Babcock & Wilcox ($BW): Selling the AI Boom a High-Cost, High-Emission Power Plant from the 1980s

B&W: 160-year-old Boiler Maker Now Hyped as AI Speed-to-Power Solution

• Multiple bankruptcies, disastrous attempts to pivot to renewables and solar.
• B&W rescued in 2018/19 by B. Riley which ended up holding ~30% of equity.
• Core business of parts and service for boiler power plants in secular decline.
• Miraculous B&W stock price recovery fueled by announced plans to build Base Electron, a massive 1.2 GW gas steam plant, in a deal coordinated by B. Riley and controlled by Applied Digital insiders, ‘guaranteed’ by Applied Digital, but kept off Applied Digital’s balance sheet.
• Base Electron plan explicitly claims to be developing a grid-connected power plant, not a behind-the-meter solution.
• Grid-connected means: getting grid interconnection; burdensome permitting requirements; working with regulated utilities.

Gas Steam Plants are an Archaic, Inefficient, Abandoned Technology

• Gas steam plants are inherently inefficient: B&W can’t fight thermodynamics.
• For baseload applications, efficiency matters and operating cost differences are huge.
• Obsolescence proven by disuse and retirement of existing gas steam plants.
• Nobody has even considered building new gas steam plants for decades; not in modern playbook of utilities or regulators.

Regulatory Reality Throws Cold Water On Speed-to-Power Claims

• Building a 1.2GW gas power plant on the grid takes 4-5 years: Permitting (1-2 years), interconnection studies, construction (2-3 years); demonstrated by timelines of real plants in North Dakota.
• Siting, permitting, construction, and electrical equipment constraints are independent of ease of steam turbine availability.

B&W Valuation Can’t Be Justified

• B&W business of maintaining and upgrading legacy thermal plants is real, but coal is in long-term decline and coal-to-gas conversion is past its peak.
• In our view, Base Electron plant is unlikely to ever be built.
• Even if it does, it’s a one-hit-wonder; we don’t believe there will be ongoing demand for gas steam power plants.
• Proposed two-stage buildout to create a combined-cycle plant represents a seemingly novel, high-cost design for a power plant, as modern combined cycle plants use heat recovery steam generator to reuse heat from gas turbines, not boilers, rendering expensive boilers almost immediately redundant if plan was executed.

Base Electron Plan Among The Worst of Many Costly, Carbon-Intensive Ideas Now Jumping On the AI Power Bandwagon

• Locks in excessive emissions for decades.
• Economics can never be competitive; North Dakota grid customers could be left paying higher prices.
• North Dakota already expanding both wind and gas capacity, responsibly.
• AI power needs better addressed through demand flexibility & transmission upgrades.
• Even if data centers need gas for speed-to-power, bridge power can be executed more cheaply, quickly, efficiently & responsibly.

Introduction

For nearly 160 years, Babcock & Wilcox (B&W) has made boilers — a technology that powered the first age of electric utilities, kept the US Navy afloat through two world wars, and then faded as combined-cycle gas turbines and renewables rendered conventional coal and gas-fired thermal plants obsolete for new, utility scale electricity generation in the United States.^2,3 Now, with data centers driving a surge in electricity demand and order books for the largest combustion turbines stretching years into the future, B&W is proposing to sell an allegedly faster route to power: a new natural gas-fired boiler and steam turbine plant, sometimes called a gas-fired Rankine Cycle plant (we will refer to it as a “gas steam plant”), to be built by a start-up Independent Power Producer (IPP) called Base Electron.

The problem is that B&W’s boilers are an entirely obsolete technology and its speed-to-power argument does not survive scrutiny.

A gas steam plant is a 30-to-40-year, front-of-meter utility asset that would generate electricity at 35–42% thermal efficiency as opposed to 58–60% efficiency for modern combined-cycle gas turbine ( CCGT)⁴ plants, requires more than 24 hours to cold-start, and will be grid-connected in one of the most oversupplied power markets in the United States⁵. Perhaps recognizing the inadequacy of its boiler-only offering, the company has suddenly started to pitch an seemingly novel design for a new power plant – where sometime in the near future⁶ new gas turbines will be fitted to B&W’s steam boilers to create a hybrid gas-turbine-and-steam-turbine-boiler plant. Applied Digital’s CEO has claimed this two-stage buildout will create a “combined-cycle plant”, even though modern CCGT plants don’t use boilers at all. Those facts alone call into question the viability of Base Electron and B&W’s power plant plans.

Meanwhile, Base Electron’s proposed plant faces the same permitting, grid interconnection, and labor constraints that any other new power generation asset must overcome. We have been unable to find public evidence that B&W or its partners have secured a site for the plant, filed for permits that are required prior to starting any construction, applied for a grid interconnection, or confirmed offtake agreements with end customers.

The deal structure itself also compounds these concerns. For example, B&W (the boiler provider and design-build contractor), Base Electron (IPP and project developer), Applied Digital (AI infrastructure developer and project guarantor), and B. Riley (the investment bank), are bound together by cross-holdings and related-party relationships that undermine the characterization of this as an arm’s-length transaction between independent commercial counterparties. Moreover, the Applied Digital guarantee, on which the deal rests, provides little credit support due to multiple termination rights in Applied Digital’s favor.⁷

What remains is a notice-to-proceed issued by a newly formed, poorly capitalized developer, backstopped by a contingent guarantee, and promoted by a network of related parties. Sunlight does not expect this project to raise the $2.4 billion that Base Electron says it needs to build the plant. In the unlikely event that it does, the project will stand as one of the most carbon-intensive and capital-inefficient infrastructure investments of the data center buildout.

