This report does not constitute a rating action.

On April 1, 1998, a British newspaper reported that, to integrate the U.K.'s traffic patterns with the rest of Europe, all traffic would drive on the right. However, the report continued, to ensure a smooth implementation of this major initiative, the change would be phased in--for the first six months the regulation would apply only to buses and trucks.

Of course, the report was an April Fool's Day prank. But it highlights the importance of taking each and every variable into account when restructuring long-established complex patterns.

This is also true for the U.S. electric grid. Wind, solar, and batteries have been the focus over the past four to five years, resulting in unintended consequences for credit quality in the sector. As a result of the ongoing largest fuel switch in the grid’s history, there are credit tailwinds behind conventional generation, and, perversely, credit pressures on competitive renewable power companies.

We published our industry credit outlook on Jan. 14, 2025. Then came the DeepSeek news that the cost of inferencing could potentially decline, significantly reducing compute infrastructure and power needs, but the sentiment in the unregulated power sector appears to remain strong despite this news. However, because we have had other major developments over the past two months, we are updating our views to reflect shifting industry trends. Here, we include topics discussed most at roundtables and investor meetings in conferences such as the Private Placement conference in Miami and CERAweek in Houston.

We have responded to some of the more nuanced questions we have been asked by investors about the energy transformation and how that may be affecting the credit quality of the sector. This commentary follows “Power Sector Update: The Piper At The Gates Of Dawn,” our commentary on the power markets published April 1, 2024, on RatingsDirect. 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.

There Is A Shift In The Power Narrative

The focus has shifted beyond renewables and now subsumes other forms of energy.

Because of an unanticipated surge in demand and ensuing concerns about the ability of the sector to provide firm power, the narrative has changed dramatically over the past two years. This shift was best described at the recent development day of NextEra Energy Inc. A company presentation highlighted some of the themes that have changed over the past two years, which we list in the table below.

Ironically, because the gas-fired generation supply chain is currently not at scale, renewables are now being considered the “bridge” to gas-fired generation.

The demand surge narrative has turned cautious but is tenacious

We expect both pricing for power capacity and energy to be underpinned by robust fundamentals through 2027.

Over the next five years, utility forecasts suggest that new data centers--including those dedicated to generative AI--will boost U.S. net on-grid electricity demand by 375 terawatt hours (TWh). In total, U.S. grid-based electricity demand could rise an average of 2.1% annually through 2030. However, in the aftermath of the DeepSeek announcement that it has lowered training costs for its language model by optimization inferencing through a process called distillation (and employing a “mixture of experts”) that potentially reduces compute needs, this level of growth is not guaranteed and depends on several factors such as data center demand being fully incremental and grid connected, and the adequate expansion of supply chains for hardware, a skilled labor force, and grid infrastructure.

While data centers represent the primary upside risk to load forecasts, incremental load growth is also driven by large industrial loads, such as new battery and chip manufacturing facilities and electrified oil and gas operations. Our affiliate, S&P Global Commodity Insights, projects the strongest growth in the Pennsylvania-New Jersey-Maryland (PJM) South interconnection region (8.4% per year), the Southeastern Electric Reliability Council (SERC)-Southeast region (3.2%), and the Electric Reliability Council of Texas (ERCOT) (2.7%) region (see chart 1).

Chart 1

Independent power producer (IPP) stocks have been seesawing on news of data center lease cancelations in Europe and the U.S. However, we saw this demand narrative as “aspirational” and more important to the equity story. We think that even if expected demand estimates are cut by 50%, generation supply-–and not lower data center demand--will be a key driver in the power sector. The ability of the grid to add incremental firm power generation is the predominant constraint function.

Through 2030, we expect new industrial demand to grow faster than timelines for commensurate power supply additions to the grid. While new data centers and manufacturing facilities typically take two to three years to design and construct, new power supply can take five years or more to come online. Regions with the greatest upside risk to demand growth would see reserve margins fall below target levels without a supply response.

The potential implementation of sectoral and reciprocal tariffs is affecting U.S. growth, which may now experience recessionary headwinds (based on the expected depth, duration and dispersion of slowdown). This slowdown is resulting in both downward adjustments to demand and increase in input costs. However, we think secular growth trends in the industry outweigh the deceleration in economic growth impacts (see growth components in chart 1). Generation supply shortages imply the increase in input costs will likely be passed through to end users in higher power contracts.

