"The Piper at the Gates of Dawn" is the debut studio album by rock band Pink Floyd released in 1967. The album's title was inspired by a reference to the god Pan in Kenneth Grahame's 1908 children's novel "The Wind in the Willows", a favorite of bandmember Syd Barrett. The Willows are presented as a setting fraught with risks, and the wind through it underscores the unpredictable aspects of nature. The Piper references a mystical force in the liminal period between night and day. He is the embodiment of nature's benevolence, coming to the aid of creatures in need of rescue and comfort.

The power sector is having its Piper moment: Can it overcome risks associated with power supply sources, including wind?

Over the past five years, the power sector has built out significant renewable generation. As low variable cost renewables came online, we expected the economics of fossil generation to erode. That has already happened to an extent; coal-fired generation has significantly reduced and a substantial amount has been retired. Yet, against the run of play, gas-fired generation has not only been tenacious, but has actually benefited in many regions as power demand has surged. Our view has been that shrinking baseload supply would tighten power markets, but the surge in demand from new "large load" needs is coming faster than anyone expected.

In this rapidly changing dynamic, what is increasingly obvious is that renewables have been challenged to provide firm (noninterruptible) power. Much like its representation in the Willows, the wind has been an unpredictable resource. In the twilight hours before dawn and dusk, as solar generation is ramping up or down, regions with significant renewable installations have confronted supply shortages, requiring the support of fossil generation. Power prices have made corresponding intraday spikes. We believe this dynamic will persist for the foreseeable future in regions such as the California Independent System Operator (Cal-ISO) and Electric Reliability Council of Texas (ERCOT).

Power sector participants will meet a Piper at the gates of dawn, all right. Whether they meet the benevolent Piper of the Willows, or the Pied Piper collecting a steep toll, depends on whether they are long- or short-generation, respectively.

In this commentary, we update the state of the power sector and explain its resurgence. We anticipate tailwinds behind the sector's credit profile and have upgraded companies and assigned positive outlooks to ratings on others.

At the outset, we clarify that we are disinterested observers on public policy--our mandate is not to influence it. We comment on policy only to the extent it affects the credit quality of a sector or the companies within it.

The Tailwinds Started in 2022

The power sector shifted in 2022. Three factors combined for an unprecedented run-up in power prices:

• Higher solar installation costs: Higher polysilicon and steel prices that nearly doubled and tripled, respectively, an antidumping investigation by the U.S. Department of Commerce, and the implementation of the Uyghur Forced Labor Prevention Act all stymied solar renewable proliferation in 2022.
• Higher wind installation costs: Wind turbine order prices from western manufacturers also steadily rose over the past few years to $1,000 per kilowatt (kW) in 2022 from below $800/kW in 2019, driven by commodity prices and supply chain pressures. Because project lead times average two years, the recent increase in turbine costs is not reflected until projects commissioned in 2023-2024.
• An increase in natural gas prices: With Europe in the middle of the Ukraine-Russia conflict, there was upward pressure on natural gas prices as Europe was energy short and needed massive imports of liquified natural gas.

Power prices have tapered off from their 2022 highs. Natural gas prices--especially this year--have declined as a production surge has encountered a milder winter, resulting in high natural gas storage and lower natural gas prices.

On a more secular trend, solar module prices and overall capital costs have fallen under 2020 levels from a sharp drop in polysilicon prices during the second quarter to fourth quarter of 2023. Module inventories in the U.S. remain strong as supply chain issues subsided in 2023 and global production capacity expanded. Moreover, a two-year moratorium on tariffs for imported modules issued in June 2022 helped slow the inflationary pressure on module prices (although its expiration in June 2024 continues to present near-term uncertainty). Similarly, wind turbine order prices from Western manufacturers plateaued in 2023, and we expect them to decline from 2024, tracking raw material prices (notably steel, aluminum, and copper).

While we viewed the high-power price phase as temporary, we note renewable power purchase agreements (PPAs) are still relatively sticky at their higher levels (chart 1). This is because supply still faces structural constraints to bring on new investment, scale substitutes, and navigate concentrated supply-chain bottlenecks, resulting in a long-term deficit in critical commodities. Solar and wind PPAs both continued to exhibit higher contracting prices through last year to reflect the effects of higher costs (labor, interconnections, etc.), the expectation of continuing scrutiny of Chinese imports, and higher interest rates.