1. Current valuation implies a new Base Electron sized project every two years till mid-2040s

B&W’s shares have soared more than 400% since announcing its contract on November 4, 2025, to build a 1.2GW gas steam power plant for the newly formed IPP called Base Electron, ostensibly to power AI data centers. Our analysis concludes that the $2.4 billion project is unlikely to be financed, and that B&W’s intrinsic value – based on its actual operating business – is a fraction of its current share price. Current analyst estimates^8,9,10 assume execution of the contract over the next three years will lead to more than a doubling of revenue and EBITDA by 2028, and that B&W can sustain that run rate with additional power projects through to at least the mid-2040s (based on analysts applying 20-25x EBITDA multiples on the growth business). Since the new growth business is project-based, to hit consensus estimates of $200m 2028 EBITDA, and maintain that indefinitely, B&W would have to launch a new $2.4 billion project every two years. However, given the technical and economic inferiority of the approach and the timing risks, we believe

the market is grossly overestimating the chances that the Base Electron project will be financed and built, much less ~9-10 new Base Electrons being built in the next two decades.

Sunlight believes B&W’s core business of providing parts and maintenance for the aging fleet of coal and gas boilers could be generously valued at 6-10X EV/EBITDA. Assuming management’s 2026 EBITDA guidance of $80-$100M and accounting for the recent $218M equity offering we believe that the legacy business is worth ~$1.20-3.60 per share.

If completed, the $2.4 billion Base Electron project at 10% margins will generate $240 million EBITDA for B&W over three years. Alternatively, if Applied Digital elects to terminate its guarantee, B&W earns a consolation prize of up to $100 million. Both scenarios result in only $1-2 in additional value to B&W equity, in our estimation.

2. Existing gas steam plants can’t compete and are being systematically retired

The last utility scale steam power plant built in the US and running on gas, Big Bend Unit 4, was commissioned in 1985¹¹. Since then, the power sector has abandoned gas steam plants because of their inherent inefficiency and inflexibility. They are significantly more costly and emissions intensive than modern CCGT plants, and surviving gas steam plants run at an average capacity factor of only 16%, versus 66% for modern CCGTs¹². The US Energy Information Administration (EIA) reports that 76% of US natural gas capacity retirements in 2026 will be gas steam plants (even though these plants represent 15% of the gas power fleet)¹³, following a persistent trend over the last two decades^14,15,16. We don’t believe investors will be willing to finance a new, $2.4 billion, 1.2GW plant that will become immediately obsolete and likely only ever operate at a small fraction of its nameplate capacity.

Importantly, the proposed Base Electron project is not intended to be a temporary or “behind-the-meter” solution,¹⁷ unlike other costly and temporary solutions to data center power demand like gas reciprocating engines and aeroderivative turbines. Instead, it proposes to deliver grid-connected capacity that would ultimately compete as de facto merchant power against the mix of resources on the grid in the Dakotas, including incumbent coal plants, modern CCGT plants, and rapidly growing wind and battery storage capacity. As a baseload resource, the ~10,000+ BTU/kWh heat rate means that B&W’s plant would burn 40% more gas per kWh than CCGT plants. For intermediate and peaking needs, a gas steam plant has slower startup times, slower ramp rates, and greater cycling costs compared to simple-cycle turbines or reciprocating engines.

Overall, B&W is proposing to build, in 2026, a type of plant that the U.S. power sector gave up building forty years ago and is now systematically decommissioning to replace with superior technologies.

3. The “speed to power” claim does not survive scrutiny

B&W’s pitch is that boilers and steam turbines could be deployed faster than CCGT or simple-cycle turbines to offer data centers a near-term solution to power bottlenecks¹⁸. Yet Applied Digital and B&W’s own recent statements¹⁹ appear to contradict this idea. The announced plant is to be “in front of the meter and developed in partnership with regional utilities” and to “add reliable power to the Dakotas.” The economics of this plan are very different from providing contracted power directly to a data center as quickly as possible. Putting this plant on the grid means the project faces unavoidable schedule constraints regardless of generation technology – including site selection and permitting, interconnection and transmission upgrades, and labor availability.

We do not believe that Base Electron, the new entity developing the project, can currently meet the stringent prerequisites for expedited interconnection in the two relevant energy markets – MISO or SPP – including site control, off-take agreements, and substantial deposits. Nor does it appear that Base Electron has started the process of obtaining necessary permits. The regulatory history of similar projects in North Dakota suggests that even if applications for permits were submitted now, construction may not be able to start until the second half of 2027, making completion of a plant targeted “to begin operation in 2028” unlikely.

4. The economics don’t work in the Dakotas

Applied Digital, the ostensible backer of the project, suggested the plant is to be built “in the Dakotas,” and more recently, B&W has specified that the project is in North Dakota. This might be one of the worst possible markets in which to build a high-cost merchant power asset. Applied Digital’s own (successful) strategy to locate its data centers in North Dakota is based on the abundance of cheap electricity in the state²⁰ and a preference for obtaining reliable power from the grid.

North Dakota generates more electricity than it consumes, exporting nearly a third of its output, and has abundant and growing wind generation capacity. Large data center loads, combined with transmission upgrades, have benefited from stranded power in North Dakota. Established utilities and co-ops are already building and planning new dispatchable capacity to handle long term growth, including two modern CCGT plants: the 1.5GW Bison Generation Station²¹ under construction, and the proposed 1.4GW River Run Energy Center,²² a partnership between Basin Electric and NextEra intended to support new large loads.

5. No other power producer or boiler manufacturer in the world is building gas steam plants

If gas steam plants were a viable engineering solution to the data center power crunch, this would represent an obvious opportunity for manufacturers, electricity producers and data center developers everywhere. Major OEMs including Mitsubishi Power, Doosan, and numerous manufacturers of large industrial boilers could easily offer this technology, yet we have seen no signs of other OEMs or EPC contractors proposing or implementing the idea. Nor has any other utility or data center developer embraced gas steam plants, despite the urgent demand for power and eagerness to consider other alternatives to gas turbines such as fuel cells and reciprocating engines.

