cost of electricity generated from conventional generators ranges from $17/megawatt-hour for nuclear reactors to $33/megawatt-hour for combustion turbines. This is significantly more than the cost of electricity generated from wind and solar generators, which is $0. Wind and solar generators do incur cost of capital, return on investment, and fixed operating costs, but those costs are all incurred regardless of the amount of electricity, if any, those generators produce.

My chart also shows the potential for renewables to reduce the prices charged in MISO’s energy markets. I explained in detail how this can happen in my most recent post (Renewable Generation Has an Outsized Effect on RTO Energy Market Prices), and am not going to repeat that explanation here. Suffice it to say that the energy market price is based on the marginal price offer submitted to the market. And that price applies to all energy sold in the market, not just to the energy produced by the generator submitting the marginal cost offer. This means that, if low cost offers from renewable generators reduce the market clearing price below an offer submitted by a combustion turbine, the price for energy paid by every customer in MISO would drop from $33/megawatt-hour, based on offers submitted by combustion turbines, to $28,based on offers submitted by coal generators, which are the next-highest price offers. And the price potentially could drop further to as low as $17 if renewable generation price offers cause nuclear units to marginal cost units.

I see in re-reading my post that perhaps my discussion is too critical of the Bad Boys’ analysis. Their analysis does do a good job of putting the lie to the claim made by some wind and solar advocates that we could reduce the costs of producing electricity on the grid by immediately replacing all existing fossil generation capacity with renewable generators. Doing so would increase, not decrease, the cost of generation on the grid during those frequent times when the grid was not shut down by blackouts. But other than refuting this non-serious claim, I don’t think that the Bad Boys’ analysis adds much to a serious analysis of the grid’s most pressing issues.

***

I hope you enjoyed this post. I enjoy writing them and will never charge for subscriptions or ask for donations. I only ask that, if you did enjoy it, you press the “like” button below. Doing so will help me evaluate interest in the book I am writing on grid operations. Of course, if you have a reaction to, or question about, this post, please leave a comment and I will be happy to respond.

I swore to myself after last week’s post that I would hold off for a while on writing about renewable generation. Many of my recent posts have covered this topic, but there are other interesting issues involving the grid that also are worthy of discussion. After making this promise to myself, I realized there was one more point I wanted to make, so here is yet another post on renewables. I didn’t make the commitment not to write about renewables to anyone other than myself, so the only one who can complain about this is me.

Up to now, my posts have focused on renewables’ costs and their effect on the costs of electric systems. Today, I am going to discuss the effect of renewables on the prices for electricity charged by Regional Transmission Organizations (RTOs). It’s a related issue; after all prices for electricity have some relationship to costs. But the effect of renewables on RTO energy prices is quite different from their effect on costs.

In order to understand my analysis, it is necessary to understand a couple of aspects of grid operations and how RTO energy prices are determined. I have discussed these issues in past posts, with the greatest detail provided in my five-part “Explaining the Grid” series of posts. However, not everyone has read my prior posts and, for those of you like me with short term memory problems, you might not remember what I wrote even if you read those prior posts. So I am going to do a quick review, aided by some basic graphics that are (barely) within my limited graphics capability.

. Let me start with two graphs I provided in one of my earliest posts, Explaining the Grid, Part One.

As explained in detail in my earlier post, this first graph shows how the demand for electricity varies during each day. Demand is at its lowest early in the morning, and increases later in the day as people wake up and engage in their daily activities. Demand peaks in the mid-afternoon when temperatures are at their highest. Demand for electricity then decreases as people return home and eventually go to bed, at which time demand returns to approximately the same low level late at night.

This second graph shows how, until recently, the grid was designed to supply these varying levels of demand. Large baseload generators, mostly fossil or nuclear, operated around the clock to supply the base level of demand that exists 24/7. Smaller intermediate generators operated during most of the daylight hours, and yet even smaller peaking generators operate during the highest hours of demand. The variable cost (primarily the cost of fuel) to operate these generators was lowest for the baseload units, second lowest for the intermediate units, and most expensive for the peaking units. Operating the least expensive units as much as possible and the most expensive units as little as possible was, and continues to be, the most efficient, least expensive approach to supplying electricity to consumers.