Chart 1 

Baseload Supply Has Been Shrinking

No incumbent utility or power generator is investing new money in coal generation units or coal mines. Rising labor costs, property tax bills, and still-lofty Appalachian coal prices point to further pressures. Coal units are increasingly undermaintained as the economics of coal-fired generation deteriorates.

In March 2023, the Environmental Protection Agency (EPA) finalized its latest "good neighbor" policies via its Cross State Air Pollution Rule (CSAPR), setting nitrous oxide levels that become more stringent in 2026 and move away from "banking allowances" by 2030. The EPA estimates this could drive 14 gigawatts (GW) of the current coal fleet to retire by 2030. However, with effluent limitations guidelines (ELGs) and coal combustion residual (CCR) regulations to follow, we believe the next large wave of coal-fired retirements by 2027 is still likely.

Bottom line, we expect coal retirements will continue, particularly in the Pennsylvania New Jersey Maryland Interconnection (PJM) market, where capacity clearing prices are too low to keep coal plants economical and where we forecast 25 GW of policy-related retirements. We believe the rules that contribute to the largest requirements include about 6 GW related to Illinois' Climate and Equitable Jobs Act in 2030, about 4.5 GW in 2026 for EPA's CSAPR rule, 3.5 GW from EPA's ELG in 2028, 2.5 GW from EPA's CCR guidelines in 2027-2028, and 2.5 GW from New Jersey's carbon dioxide rule.

As of November 2023, about 135 GW of U.S. grid-connected, coal-fired capacity that existed in 2010 has retired. Last year, our affiliate, S&P Commodity Insights, expected about 100 GW of additional retirements through 2030, but now it expects only about 60 GW to retire as generation supply tightness has pressured utilities in regions such as the Midwest ISO to extend the retirement dates for several coal plants. We believe more capacity would have delayed shuttering but cannot as it is mandated to close subject to EPA rules (or agreements with environmental organizations such as the Sierra Club). Absent a change or repeal of those rules, we assume such mandated units will shutter.

Pick Any Two: Cheap, Clean, and Consistent Power

Renewables have fundamentally changed the shape of the dispatch curve.

Let's assume the cumulative generation supply capacity along the horizontal X-axis of a graph and the marginal costs of these generation units along the vertical Y-axis. Then a plot of dispatch costs (or prices bid by the units in the day-ahead market) represents the generation supply stack. Along the horizontal axis, new, near-zero marginal cost renewable resources have pushed the curve to the right. At the same time, reduced marginal fuel costs of natural gas lowered the slope of the curve. Not only has the curve flattened, but it has also effectively flipped, and units that were formerly committed as base or mid-merit power supply are now more regularly the marginal resources needed to meet demand.

Normally, lower variable costs of the power supply dispatch stack would be a good thing, but three factors have prevented this:

• Investment and production and tax credits (PTCs) were available to solar and wind installations much earlier than the Inflation Reduction Act (IRA) made those available for stand-alone storage.
• Storage technologies have lagged solar and wind technologies.
• More broadly, technology has lagged aspirations ; a four-hour storage solution helps but is not a long-term solution.

The likely reason why storage could not avail tax credit incentives earlier is that battery technologies were expensive and still relatively nascent when solar was scaling up over the last 20-25 years. So when the first solar tax credit was enacted around 2005, solar technology was proven, but energy storage was far from proven. Typically, until a technology matures for utility scale deployment, there is little support to subsidize it.

Put together, the lower-cost renewable capacity came with the drawback of interruptible power supply that could not be firmed up by commensurable co-located storage solutions, which would help to time-shift load away from peak hours.

A Good Idea Taken Too Far Becomes Indistinguishable From A Bad One

While as-produced renewable contracts deliver cheap and clean power, they do not deliver firm power. For instance, all solar units produce at the same time. This means that as solar installed capacity goes up, its effective load carrying capability (ELCC) goes down. This is because ELCC values depend on when electricity shortages are most likely, and the type and quantity of resources already on the grid have a significant impact on the likely timing of electricity shortfalls. As the grid adds more solar plants, it reaches a point where they prevent daytime reliability issues so effectively that the remaining reliability challenges move into the evening hours when solar can't produce. At this point, adding more solar does very little to prevent electricity shortages, and the unit's ELCC often falls to 0%.