6. The emission profile creates regulatory and financing risk across the full asset life

A gas steam plant operates at a thermal efficiency implying roughly 40–50% more CO₂ per megawatt-hour than modern CCGT plants. Tightening emissions standards and carbon pricing in various forms remain on the medium-term policy agenda irrespective of the current administration’s posture. A plant commissioned in 2028 (or later) will reach only mid-life by the time most credible decarbonization scenarios require deep reductions in gas generation. Lenders providing long-tenor project finance would be underwriting regulatory and stranded-asset risk across precisely the window when gas generation faces its greatest policy headwinds.

7. The related-party structure raises concerns

While the engineering and economic motivations for this project seem confusing, the story makes more sense considering the incentives of the related parties behind the deal. B&W itself, facing existential risk from its weak balance sheet last year, wasted no time raising $67.5m through an at-the-market equity offering that closed on November 7, 2025, three days after its stock jumped following the project announcement, and on

May 15, 2026, the company sold $200m of its stock in an offering largely underwritten by B. Riley Securities. Base Electron, the project developer created a month after the deal was announced, appears to be controlled by Applied Digital and B. Riley insiders,²³ yet Applied Digital has yet to commit to buying Base Electron’s electricity and is now walking back any suggestion that the power is specifically intended for its own data centers.²⁴ In consideration for Applied Digital guaranteeing Base Electron’s obligations, B&W issued equity and warrants to Applied Digital that are now worth over $100m.^25,26 B. Riley, which has been deeply involved in restructuring B&W since 2018, once held nearly 30% of B&W equity and has started to sell its position.²⁷

1. B&W’s History and Pivot to AI Data Center Power

Pioneer in steam power

Founded in 1856, B&W is a pioneer of American industry that invented the water tube boiler and played a major role in power generation over the 20th century, building boilers for hundreds of power plants, developing the first supercritical coal-fired boilers in the US, and expanding into nuclear reactors for both naval propulsion and power generation. The company literally wrote the book on steam – Steam: Its Generation and Use – now in its 42nd Edition.²⁸

In the 1970s, B&W was acquired by McDermott, a major energy EPC firm, where it operated as a subsidiary. B&W was first forced into bankruptcy in 2000 by litigation stemming from the asbestos used over decades in its boiler systems. After a succession of reorganizations and spinouts and a failed early effort to develop small modular reactors,²⁹ today’s B&W contains the remains of the fossil-fueled boiler business, which separated in 2015 from BWX Technologies (BWXT), the naval nuclear power division that continues as an independent company.

Faced with the accelerating decline of coal-fired power in the US,³⁰ B&W pivoted to try to develop new sustainable businesses in renewable, environmental, and industrial markets. In 2015, B&W announced hundreds of millions of dollars of fixed-cost contracts awarded through its Danish subsidiary, B&W Vølund, to build and operate biomass and “waste-to-energy” incineration plants, but in 2018 the Vølund and other renewables segment reported negative EBITDA of $275.9 million due to delays and cost overruns across six European deals, leaving B&W once again in financial distress.³¹

B. Riley Rescue, Eos Energy and balance sheet issues

To shore up its balance sheet, in 2018 B&W raised capital through a rights offering involving B. Riley Financial and Vintage Capital Management,³² a fund controlled by Brian Kahn, who last year pled guilty to conspiracy to commit securities fraud in a case the SEC alleged led to $400 million in trading losses.³³ Kenneth Young, then President of B. Riley, was appointed B&W CEO in November 2018 under a consulting agreement, while continuing to serve at B. Riley, and the rights offering led to B. Riley-affiliated parties holding approximately 30% of B&W equity. Not all B&W shareholders were happy, and a derivative action was filed in Delaware against Young, B. Riley, and Vintage alleging “breach of fiduciary duty” and “unjust enrichment.” The lawsuit settled in 2023 for $4.75 million.³⁴

Following the 2019 rescue, B&W was reorganized around “Renewable, Environmental, and Thermal” segments, but continued to struggle with a weak balance sheet. In October 2020³⁵ B&W announced a “strategic partnership agreement” with a battery storage company called Eos Energy to market and install Eos’s zinc batteries globally. That agreement materialized just weeks before Eos Energy went public through a SPAC merger with B. Riley Principal Merger Corp II, a deal announced Sept 8, 2020. On November 13, 2020, B&W also bought 200,000 shares of Eos for $2 million in a PIPE just prior to the SPAC merger closing on November 16, then sold the shares in January 2021 at all-time highs for $4.5 million.³⁶ The partnership didn’t proceed, and six years later Eos shares trade ~75% lower.

Over 2024 and 2025, B&W shut down or sold numerous subsidiaries after an audit opinion in 2024 raised doubts about B&W’s ability to continue as a going concern,³⁷ and the company faced delisting in April 2025.³⁸ Although the auditors’ going concern opinion was lifted for 2025, B&W’s long-time COO quietly departed in August of that year, an exit buried in the second quarter 2025 10-Q. Then, later in 2025, B&W unveiled a surprising deal creating an entirely new narrative tied to AI, which dramatically reversed investor perceptions.

Data center pivot

On 4 November 2025, B&W announced its “AI Data Center Power Generation Solution” and “limited notice to proceed” for a project valued at over $1.5 billion to deliver one gigawatt of power for an “Applied Digital AI

Factory.” The B&W website suddenly featured photos of blinking servers and an “AI Data Centers” section. B&W stock rapidly doubled, and the company immediately seized the opportunity to raise $67.5 million on November 5, 2025, significantly improving its balance sheet, paying off bonds that were due Feb 2026, and eliminating much of the existential risk that had been looming over the company across 2025.