Today, we also have renewable generators. These generators are not capable of supplying baseload service because they cannot operate on a 24/7 basis. As a result, they are in some respects like the peaking units or, perhaps, the intermediate units shown on the second graph. But there is one major difference. Unlike intermediate units and, especially peaking units, renewable generators are not expensive to operate. To the contrary, renewable generators are very inexpensive to operate because they do not require fuel to generate electricity.

With this basic understanding of how different types of generators operate to meet the demand for electricity as it varies throughout the day, lets now look at how RTOs set prices for energy. Prices are established through competitive auctions, in which each potential seller submits an offer to sell electricity at a stated price. As I explained in my earlier posts, these offers tend to be at or near the variable operating cost the generator incurs to produce electricity. RTOs then establish their prices based on these offers, using what is called a stack. Here is a graphic that is useful in conceptualizing how a stack works. It is similar to graphics I used in Explaining the Grid, Parts Three and Four.

This illustrative stack includes four generation facilities. The two submitting the lowest cost offers ($20 and $30) are submitted by baseload units, which tend to have the lowest operating costs of all conventional generators. The facility submitting the $50 offer is an intermediate facility. And the $90 offer is submitted by a peaking unit, which tends to have the highest costs in an electric system.

The graphic also shows that customer demand at the peak time of demand (i.e. the of the greatest demand for electricity on a particular day) is 950 megawatts. This means that the RTO will need to select 950 megawatts of capacity out of the 1,000 megawatts of offers submitted. This selection is done using the stack, in which the offers are stacked in price order, with the lowest price offer ($20) at the bottom and the highest price offer ($90) at the top. The RTOs then see how far up the stack they have to go to meet customer demand at the time the auction is run. In this case the auction is run at the time of the highest demand of the day, which means that 950 megawatts of generation capacity is needed to meet the demand. In order to meet this large amount of demand, the RTO has to select 50 megawatts of the $90 peaking facility, plus all 900 megawatts of capacity whose offers were lower than $90.

Now for the part that is crucial for understanding how renewable generators affect RTO prices. The highest price offer for electricity that was needed by the RTO to supply customer demand (the $90 offer submitted by the peaking unit) is considered to be the marginal, or market-clearing offer. AND ALL GENERATORS SELECTED TO SUPPLY ENERGY ARE PAID THE $90 OFFER EVEN THOUGH ONLY 50 MEGAWATTS OUT OF 950 MEGAWATTS OF CAPACITY WAS OFFERED AT $90. The units offering at $20, $30, and $50 are all paid the same $90 as the peaking unit submitting the $90 market-clearing offer. There are good reasons for this counterintuitive result, as I describe in Explaining the Grid, Part Four.

Why is this fact crucial for understanding the rate effect of renewables? Let me explain with a graphic almost identical to the one above, except that I have added an offer submitted by a 75 megawatt solar generator. Solar generators have low variable operating costs because they do not need fuel to generate electricity—they generate electricity from sunlight. As a result, the solar generator is able to offer its electricity at $1.

As this graphic shows, the $1 offer is the lowest cost offer, so it is placed at the bottom of the stack. This moves all other offers up the stack, with the result being that the $50 offer is now the marginal market-clearing offer, and the price of all electricity is set at $50 instead of $90. This represents a tremendous rate saving for customers. Without the solar panels, the total rate paid for the 950 megawatt-hours of electricity demand is $85,500 ($90 x 950). With the addition of just 75 megawatts of solar panels, the total rate paid for 950 megawatt-hours of electricity is reduced to $47,500 ($50 x 950). This is an approximately 45% rate reduction enjoyed by all of the RTO’s customers resulting from an offer representing less than eight percent of the total amount of energy sold in the RTO market.

I should note it is very unlikely there ever would be a price difference of $40 between two offers in a stack, as I showed in my hypothetical. However, the cost differences between offers in RTOs are greater at the top of the stack than they are between lower cost offers lower down in the stack. I used this large price difference to highlight the fact that a relatively small number of low cost offers can have an out-sized effect on the price in the RTO energy markets during peak demand periods. Moreover, the size of the RTOs is much greater than the 950 megawatts shown in my hypothetical. The peak demands of RTOs range from about 20,000 megawatts to over 130,000 megawatts. Consequently, the reduction in the total price paid for all energy sold in an RTO energy market is much greater than I have shown, even if the reduction in price per megawatt-hour is much less than in my hypothetical.