At the heart of the issue is that the generation that renewables are displacing is dispatchable, meaning it can be reliably called on to meet power demand when needed by the grid. This situation raises an economic challenge: The less opportunity dispatchable resources have to sell energy into the grid during a day, the less financially competitive they become, which, if unmitigated--for example, via an increase in resource adequacy (RA) prices in California--would ultimately lead to shutdowns and retirements.

We believe the market neither anticipated the rate of uptake of renewables, nor understood the impact of such a meaningful influx of solar penetration on the grid. It is only now that there is broader recognition of the fact that storage anywhere on the grid is beneficial to the absorption of renewable MWhs, which may otherwise be curtailed and not reach the system, or cause energy shortages as they ramp off at twilight.

As the grid transforms, dispatchability has become the critical attribute, and concerns about system reliability have mounted. So, when it comes to delivering firm power, without attendant storage to optimize production, one could say that renewables have temporarily become destructively disruptive--they have displaced dispatchable baseload generation without the ability at the present time to supplant them.

We believe dispatchability concerns could be a major reason as to why the EPA dropped requirements covering existing natural gas-fired power plants in its final Section 111 rule regulating power sector greenhouse gas emissions in February 2024. Only new gas-fired installations are subsumed in the rules the EPA will release in April 2024.

California And ERCOT Show This Deficiency The Most

California's now notorious "duck curve" emerged as early as 2012 and has since worsened (chart 2). This has placed emphasis on dispatchable resources. Similarly, ERCOT's record electricity demand last year, with 10 peaks breaking records in the summer and reaching an all-time peak of 85 GW. A California-like "armadillo curve" is now hitting ERCOT.

Chart 2 

We illustrate this problem with examples from ERCOT's real-time day pricing, one each from summer 2023 and winter 2023-2024.

In chart 3 below, we've shown intraday power pricing on August 10, 2023. We highlight the significant change in load to 85 GW at the peak from about 57 GW earlier in the day. Of note is the decline in wind generation (as a percentage of peak load) through the day, resulting in a sharp increase in power prices to $3,500 per megawatt hour (MWh) at 4 p.m. when solar generation began to ramp down that day. In the summer, price volatility is caused more from the sustained increase in load through the afternoon.

Chart 3 

Chart 4 illustrates an example from the winter. In this instance, the peaks are caused by the relatively lower production from solar generation, given shorter daytime hours, and also the substantial variability in wind, resulting in two peaks, one each at dawn and at dusk. Increasingly, prices are higher in the morning hours too.

Chart 4 

Based on Vistra Corp.’s earnings calls, we've made the following observations about ERCOT's summer 2023 prices:

• For August, ERCOT North's all-hours, real-time power pricing averaged $196/MWh.
• 43 hours in August cleared over $1,000/MWh.
• From 6 p.m. to 8 p.m., power prices in August cleared, on average, $843/MWh.

We believe "unfirmed" growth of renewables without storage no longer accentuates their gains; it instead amplifies their losses. In other words, a region can pay for its reliability risks periodically in spiking power prices from potential energy shortages, or it can pay the costs proactively and avoid reliability risks by firming renewables and offering other reliability incentives.

It is somewhat unclear to us how ERCOT's energy-only market will address this (a large deployment of storage could do it but will take time). The California market has started reacting to this, as its RA prices reflect the shortage. RA prices in Cal-ISO rose substantially over the past two years and are now at unprecedented highs, with some bilateral contracts reportedly inked at above $20/kW per month. Aggressive California Public Utilities Commission clean capacity procurement orders and potential noncompliance penalties support strong RA prices in the medium term.

The California market has recognized this need by pivoting largely to storage deployment, and battery storage has been the source of new capacity since 2021 (chart 5).

Chart 5

Based on these developments, we expect RA prices in California of at least $10/kW per month through 2030, compared with RA prices of about $2.5/kW per month as recently as 2018. To address the firming problem, we expect standalone batteries, or batteries co-located with solar, to comprise most future installations (chart 6).