While B&W is unquestionably in a much better place than it was a year ago, we fear that investors may be misunderstanding B&W’s prospects for addressing data center power needs using its steam boilers. In March of this year, B&W announced the Base Electron agreement had progressed to a “full notice to proceed,” and the project cost had increased to $2.4 billion. B&W’s stock price has continued to rise, and most of B&W’s value now appears to be founded on enthusiasm for this AI data center power business.

2. Gas Steam Utility Generation: Archaic, Inefficient, Abandoned Technology

B&W’s proposed plant consists of four 300MW gas-fired boilers driving steam turbines. When industry experts with experience designing and managing the construction of power plants learned about the project, one of the kindest reactions we heard is “that’s… interesting. We haven’t been building plants like that since the 1980s.”

In fairness to B&W, none of the engineers we spoke with doubted that it is possible to build a plant of the kind proposed, nor questioned B&W’s bona fides in boiler engineering and maintenance. There are, of course, industrial applications that use steam to provide building or process heat, often with cogeneration, and B&W remains a key and respected player in those markets. But the world long ago stopped using steam boilers for new utility-scale gas-fired power plants. While a greenfield gas steam plant may be technically feasible, it is the wrong technology for modern generation. Existing plants survive only because they are sunk-cost legacy assets.

Inherent inefficiency of gas steam plants

The core problem with gas steam plants is that they are uncompetitive for both baseload and peaking/ intermediate power generation. For baseload generation, they are much less efficient than the CCGT plants that supplanted them. For peaking and intermediate generation needs, simple cycle gas turbines are faster and more flexible to start up than gas steam plants.

The average HHV heat rate of a gas steam plant is 10,337 BTU / kWh according to 2024 EIA data,³⁹ corresponding to ~33% efficiency. The 4 x 300 MW drum-type boilers proposed for the Base Electron project appear to be subcritical units⁴⁰ unlikely to have heat rates substantially better than this. In contrast, the average CCGT plant in the US operates at 7,548 BTU / kWh, and the newest CCGT plants using GE or Siemens H-class or Mitsubishi J-class turbines can achieve heat rates of under 6,500 Btu/kWh and efficiencies over 62%.⁴¹

These efficiency differences are determined by thermodynamics, and B&W can never close the gap. Although new steam turbines may be slightly more efficient at the margin than those used in legacy plants, according to an engineer with 40 years of experience in designing power plants, the differences are “one percent, half a percent, 1.5 percent. There are no miracles there.”⁴²

If the goal is to provide steady baseload power, the efficiency penalty of B&W’s boilers would lock in higher fuel costs and higher emissions for the life of the plant.

Even compared to simple cycle combustion turbines, a subcritical gas steam plant will likely be less efficient. Example configurations analyzed for the EIA Annual Energy Outlook 2025 report describe a 4x54 MW aeroderivative gas turbine configuration with a heat rate of 9447 BTU/kWh, and an H-class gas frame turbine operating at 9142 Btu/kWh⁴³ . Aeroderivatives offer maximal flexibility suitable for peaking and balancing renewable capacity, with fast start times and ramp rates. Larger frame turbines, while slower to start and somewhat less flexible than aeroderivatives, have lower capital costs per kW. Reciprocating internal

combustion engines (RICE) are also routinely considered for modular, flexible peaking capacity with good part-load efficiency.

Although the heat rate difference compared to simple cycle gas turbines is small, a gas steam plant also competes unfavorably on the operational characteristics required to provide flexible capacity. Gas steam plants are slow to start. An expert who worked on conversions of coal steam plants to run on gas explained: “The ramp rate is going to be very slow. So, whoever’s buying this thing is going to have to either let me run full load or let me ramp relatively slowly compared to a combustion turbine.”⁴⁴ Another engineer who led technology assessment at a large municipal utility agreed that steam plants are generally inflexible, and to ramp up capacity one would typically want to know “the day before.”⁴⁵ A 2012 NREL analysis of cycling costs reported a typical startup time of 4 to 48 hours for a gas steam plant, compared to 0 to 1 hour for an aeroderivative or 2 to 3 hours for a frame turbine.⁴⁶ Capex and maintenance costs attributable to cycling and load following were also estimated to be 2 to 3 times higher for a gas steam plant versus turbines, and those costs increase faster as cycling frequencies and ramp rates increase.

The poor operational performance of a gas steam plant is especially unappealing when it comes at a capital cost higher even than today’s inflated turbine prices. B&W’s reported price of $2000/kW exceeds the median all-in cost of $1557/kW for new projects deploying simple cycle turbines between 2027 and 2029 in a sample analyzed by Gridlab and Halcyon.⁴⁷ It’s unclear to us why any power producer would choose to pay a premium for a plant that is both intrinsically inefficient and ill-suited for flexible dispatch.

Source: “The New Reality of Power Generation: An Analysis of Increasing Gas Turbine Costs in the U.S.”, Gridlab, Appendix (p. 9).

Power plants operate for 25-50 years, and fuel costs dominate

Compared to “bridge power” solutions, such as aeroderivative gas turbines and gas reciprocating engines, the inefficiency of a large power plant like the one proposed by B&W is especially important because of its expected life. We recognize that a data center operator might be willing to pay a premium price for electricity in the short term to operate sooner, given the high potential revenue opportunity. But locking in unnecessarily higher electricity prices for the 25 to 50 years that power plants typically operate is a different proposition. As a project developer at a major IPP put it: “To think of a project as being just solely a data center served by

an on-site power plant and that’s the only place to get their power, that only works for the bridge time period; longer term you would economically dispatch the unit. In other words, shut it down when grid power is less expensive. What I’d say is the economics and the PPAs are complicated by that fact.”⁴⁸

Gas steam plants are being retired and replaced by CCGT

The inferiority of gas steam plants is not a theoretical criticism; the market’s wholesale rejection of this approach over the last 40 years provides evidence that gas steam plants are not competitive. Existing gas steam plants are being shut down and replaced with better technology, and those that survive are used as little as possible. Sunlight has been unable to identify a single electricity producer or government agency considering gas steam plants in systematic planning and technology evaluations.