I decided yesterday to do a bit of a reality check on my theory that renewables can play an important role in dampening the high energy prices that can result from the high demand that exists in hot weather. Yesterday was not as hot in most parts of the country as it has been for much of the last week, but it still was a reasonably hot day. Each of the RTOs has real time data displayed on the first page of its web site—updated every five-ten minutes— that shows the currently operating generation mix, prices, demand, and other data.[1] I looked at this data yesterday (Sunday, July 13) for five of the seven RTOs (all but New York ISO and ERCOT) at four different times; approximately 10 a.m., 1 p.m., 3 p.m., and 6 p.m. For each of these times I noted the generation mix—particularly the amount of solar and wind generation—and the prices prevailing in each RTO.

Below, I am showing the screen shots I took of the PJM website. Each website is somewhat different, but these screenshots are representative of what the other RTO websites also showed.

1. PJM 10:18

2. PJM 12:45

3. PJM 3:16

4. PJM 5:40

As you can see, the web sites don’t show the prices at which generators offer their energy. As a result, I couldn’t directly check whether low cost renewable price offers were being made that pushed high cost offers from peaking units above the marginal cost offer, as shown on my illustrative graphic. But I could see that the amount of electricity produced by renewables generally increased as demand increased. The screenshots above show that, in PJM, the amount of renewable energy produced in PJM was at 8,164 megawatts at 10:18 a.m. when demand was a bit over 105,000 megawatts. The total amount of renewables rose to 13,807 megawatts at 5:40 p.m., when demand had risen to over 133,000 megawatts. Similarly, the total amount of renewables rose to over ten percent of all electricity being produced in three out of the other four RTOs, and over nine percent of all electricity produced in SPP, the other RTO I was monitoring.

Most, if not all, of this renewable energy offered into the RTO energy markets was likely offered at relatively low prices. Consequently, it seems almost certain that energy prices were dampened below what they would have been if no renewable energy had been produced. This conclusion is supported by the fact that, even though it was a relatively hot day, the price increases in every RTO as the temperatures increased were for the most part quite modest. For example, you can see from the PJM screenshots that prices were generally in the $30s at 10:18, and had increased only to the $50s at 5:40 even though demand had increased substantially. There were a few pockets where prices spiked higher, but these were caused by transmission constraints that prevented lower cost energy from being transmitted into the high cost area. It seems likely that prices in the RTO energy markets would have been higher if there had been no renewables in the generation mix and instead high cost peaking units had been used to meet demand.

And it is important to keep in mind the RTO market approach to pricing, where every generator is paid the market-clearing price. If, as the price data suggests, low cost offers from renewables prevented the market clearing price from rising higher, that lower price did not apply only to the energy produced by renewable generators. It also reduced the price paid to the other generators, who produced about 90 percent of the electricity sold in the RTO markets.

I have read arguments to the effect that the rate-reducing effect of renewable generators is due to the federal tax credits paid for each megawatt-hour of electricity wind and solar generators produce. There is some truth to this argument during periods of very low demand during the night, when RTO energy prices sometimes plunge down to, and even below, zero (i.e. generators must pay the RTO for each megawatt-hour of energy they deliver). In those low demand periods, the tax credit makes the production of energy profitable even when the RTO price for energy falls below zero. But most of the time, energy market prices are positive. At these times, renewables can offer their electricity in the RTO energy markets at prices close to zero even without a tax credit. This is because renewables do not incur the cost of fuel to generate electricity.

We can argue about whether the price-suppressive effects of renewables on RTOs is a good thing or a bad thing in the long run. As I have noted in other posts, low RTO energy prices have created economic pressures on large baseload generation facilities that are necessary to ensure the grid’s reliability. Serious reliability issues will result if nothing is done to provide the necessary economic support for the continued operation of existing baseload generation and the construction of new baseload generation. But it cannot be reasonably disputed that, at least in the short term, renewable generators can provide an outsized rate benefit in the RTO energy markets that extends far beyond the energy renewable generators produce themselves.

***

I hope you enjoyed this post. I enjoy writing them and will never charge for subscriptions or ask for donations. I only ask that, if you did enjoy it, you press the “like” button below. Doing so will help me evaluate interest in the book I am writing on grid operations. Of course, if you have a reaction to, or question about, this post, please leave a comment and I will be happy to respond.

[1] These website addresses are: pjm.com (PJM); Iso-ne.com (ISO New England); Nyiso.com (New York ISO); Misoenergy.org (MISO); SPP.org (SPP); Ercot.com (ERCOT); and Caiso.com (California ISO).