Chart 6

A market is never irrational; it only seems irrational because it comprises different players with varying priorities. When solar and wind tax incentives were announced, developers with shorter-term objectives dominated the market, which then appeared irrational to the long-term players. That is our long way of saying that the renewables market has overinvested in renewables relative to storage.

Developers that have a shorter-term horizon (build and sell to monetize tax credits) may not factor this firming problem into their assessment. So a good way to gauge how the market is pivoting is to observe how incumbents with longer-term interests behave.

As storage technologies improve, and as the economics improve after incorporating with tax incentives, we anticipate a tsunami of storage deployments coming. Almost all major players--Vistra (at Moss Landing), Calpine Corp. (Nova), Clearway Energy Inc. (Daggett and Victory Pass, etc.), and AES Corp. (Alamitos, Luna, etc.) have pivoted to storage projects as a stand-alone or colocated solution.

Earlier, there was a view that storage needs to be physically tied to renewables because that was the only way it considered a green investment. Now, there is a growing appreciation that standalone storage can also reduce marginal grid level emissions by charging during periods of low grid emissions and discharging during periods of high grid emission. With costs declining in recent years, standalone storage is now an increasingly viable option for grid stability and ancillary services relative to traditional quick-start fossil units.

Risk Is What You Don't See--Now Comes The Demand Surge

In January 2024, the PJM Interconnection, the biggest ISO in the U.S., increased its demand forecast to a compound annual growth rate (CAGR) of 1.7% through 2030, up from 0.8% last January. We note that compounding is deceptive; it hides growth in a manner humans don't intuitively comprehend. This differential over a year is a significant revision, especially when compared with demand that has literally stagnated. Over the past decade, increase in demand from customer growth and electrification needs was offset by energy efficiency (building codes, efficient appliances) and behind the meter solar installations.

An unanticipated, and sudden, demand surge

S&P Commodity Insights estimates that by 2030, the contiguous U.S. 48 states' net on-grid demand will be about 200 terawatt hours (TWh) higher than its previous forecast. This uptick in demand growth expectations stems from three specific reasons:

• Large loads;
• Heating electrification; and
• Green hydrogen production.

Large loads refer to demand driven by segments like:

• Investment in new data centers;
• Manufacturing and industrial facilities (e.g., onshoring of semiconductor fabrication plants, electrification of oil and gas operations); and
• Cryptocurrency mining facilities.

By 2030, S&P Commodity Insights estimates incremental demand growth attributed to large loads will aggregate 150 TWh, effectively doubling the current estimated demand of 155 TWh. For perspective, 150 TWh is equivalent to the New York ISO's 2023 annual demand.

But here's the problem. All of us are just guessing (albeit with some fancier math). What we do know is that data centers will drive strong regional growth in peak demand because they tend to operate around-the-clock. What we don't know is the amount of actual growth. For instance, McKinsey and Co. believes this demand could increase at a 15% CAGR over 2023-2030, with about 400 TWh tied to data centers (or about 9% of the current 4,250 TWh of U.S. power demand). In terms of capacity, these demand figures translate to about 60 GW, with a potential supply shortage of 20 GW-30 GW.

Regional growth expectations

Growth from large loads is also regionally concentrated. PJM's annual load growth report shows strongest growth is in PJM South (5% annually), ERCOT (2.3%), and the Southwest Power Pool (2.1%). S&P Commodities Insights expects PJM South and ERCOT to account for approximately 75% of incremental growth in large loads by 2030.

Within PJM South, S&P Commodities Insights expects data centers to double their peak load by 2030 (about 12 GW of incremental peak load). For context, total PJM load in 2030 is projected at 168 GW, up from 150 GW-152 GW. Thus far, delayed coal retirements and new transmissions have been the supply-side response as dispatchable new build construction faded after the PJM capacity auctions returned weak prices.

Within ERCOT, large loads rise to account for 20% of net on-grid demand by 2030, driven by the electrification of the Permian Basin as well as cryptocurrency mining and data centers. ERCOT's large load interconnection status reports that the queue for these types of load sits at about 40 GW by 2027. ERCOT also estimates its 2030 load peak at 93 GW (from 85 GW in 2023). Supply response here has been smaller natural gas peaking assets, but the low-interest loan bill that is yet to be finalized could change that.