According to EIA data, 66.9GW of natural gas fired steam turbine capacity across 685 generators has been retired since 2022, and in that time no new gas steam plants have been built dedicated to electricity generation. The only such plants built in recent decades are smaller units providing combined heat and power.

Surviving gas steam plants have low utilization

Because of their inefficiency and position in the merit order, most gas steam plants have capacity factors of 10-20% or less. That is, because the plants only run a small fraction of the time, they produce only 10-20% of the energy that they have the capacity to produce. In recent years, average utilization has increased modestly as some of the oldest and least-utilized plants are retired. Surviving resources – particularly large coal facilities that have been converted to gas – are having retirements delayed⁴⁹ to support demand growth and grid stability. Yet the only reason that it makes economic sense to continue operating these units is that they were built and paid for decades ago.

Gas steam plants are not considered a serious option by regulators or utilities for new capacity

The EIA publishes a regular analysis of “Capital Cost and Performance Characteristics for Utility Scale Electric Power Generating Technologies” to guide capacity planning decisions.⁵¹ The latest 2024 EIA report evaluated 19 case studies including diverse fossil fuel, nuclear, and renewable technology options, yet a gas-fired steam turbine plant was not even considered. Similarly, an exhaustive 2019 Department of Energy (DOE)/National Energy Technology Laboratory (NETL) analysis, “Cost and Performance Baseline for Fossil Energy Plants” did not view gas steam plants as even worthy of discussion.⁵²

Regulated utilities prepare Integrated Resource Plans (IRPs) to systematically evaluate the full range of technology options available to meet generation requirements and help justify rational long-term investment decisions.⁵³ We have been unable to find any IRP in which a gas steam plant even merits consideration. In North Dakota, the Basin Electric 2023 IRP⁵⁴ assessed simple cycle aeroderivative and frame turbines, reciprocating engines, CCGT plants, coal-to-gas conversions, wind, solar, battery storage, nuclear, and even new supercritical coal plants, carefully documenting capital and operating costs, best applications, and limitations of each. Greenfield gas steam plants were not included. We are doubtful that an established

electric utility or informed energy project investor will be willing to invest in building grid-connected generation capacity without fully studying the approach through standard planning processes.

North Dakota, with its reliance on coal, has not historically prioritized minimizing emissions. Yet even in a state with such strong public and policy support for fossil fuels including coal, recent deployments have focused on efficient gas plants and wind farms.

In a puzzling narrative shift, both B&W and Applied Digital have now suggested, six months after the initial announcement, that their new steam boiler plants could be later repowered with combustion turbines. In his “Investment Reports” interview on May 12, Applied Digital CEO Wes Cummins said: “ We’re deploying steam turbines on that gas, building the steam turbine portion of a combined-cycle plant first and getting delivery in 2028, with the traditional gas turbine to follow.”⁵⁵ Then, on May 15, KennethYoung appeared on Mad Money with Jim Cramer and claimed: “We can take our boilers and combine those with combustion turbines in the future. So, for example, we could put in a steam boiler today and produce, let’s just say, 100 megawatts out of that particular unit. We can combine that in the future when the combustion turbine becomes available at 100 megawatt, combine that with that steam boiler and turbine, and now produce 200 megawatt from that particular site without increasing the land mass.”⁵⁶

This “parallel repowering” strategy has been used in a handful of thermal plant retrofits, including the Peterhead Power Station in Scotland.⁵⁷ Peterhead was an oil-fired plant built in the 1970s, converted to burn gas in the early 1980s, and then repowered by adding gas turbines and a heat recovery steam generator in the late 1990s, while leaving the boilers connected to the steam system. The plant can operate both in a CCGT-only mode and a combined CCGT/boiler mode, which has slightly higher power output but is less efficient.⁵⁸ We have not been able to identify any plants like this in the US; the idea was briefly considered in 2011 for the retrofit of the decades-old A.B. Brown plant in Indiana but never pursued⁵⁹.

We see at least four potential problems with this concept. First, the unusual parallel repowering approach proposed does nothing to improve the efficiency of the gas boiler part of the system. To the extent that the boiler continues to operate, the heat rate of the plant will continue to be inferior to an ordinary CCGT.

Second, the strategy is wasteful in building a boiler that will soon be underused, since the steam from a CCGT is produced by a heat recovery steam generator using exhaust gas from the combustion turbine. Either the first iteration of the plant will not take full advantage of the capacity of the steam turbines, or the ultimate plant will not use the full capacity of the boilers. Either way, the plan also adds considerable design complexity and new risks of a novel plant design, rather than using standard, optimized packages available from turbine OEMs.

Third, replacing or integrating new gas turbines will massively increase capex and likely add significant downtime to later modify the otherwise brand-new plant. While repowering could be justified for old plants that are already paid for and need refurbishment, the Base Electron plan promises to build a brand-new inefficient steam gas plant, then almost immediately refurbish it into an even larger, inferior CCGT+boiler plant.

Finally, as we explain in the next section, using more readily available steam turbines instead of supply-constrained gas turbines will not on its own deliver ‘speed-to-power’, as a grid connected plant faces additional permitting and construction delays that are likely to neutralize this advantage.