For instance, the impact of earlier tariff increases and supply chain disruption has already resulted in the pass through of costs in the power purchase prices of renewable contracts. We note the continental index of renewable purchase power agreement prices has doubled since the first quarter of 2021 (chart 2) related to uncertainties around tariffs, tax credits, and transmission buildout.

Chart 2

The market needs dispatchable generation support

We are seeing a resurgence in gas-fired generation in PJM and ERCOT.

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.

While there are regional differences, the U.S. has mostly built wind, solar, and batteries over the past four to five years. This unintended lean toward intermittency has pressured winter dynamics significantly. On-peak demand hours in winter typically occur when there is no solar output. While wind output tends to increase year over year, it has not always been the case. For example, in ERCOT, the winter peak in 2021 saw a lower wind output at the peak demand hour than in the previous year (see chart 3 below). As a result, the reliance on natural gas-fired generation for peak hour demand, and for total energy on the highest energy demand days, appears to be growing. The chart below presents how much of demand was serviced by different fuel types. The need for gas-fired generation has increased virtually each year over the past seven years, and this is happening even as the grid is not building much new gas-fired generation.

Chart 3

Renewable deployments are lower relative to expectations

Renewables cannot address expected demand alone. In fact, we expect slower renewable growth than currently in the interconnection queue, further aggravating supply upside.

Based on the interconnection queues nationwide, grid-facing solar and onshore wind comprise nearly 60% of forecasted additions to the power supply stack (see chart 4), which is consistent with installations over the past four years.

However, against the run of play, renewables now face challenges that may persist and lower their development potential. Most land-based renewables are sited in rural areas where surveyed measures of wind and solar favorability have been declining. This decline has coincided with a rise in local challenges to siting and permitting of renewables. In PJM, for instance, 30%-40% of all proposed renewable projects are in counties with significant local opposition.

These challenges often result in higher development costs by increasing the chance of cancellations and delays and contribute to higher costs if they push projects into less cost-effective areas. We note renewable development costs are up to about $120 per kilowatt (kW; or 15% higher) in states with siting and permitting challenges.

Development headwinds are more significant in the east than the west and Texas, which may reflect differences in population density. Compared with solar, which is a much more consistent resource across geographic regions and can be sited more closely to load, onshore wind is more sensitive to development restrictions that may arise from local opposition.

We see new onshore wind projects installed at lower levels than what the interconnection queue would suggest. Although most onshore wind installations tend to be on private land, the current moratorium on federal land usage also adds to that risk. The lower additions are offset to an extent by repowerings.

Installations also appear to have slowed down because hyperscalers want tariff risks offloaded on developers, while developers and suppliers often cannot agree on tariff risk sharing.

Chart 4

Renewables have not delivered firm power at scale

Renewables cannot address large load requirements on their own because of intermittency limitations. In that context, planned battery energy storage system (BESS) deployments are still substantially lower.

Chart 4 also underscores an important aspect of generation supply deployment. While as-produced renewable generation delivers cheap and clean power, it does not deliver firm power. Yet, batteries are not being deployed at levels commensurate with renewables deployment.

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 cannot produce. At this point, adding more solar does very little to prevent electricity shortages, and the unit's ELCC often falls to 0%.

We believe the market neither anticipated the rate of uptake of renewables nor understood the impact of such a significant 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 megawatt hours, which may otherwise be curtailed and not reach the system, or cause energy shortages as they ramp off at twilight.

Nationwide developers are still not installing BESS commensurate with renewable deployment. That makes sense given the low solar capacity utilization that still averages 28%-30%. In the initial phase of investments, the one exception is California, where we expect the ratio of battery capacity to solar gigawatt (GW) capacity to be about 1.0 by 2026 because batteries can now be used to optimize the large solar generation installed base (see sidebar 1).