The Epiphany Arrives

On March 4, 2024, Talen Energy Corp. announced a data center transaction with Amazon Web Services (AWS). This is the proverbial epiphany moment for the sector. The market recognized it as a gamechanger for the IPP sector because it highlights the value of clean, baseload generation in a growing demand environment.

The transaction comprised AWS's purchase of the data center shell and related infrastructure that Talen constructed. We view this component of the transaction as specific to Talen, unlikely to be replicated by other IPPs. The read-through for the sector is in the other components of the transaction, specifically, a 10-year power supply contract with AWS's data center and the incremental revenues paid by AWS for the emission free energy certificates (EFEC). We note that MW involved are not insignificant--AWS has minimum offtake commitments (120 MW) that could rise to 960 MW in stages. The purchase of the EFEC resoundingly underscores that hyperscalers value firm and clean generation. The fact that the Susquehanna nuclear station is a dual-unit site and provides power supply redundancy that a data center would need makes the transaction extremely attractive to the hyperscaler.

This read-through is the strongest for Constellation Energy Generation LLC, which has 11 dual-unit sites (albeit some are jointly owned and the partner would have to agree). Vistra Energy and PSEG Power LLC are also candidates.

What kind of clean energy works better with a large load is a matter of debate. Renewable developers will–-rightly--tell you that one cannot take a nuclear unit to where a data center is needed, and there are only so many dual-unit nuclear reactors available. However, nuclear units can take a data center truly behind-the-meter (off the grid). Market consultants we spoke to asserted that technology that can take a renewables powered datacenter off the grid is currently not available, at least through mid-2030.. Storage-based firm power will likely still need to utilize the grid for 20%-30% of its needs. The datacenter can contract for clean power from the grid to be deemed 100% green (called a green wrap) but will still rely on the grid.

Storage paired with solar also requires a lot of land. Yet, the potential for renewables paired with storage to serve demand is vast.

Credit Implications For The Unregulated Power Sector

The sector's credit profile has improved

We forecast credit stability through 2026 for a sector that typically sees credit volatility. On the financial side, we anticipate relatively strong free cash flow, which is now underpinned by a floor price for nuclear generation. We also project leaner balance sheets as deleveraging remains an objective for most IPPs. On the business side, companies have progressively cleaner generation fleets and a reasonable mix of regional, fuel, and retail/wholesale diversity.

Owning baseload generation assets is now a credit positive

Unregulated generators that are long-generation are well positioned to respond to the demand. In the sector, we believe Constellation Energy and Vistra are best positioned to benefit. We expect Talen Energy and PSEG Power to benefit as well, but to a lesser extent because of their smaller generation footprints. We also expect developers such as NextEra Energy Inc., Brookfield Renewable Partners L.P., Clearway Energy Inc., and Pattern Energy Group L.P. to allocate significant capital to firming power. As a separate point, we are highlighting Pattern Energy's Sunzia project because we're convinced that there can be no energy transition without transmission.

For example, we believe for every $1/MWh price move, Constellation's EBITDA would rise by $200 million, all else equal. This is about 5% of its expected EBITDA. Our calculations make a number of simplifying assumptions, like ignoring any dynamic hedging of its economic generation. We also ignore the fact that about 55 TWh of its generation is serving load at $32 to $34/MWh under Illinois' carbon mitigation credits plan (a price below the $40/MWh-$45/MWh PTC floor before inflation adjustments).

Conversely, the New York units are already earning a higher price compared with the PTC through the state's zero-emission credits program. Therefore, prices in New York would have to rise above $60/MWh before Constellation benefits. However, these calculations illustrate that its nearly 180 TWh of clean, baseload nuclear generation is now seen differently as it is valued more for its 24 X7 clean generation.

A similar calculation for Vistra, which assumes about 75 TWh of nuclear and renewable generation (including Energy Harbor's 30 TWh) and 105 TWh of gas-fired generation, also results in a $200 million-$210 million uplift for every $1/MWh increase in power prices. At current prices, we expect generation would also potentially increase, so the range could be a little wider. We note that our calculations assume about 25 TWh of coal-fired generation that the company has classified as "sunset" generation intended for retirement. Most of this retirement is mandated, but we believe some could eventually convert into natural gas units.