3. Unrealistic Timelines Undermine Base Electron’s Speed-To-Power Promise

In Applied Digital’s Q3 2026 earnings call on April 8, 2026, CEO Wes Cummins reiterated that “Base Electron will work with Babcock & Wilcox to build a power plant that will supply, initially, roughly 1.2 gigawatts of natural gas-fired generation capacity to the grid in the Dakotas region. This power will be in front of the meter and developed in partnership with regional utilities.” He added: “towards the end of 2028 […] Base Electron will start to commission some of these new power generation assets.” B&W also previously stated

in its initial announcement that “the plant is targeted to begin operation in 2028.”⁶⁰ So far, no site has been disclosed; and as of Jan 2026, “the location is still to be determined,” according to an interview with B&W Chairman and CEO Kenneth Young.⁶¹

The Base Electron claims seem impossible to reconcile with the logistical and regulatory realities of building a new power plant. Even if a site that meets requirements for a plant of this scale is found (including gas supply, water for cooling, and community support), the announced timeline will collide with at least two unavoidable obstacles: 1) permits; and 2) grid interconnection studies.

In searches for Public Service Commission and Department of Environmental Quality filings, we have found no evidence to date of Base Electron initiating the regulatory submissions required to start construction in North Dakota. Critical stages in these processes have strict prerequisites that Base Electron must satisfy. Realistic timelines to build new gas-fired capacity can be calibrated against progress of real projects already under construction in the state, and mapping the timing of these recent projects onto the regulatory gauntlet faced by the imagined Base Electron plant implies that a construction project only starting to seek permits now would extend years beyond the 2028 timeframe promised, nullifying any speed to power advantage supposedly achieved by adoption of B&W’s inefficient and obsolete boilers.

Bison Generation Station: what a real plant under construction in North Dakota looks like

The Bison Generation Station⁶² is a 1,490 MW plant under construction in Williams County, North Dakota by Basin Electric, one of the largest electricity cooperatives in the state. The project has been planned since at least 2024, and publicly available disclosures of its development progress help calibrate how long it takes to permit and build a power plant in North Dakota.

Site and air permits

Starting construction of a power plant in North Dakota⁶³ requires, among other approvals, a Certificate of Site Compatibility from the Public Service Commission (NDPSC), and an Air Permit to Construct from the Department of Environmental Quality (NDDEQ) for compliance with EPA Title V.⁶⁴

The Public Utilities Division of the North Dakota Public Service Commission regulates all sitings of energy conversion and transmission facilities in the state, and communication with the commission occurs through

submissions to cases that are publicly posted on the PSC website and considered in public commission meetings and hearings.

Looking at the Bison project as an example, we found that Basin Electric filed its 345-page Application for Certificate of Site Compatibility on March 5, 2025. This document included an exhaustive analysis of need determination, a discussion of the resource alternatives considered based on Basin Electric’s 2023 Technology Assessment, a consideration of potential environmental impacts, and documentation of required correspondence to regulators, including the Bureau of Land Management, FAA, Army Corps of Engineers, Fish and Wildlife Service, ND State Water Commission, ND Transmission Authority, and local legislators, among others. The application enumerated at least 29 permits required for the project from various state, local, and federal regulators.

As of May 24, 2026, we have been unable to locate an Application for Certificate of Site Compatibility for the Base Electron project, nor any other cases or files submitted to the North Dakota PSC by Base Electron or B&W.⁶⁵ Applied Digital is directly associated with three existing PSC cases (see below), none of which concern a proposed generation facility, and several 2023 and 2024 additional submissions relate to power supply agreements and interconnection with Montana-Dakota Utilities to power existing data center projects.

3 cases:

5 cases:

Our conclusion, based on our searches for public records, is that neither Base Electron nor any of the related parties have filed for a Site Compatibility permit. If our assumption is true, we believe it unlikely that Base Electron can obtain this permit or start construction in 2026, because obtaining the Certificate of Site Compatibility is expected to take at least 5 months from submission based on the Bison example (which reflected the work of an established and experienced utility and planning that had started at least 9 months before submission).

Also needed prior to starting construction is an air permit, which is likely to take even longer. The North Dakota Department of Environmental Quality estimates a 12- to 15-month timeframe for permit issuance for a Major PSD Source like a new 1.2 GW gas-fired power plant:

ND Air Quality Permit timeframe (source: https://deq.nd.gov/AQ/permitting/PTCInProgress.aspx)

Information about sites and permits in North Dakota is published on the CERIS-ND website. For example, for the Bison project,⁶⁶ a New Air Permit to Construct application was submitted 2/10/2025, and the permit was issued 12/29/2025 following public hearings and submission of additional documentation. Basin Electric then began compliance submissions providing updates on the status of the project.

Once again, we have been unable to find any issued permits in the CERIS-ND database connected to Base Electron or Babcock & Wilcox. Applied Digital has applied for air permits to install backup diesel emergency generators at its FAR01 and FAR02⁶⁷ sites and at the Ellendale⁶⁸ facility, but none appears to relate to Base Electron or a new power plant. Based on the absence of an Air Permit to Construct, we do not believe that there is any Base Electron site currently under construction in North Dakota, and no construction can begin until a permit is issued, approximately 10 to 15 months after an application.

Grid interconnection

A key element of Base Electron’s plan is connecting the plant to the grid, and every power expert we spoke with viewed this as a timing risk. According to one IPP developer:

“I guess the real question is, can you go through the interconnection quick enough? Because if you’re going through a formal interconnection process, what’s the timing of that? Is it gonna take 4 years, 3 years to go through that process? Or is this a one-off situation where the utility will analyze the interconnection point, and in 6 months or 12 months they’ll be good to go? Because that’s going to be the critical path to getting a project ready for construction.”⁶⁹

B&W and Base Electron’s aspiration to obtain grid interconnection seems even more challenging than obtaining air and site permits. None of the parties involved have disclosed a site for this project, so we cannot know for certain where it is planned to connect to the grid. Coordination of the electric grid in North Dakota is split between two regional grid operators MISO and SPP. Applied Digital recently disclosed that the project would “expand power and capacity supplied to the grid and utility customers in the MISO region,”⁷⁰ suggesting the site will be within the Eastern part of the state where the grid is regulated by MISO.