We think the likely reason storage is not growing commensurately with renewables is that battery technologies are still expensive and relatively nascent, while solar has scaled up over the past 20-25 years. So, when the first solar tax credit was enacted around 2005, solar technology was proven, but energy storage was still nebulous. 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 an interruptible power supply that has not been firmed up by commensurable co-located storage solutions, which would help to time-shift load away from peak hours.

We note the recent fire at Moss Landing, among the largest battery installation in the world, also soured lender confidence about the pace at which battery technologies have improved. Because of perceived thermal runaway risks, it will also attract additional regulatory scrutiny on all new BESS installations.

Chart 5

Batteries still have a duration issue

We are agnostic toward whether a region pays for its reliability risks periodically in spiking power prices from potential energy shortages or pays upfront costs and avoids reliability risks by firming renewables and offering other reliability incentives (see also sidebar 2). In our view, these approaches are the difference between the CAISO and ERCOT markets. From a credit perspective, we believe the ERCOT market will remain in transition for longer, offering generators volatile but higher cash flows. Historically, all else equal, we rate project finance transactions with merchant exposure in ERCOT a notch (or two) lower than transactions in CAISO or the PJM Interconnection.

We note longer duration batteries are significantly more expensive because as durations increase, battery cell costs account for a larger proportion, increasing sensitivity to raw material prices and supply dynamics. For context, the capital costs of a one-hour battery are about $500/kW, rising to over $1,150/kW for a four-hour battery. S&P Global Commodity Insights estimates the cost of a 10-hour battery at over $2,500/kW (see chart 6).

Chart 6

So, batteries need some form of a visible revenue stream. We believe California is addressing the intermittency issue well. CAISO's solar fleet benefits from large investments in longer-duration battery fleets, facilitated by its market rules. The California market construct has incentivized long-duration batteries through its resource adequacy (RA) payments. RA prices in CAISO have risen substantially over the past two years and are now at unprecedented highs, with bilateral contracts reportedly inked at above $20/kW-month. As a result, over 98% of proposed solar capacity in the CAISO footprint is now co-located with storage. According to data provided by S&P Global Commodity Insights, the average duration of batteries in California, including the ones in the interconnection queue, is about 4.5 hours.

Critically, because of the absence of an RA type recovery mechanism, batteries in ERCOT average substantially shorter durations compared with California. Developers in ERCOT have invested mostly in one- to two-hour batteries owing to high prices for ancillary services, combined with a growing arbitrage opportunity. We have noted before that we think a short duration battery is merely a butterknife in the intermittency gunfight.

The emphasis on carbon-free generation has eroded the supply chain for gas-fired alternatives

Given the poor results for recent new builds and weak economics, there had been no active interconnection requests for natural gas plants since 2021 (through the third quarter of 2024), a first since 1997. Besides, the expansion of transmission infrastructure assets is a long-term planning process that requires permitting and siting and is typically done at a measured pace. It requires regulatory approvals that often take many filings and considerable time.

In our opinion, the bigger issues are supply chains. The supply chain for gas-fired generation has been disrupted, especially after the peak of the COVID-19 pandemic. Chart 7 below shows the backlog of orders at the three primary gas-turbine suppliers. Here, backlog refers to the total value of orders that have been placed by customers but have not yet been fulfilled or delivered. It represents the pipeline of confirmed sales that companies expect to execute in the future.

Tightness in the supply chain is evident in six dropouts from the Texas Energy Fund’s (TEF) lists of accepted projects. While the latest one to drop (Constellation Energy Corp.’s Wolf Hollow III) cited uncertain timing of the issuance of an air permit for the project as the reason, Engie cited equipment procurement constraints and other factors that would have delayed its project’s schedule (the 930 MW Engie Perseus peaker) such that it would be unable to make the statutorily mandated initial loan disbursements by year-end 2025. Developer WattBridge also filed to drop out of the TEF, which included its Longleaf (600 MW) and Elmax (510 MW) plants.

Once supply chains are disrupted, they take time to regroup. We expect supply chains for gas-fired generation to start correcting only in 2027. For instance, GE Vernova Inc. announced it is building furnaces (for forging and castings) at its Greenville, S.C. facility through the first half of 2026. Heavy machinery will be in place by end of 2026, and 20 GW/year incremental delivery of gas turbines will ensue in 2027.