We expect NRG Energy Inc. to benefit as well, but less so as its asset-lite asset retail strategy balances load to generation through a combination of owned generation and physical and financial products (PPAs, options, etc.). We're curious to see how NRG's asset-lite strategy handles growth in its retail business should procuring incremental power to serve load become more expensive.

The IPPs lost the long-term investor over the debate about who should be the rightful owners of a fossil generation fleet that has an uncertain terminal value. This has made companies go private. One such company not receiving as much attention as it could since it went private is Calpine Corp. We believe Calpine's sponsors figured out that the transition to clean energy is a longer one than contemplated, but one that public markets would not appreciate. The company owns mostly natural gas fired assets (Calpine does own about 725 MW in the Geysers, a unique set of geothermal assets in California that come closest to a proxy for firm renewables).

We acknowledge that Calpine has exceeded our expectations in each of the six years since it went private. To address its carbon footprint, Calpine has now pivoted to carbon capture and sequestration (CCS) for use with its CCGT fleet and has won two Department of Energy grants that it must now negotiate terms for. We believe these developments are worth watching.

The ability to secure sites with grid interconnections and expertise in equipment procurement are key

Given how pervasive transmission issues are, we believe for the foreseeable future, it is all about securing generation sites that come attendant with grid interconnections infrastructure.

The time it takes for new generation capacity to go from planning to commercial operations has increased as grid interconnections are harder to find. We believe companies such as NextEra Energy are skilled here because they have an early mover advantage in securing sites and advanced placement in interconnections queues. NextEra also has experience and expertise in constructing and commissioning renewables. We view AES as another company with significant advantages and believe it's well-positioned for renewable development. NRG Energy and Vistra could also convert several older coal-fired sites that already have transmission access and brownfield infrastructure.

Circling Back To The Piper

Given the nature of recent capacity installations, intermittent capacity has replaced baseload capacity, resulting in not only seasonal shortages but increasing intraday price spikes. Regardless of the fact that storage is being installed aggressively, we believe this deficit is difficult to overcome and a "firming premium" will stubbornly persist through 2030, all else equal.

That said, all else is no longer equal. We believe there are credit tailwinds for the power sector as a result of greater conviction that the demand growth is real. Moreover, the nature of this demand growth needs reliable, baseload power.

Rate-based utilities will find it difficult, if not impossible, to offer their nuclear units to serve large loads as state regulators would not allow baseload generation to be diverted away from ratepayers. But there is nothing preventing an unregulated generator from doing so. For instance, we believe unregulated nuclear units could withdraw from PJM's capacity auction by participating in a a fixed resource requirement (FRR), assuming a load serving entity initiates the withdrawal of load from the capacity auction and contracts with that nuclear generator bilaterally to include in its FRR plan (i.e., to fulfill the capacity obligation for this load via FRR).

In fact, it is reasonable to think that this will happen. Because the data center will sit adjacent to the nuclear units, transmission and distribution grid charges are saved. As a result, the long-term PPA pricing that the generator will negotiate will be higher than the wholesale power price that it would otherwise attract, yet this PPA price will still be lower than the retail power price that the hyperscaler would normally pay.

Moreover, should more nuclear units commit to a similar Susquehanna-style transaction, the tighter the rest of the PJM market becomes. This is another wrinkle that the PJM needs to iron out before scheduling its already delayed capacity auction. In fact, on March 22, 2024, the PJM issued its latest guidance for sites that plan to operate co-located power loads. As a separate, yet related, point: We are a rating agency primed to be conservative, but expect prices established in PJM's capacity prices for 2025-2026 to be meaningfully higher.

Importantly, the demand theme plays into the asset life assumptions for natural gas-fired generation. No one we speak with now views natural gas generation as a bridge. Because of the anticipated demand growth, no one doubts that the natural gas fleet is required to address not only demand, but also to support the grid's reliability. Transitions take time and have unintended consequences. This is one of them.

For the foreseeable future, during peak summer and winter days, we will likely see price spikes in the twilight hours. We do not expect that to correct anytime soon.

The Piper will be at the gates of dawn, and dusk, for a while.