Recognizing the urgency to add new generation resources, both MISO and SPP have recently implemented Expedited Resource Addition Study (ERAS) processes that can speed through interconnection studies in as little as 90 days. Yet, to ensure prioritization of projects that will really get built and serve documented demand, MISO and SPP impose stringent requirements (and hefty deposits) to apply through these programs.

The MISO ERAS process has “strict eligibility requirements due at the time of ERAS application submittal”, including “100% site control”, off-take agreements, and a commercial operation date no later than three

years from submission.⁷¹ MISO ERAS requests are dominated by established utilities with projects meeting those requirements.⁷²

Similarly, the SPP ERAS process is limited to “qualified load responsible entities … that meet clearly defined threshold related to near-term resource adequacy needs.”⁷³ Basin Electric, a major electric cooperative in North Dakota, entered the SPP queue in 2024 for the 1.5GW Bison plant expected to come online starting in 2029, and also applied through the 2025 ERAS process (SPP-ERAS-2025-011 and 012) for related transmission development.⁷⁴

As a speculative IPP with no site selected or under control, no committed customers, minimal capital, and no existing generation resources, Base Electron does not appear to meet threshold criteria to apply for an expedited grid interconnection study. Working through the standard process could take 3-5 years or more. Searches of the MISO Generator Interconnection Queue⁷⁵ and MISO ERAS Queue⁷⁶ for projects in North Dakota found no generator interconnection requests corresponding to Base Electron or any thermal power plant of the type proposed. We see interconnection as a major blocker to Base Electron building a grid-connected power plant by 2028.

Labor bottlenecks

If and when Base Electron obtains permits to start construction, it will need construction workers. As Applied Digital acknowledges in its own AI Factory Whitepaper⁷⁷, “North Dakota has the most constrained labor market” compared to other states like Texas or Virginia that are popular for data centers. Moreover, if the Base Electron plant is to be built in North Dakota, where most of the Applied Digital campuses are, it will be directly competing for highly skilled labor with at least one major project already underway, Basin Electric’s Bison plant.

Recognizing the labor challenge, the Bison project planned an extended schedule deliberately intended to reduce peak workforce requirements. Even so, Basin Electric expects to need nearly 1000 construction workers over 2027 and 2028. We believe that access to labor presents yet another substantial risk to the timing of the Base Electron project.

Transformers and transmission equipment

Electrical equipment supply chains will also impose scheduling constraints on a new generation project independent of turbine supply, with lead times for power transformers now exceeding 2.5 years⁷⁹ according to a 2025 Wood Mackenzie survey, and up to four years for high-capacity units.⁸⁰ The specifications for generator step-up and transmission equipment will also likely rely on results of interconnection studies that

do not appear to have started. Obtaining this equipment could impose additional scheduling constraints on the project that could further threaten the speed-to-power promise.

Water, gas, and local politics

Finding a suitable site for a new power plant of the scale proposed by B&W and Base Electron is not simple, and it is hard to take the plans seriously until a site is announced.

Gas fired boilers typically use significant amounts of water for cooling. Once-through water cooling is most common and most efficient in large plants of this type, although wet cooling towers are possible. While it may also be possible to use dry cooling,⁸¹ this would be an unusual design choice and come with a further cost to efficiency.

Base Electron will need a reliable gas supply. Redundant supplies are preferred, though not essential, but the volume of gas needed for a plant like this is not trivial. Proximity to gas supply is thus an important issue and additional source of permitting and construction delays. For example, the Line Section 32 Expansion Project,⁸² a 17-mile pipeline to serve the Bison Generation Station had its Federal Energy Regulatory Commission (FERC) application submitted in March 2026 and is only planned to be completed in late 2028.

Base Electron power by 2028 (or 2029?) seems unlikely

Taken together, the lack of an announced site, the apparent absence of permits and interconnection studies, and industry-wide constraints on labor and key equipment, lead us to doubt that Base Electron can deliver on its plan to “begin operation in 2028.” All the experts and engineers we spoke with had similar concerns about a 2028 timeline for any project of this scale starting now. “They’d be very well challenged to do that,” said one former IPP developer.⁸³

In fact, even B&W’s own disclosures now appear to contradict earlier promises of power by 2028, describing a 36-to-40-month timeframe for the 4 x 300 MW boiler configuration that would seem to preclude a project not yet permitted for construction from 2028 completion. B&W appears to have recently recognized the scheduling problem: Its most recent 10-Q now suggests that “the plant is targeted to begin operation by 2029.”⁸⁴

Yet an interview with Applied Digital’s CEO Wes Cummins published one day later⁸⁵, on May 12, claimed “generation capacity coming online by 2028” and for the first time, described an entirely new concept for the plant, stating “we’re deploying steam turbines on that gas, building the steam turbine portion of a combined cycle plant first and getting delivery in 2028, with the traditional gas turbine to follow.” If the B&W boiler approach is as efficient and effective as claimed, we wonder why Base Electron is now apparently considering such a strange plan. As stated above, we believe the strategy would impose considerable new engineering complexity and cost. In the meantime, we await filing of applications for a Permit of Site Compatibility and Air Permit to Construct with details on actual plans.