Chart 7

What Does This Mean For The Industry’s Credit Quality?

Owning baseload generation assets is now a credit positive

Unregulated, ‘generation long’ companies are well positioned to respond to the demand. In the sector, we believe Constellation Energy and Vistra Corp. are best positioned to benefit. We expect Talen Energy Corp. and PSEG Power LLC to benefit as well, but to a lesser extent because of their smaller generation footprints. We also expect developers such as NextEra Energy, Brookfield Renewable Partners L.P., Clearway Energy Inc., and Pattern Energy Group L.P. to allocate significant capital to firming power. Pattern Energy's Sunzia project is well positioned in our view, because there can be no energy transition without transmission (Sunzia’s transmission offers a firmer renewable energy product by bringing wind generation from Arizona and New Mexico that would blend with California’s daytime solar generation).

Securing sites with grid interconnections and expertise in equipment procurement are key

Given how pervasive transmission issues are, we believe, at least for now, it is all about securing generation sites that come with grid interconnections infrastructure. The time it takes for new generation capacity to go from planning to commercial operations has increased because grid interconnections are harder to find.

We believe companies such as NextEra Energy are skilled in this respect because they have an early mover advantage in securing sites and advanced placement in interconnections queues.

NextEra Energy also has experience and expertise in constructing and commissioning renewables. We view AES Corp. as another company with significant advantages and believe it is well positioned for renewable development (but has pared growth somewhat recently). Companies like NRG Energy Inc. and Vistra could also convert several older coal-fired sites that already have transmission access and brownfield infrastructure. We are already hearing of potential large gas-fired projects at the former Homer City and Bruce Mansfield coal-fired sites.

Perversely, renewable companies face credit headwinds

Renewable assets in this country have historically been funded with a combination of tax equity and debt. A typical project comprises about 40% project debt, 40% tax equity, and a modicum of sponsor equity (the composition differs for solar, wind, and batteries, but equity isn’t substantial in the capital stack). That means much of the cash generated is encumbered to service tax equity and debt, leaving relatively lower flows to equity. So, the secret about renewable financing is that it typically cannot fund growth. Sponsors need constant access to external capital.

This became all the more relevant with the implementation of the Inflation Reduction Act (IRA) because of that turbo-charged growth, which was funded with debt and tax equity. Transferability provided another fillip as it became possible to monetize tax credits quicker and from a wider base of tax equity counterparties. Even as rates were increasing, access to competitively priced debt existed because of sustainability mandates. Despite the fact that those were the best of times, renewable companies did not raise much equity.

Then came the combination of increasingly higher interest rates, supply chain disruptions, allegations leading to AD/CVD tariffs on southeast Asian imports, and higher development costs due to siting and permitting considerations. Sponsors were also now unable--or unwilling--to issue equity as stock prices declined substantially from erstwhile levels.

We’ve seen growing credit pressures across the competitive renewable segment with a majority of negative actions on ratings, or outlooks, coming in this space.

Tightness in markets will likely keep capacity prices elevated for longer

Despite the latest news about DeepSeek, the belief now appears to be that the cost of inference will significantly come down, and that will make it easier for companies to infuse other applications with inference and generative AI.

We note that large hyperscalers have actually increased their investments decisions, with Google’s up to $75 billion for 2025 (the market consensus was $58 billion). We note Amazon’s headline news of fourth-quarter 2024 capital expenditure (capex) of $26 billion, which analysts expect will be annualized into a run-rate $100 billion, ahead of street expectations of $80 billion for 2025. Similarly, Meta Platforms Inc.’s (Meta) full-year capex plan in 2025 is $60 billion-$65 billion and Microsoft expects a capex spend of $80 billion in 2025. Still, there are some potential headwinds on the data center front—there was recent conjecture that Microsoft may pull back on some data center projects around the world. We’ve yet to see official changes in its capex plans, or revisions to any already announced data center projects in the U.S. Despite a strong 2025 spend, the jury is out on how much the hyperscalers will spend in 2026 and beyond.

Based on demand/supply dynamics, we expect markets to be unusually tight through 2028. For instance, in the PJM, if we were to estimate new generation based on its stage of construction and its financing status, we see only one unit totaling 925 MW coming online through 2028 (Trumbull). Further, we project only 4.5 GW online through 2029 (Sycamore and Glen Falls in 2028 and Chesterfield and Shay in 2029). We note that new combined cycle gas turbine (CCGT) builds are still bound by the April 2024 Environmental Protection Agency rules, which require investments in carbon capture by 2032. While we expect these rules to be challenged, they add to the new build uncertainty.

Last week, the PJM Interconnection published its parameters for the upcoming 2026-2027 capacity auction in July 2025. Overall, we saw a combination of stronger demand forecast (5.45 GW or 3.5% higher); a higher reserve margin requirement (19.1% versus 17.8% in the initial parameters); higher gross cost of new entry (CONE; gross CONE for regional transmission organizations set at $505/MW-day), and lower capacity emergency transfer limits for several locational deliverability areas as constructive for prices.

Finally, U.S. power demand growth remained solid (up 1.5% year over year on a weather-adjusted basis, a six-month average until February), despite the slowdown in U.S. GDP growth. The current tariff regime could eventually influence economic growth and impact power markets. We believe any slowdown will weigh first on the energy market, making it all the more possible that revenue requirements will be bid into the capacity market. As a result, we expect PJM’s capacity prices to remain elevated for longer (our latest assumptions are linked at the end). Our affiliate, S&P Global Commodity Insights, recently updated its expectations for capacity prices, which we present below (chart 8).

Chart 8

Despite the near-term tightness, the longer-term forecast for capacity prices is less certain as more generation comes on. Last month, the Federal Energy Regulatory Commission approved PJM’s fast-track initiative for shovel-ready new generation. PJM recently announced its new fast track for dispatchable generation (reliability resource initiative) received 94 applications totaling 26.6 GW. Half of these applications would be uprates at existing plants and half would be new builds, including batteries.

We expect asset valuations to be higher and asset lives to be longer

In order to address reliability, we now see a kitchen-sink approach. First, we think more storage alongside new renewable installation is certainly needed. But now we also see more gas-fired additions (and perversely, slower coal retirements). Build time on new CCGTs is four to five years and costs have also gone up, but peakers will be built more rapidly to address immediate issues. Coal-to-gas conversions (or entire rebuilds like at Homer City) can also make use of existing interconnection queues.

While gas-fired generation supply chains are moving forward, they are not available at scale until 2028. Cost have also increased as sponsors vie for skilled labor (e.g., craftsmen and welders) against a backdrop of competing construction of liquefied natural gas facilities, chip manufacturing facilities, onshoring of industrial complexes, and data centers, among others. We see engineering, procurement, and constructions costs higher by 25%-30% over prepandemic levels. Supply of equipment, notably gas turbines, transformers, switchgears, for example, place additional upward pressure.

In recent filings for new-build cost estimates for generic CCGT plants by utilities in their integrated resource plans, we note capital costs have increased to average $1,650/kW. Recent sales (minority ownerships as well as full asset sales) of CCGTs in the PJM (see chart 9) average $1,175/kW. Adjusting these prices (to remove depreciation to date assuming a 40-year useful life and straight-line depreciation and escalating the result to 2027) would imply an adjusted replacement price of about $1,600/kW or more.

Chart 9

Importantly, the demand theme plays into the asset life assumptions for natural gas-fired generation. The demand supply mismatch will likely result in gas-fired plants to remain in service longer and also command higher valuations. 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.

Related Research

• Capacity Market Assumptions For Power Project Financing, April 8, 2025
• Power Sector Update: Credit Drivers In The California And Texas Power Markets, June 18, 2024
• Power Sector Update: A Rising Tide And Other Credit Views, June 10, 2024
• Power Sector Update: The Piper At The Gates Of Dawn, April 1, 2024

Primary Contact:	Aneesh Prabhu, CFA, FRM, New York 1-212-438-1285; aneesh.prabhu@spglobal.com
Secondary Contact:	Vishal H Merani, CFA, New York 1-212-438-2679; vishal.merani@spglobal.com
Research Assistants:	Valeria Morillo, New York 3322680692; valeria.morillo@spglobal.com
	Joseph Moreno, New York ; joseph.moreno@spglobal.com