4. Inefficient New Build Gas Steam Plants Can’t Compete on the North Dakota Grid

Electricity prices and fuel cost matters

In the frenzy to build data centers, it has been suggested that “speed to power” is all that matters rather than the cost of energy since electricity might represent only about “20% of the total cost base”⁹⁵ of a data center business and AI data centers can generate significant revenue. We don’t believe that this proposition justifies the Base Electron project, for at least two reasons: In reality, Applied Digital itself claims to be highly sensitive to electricity prices, and a generation resource connected to the grid will compete as merchant power against the grid, in a state where electricity prices rank among the lowest in the country⁹⁶.

The low cost of electricity in North Dakota is, after all, one of the major motivations behind Applied Digital’s strategy of building data centers in that state. As Applied Digital explains in its own whitepaper,⁹⁷ “electricity costs, [are] the largest operating expense in a data center” and “stranded power along with the lowest commercial and residential electricity rates in the United States makes North Dakota a great candidate for AI Factories.” Relying on a power plant that might use up to 50% more fuel per MWh is certain to increase the ultimate price of electricity to a data center and expose customers to additional fuel price risk. We are skeptical that any customer will commit to a long-term take-or-pay PPA at prices far above those available from the grid, and so far, no such customer has been disclosed.

Steam boilers are ill-suited to peaking or intermediate generation

While the inefficiency of B&W boilers for baseload energy is irrefutable, some might suggest that even an inefficient plant could be useful as a capacity resource. It is true that old steam plants have been kept in service to support grid reliability, and while inefficient, these plants can be economically viable due to capacity payments and high electricity prices at times of peak demand. The problem with this idea for Base Electron is that a steam plant is a suboptimal technology to provide these capabilities. As described earlier, steam plants are relatively slow to start and ramp. For true peaking capacity, aeroderivatives, smaller frame

turbines, and reciprocating engines are all more effective and flexible, and available at comparable or lower capital cost than the apparent $2000/kW capital cost of Base Electron’s plan⁹⁸.

North Dakota has better options to add capacity

North Dakota is rapidly adding both baseload and dispatchable capacity using technologies that are more efficient, more economical and less emissions-intensive than a gas steam plant.

Wind farms already provide approximately 35% of the electricity in the state and nearly 5GW of capacity. According to Clearview, there are 17 wind projects under development in North Dakota representing an additional 3.3GW of capacity,⁹⁹ along with nearly 2GW of battery storage projects.¹⁰⁰

While renewables and battery storage are increasingly competitive for many purposes, some gas-fired dispatchable resources may remain necessary. Pioneer Generation Station Phase IV is an example of a facility recently built in North Dakota focused on dispatch flexibility and grid stability.¹⁰¹ The addition to the Pioneer plant was permitted in July 2023, started construction in 2024, and entered service in 2025, including both a simple cycle combustion turbine and a set of reciprocating internal combustion engines to provide 580MW of capacity. If more flexible generation is required in North Dakota, we would expect electricity producers to continue to use similar technologies.

The ongoing deployment of new combustion turbines and engines in North Dakota and across the United States calls into question one of the underlying claims of B&W’s narrative – that boilers are needed because combustion turbines are unavailable. The reality is much more nuanced. While lead times for the largest turbines used in CCGT plants may have stretched to five years or more for orders being placed today, many combustion turbines are still being installed across the US. According to an EPC engineer, “the combustion turbines, and even engine generators, a lot of bridge power companies are securing those,”¹⁰² and many of the experts we spoke with noted that the availability of generation equipment depends enormously on who is buying it, long-term relationships with vendors, and ability to pay.

Project financing will depend on a pro forma model that makes sense

As an experienced IPP developer told us, “It’s all going to come down to the pro forma.”¹⁰³ We asked an expert who has developed power plants and now consults on new project evaluation and financing what an investor evaluating a power plant project would be looking for. They suggested that “the absolute minimum” would be a PPA commitment on paper with a counterparty that can be assessed, and a detailed financial model of the plant:

“I was a lender’s engineer, an owner’s engineer, and then I’ve been the owner. I’ve done a lot of refinancing, and what they look like is the sewing together of all the contracts… All the lawyers do is they go through and make sure everything is matched, right? They look at the permit. What’s your emissions?… Effectively, what you’ll need at a minimum is site control, permits... And then you have to have all your contracts effectively substantially negotiated... Every assumption that’s in the financial model has to tie back”¹⁰⁴

With no PPA, no site control, and no permits, we are skeptical that any party will be willing and able to finance this project. Applied Digital doesn’t intend to pay to build the Base Electron plant by itself. In a May 12, 2026 interview, CEO Wes Cummins confirmed: “We’ve been very deliberate about separating these businesses precisely because they carry very different risk profiles.”¹⁰⁵ Cummins also said: “Crucially, we’re not doing behind-the-meter generation, where you just build power right next to your own data center. We’re siting these plants on the grid itself, working directly with utilities in North Dakota, so that the infrastructure benefits all ratepayers across the broader grid, not just our campuses.”¹⁰⁶

Cummins’ suggestion that the plant is for the benefit of “all ratepayers across the broader grid” should be of concern to all ratepayers who would likely be paying the inflated cost of these resources for the next ~40

years. If Base Electron expects that ratepayers will finance development, Base Electron might also expect its project to be scrutinized within the regulatory frameworks that exist to protect ratepayers and to encourage financially responsible power development that serves the long-term interests of all North Dakotans.

In addition to the higher operational costs imposed by the inefficiency of the plant design, Base Electron will also compete for project finance with others now building new capacity in the Dakotas, including established utilities. Basin Electric, for example, a 65-year-old cooperative with three million customers, raised $700m in October 2025 through the issuance of 30-year bonds paying 5.85%.¹⁰⁷ The River Run Energy Center, now being planned as a generation resource for large loads, is a collaboration with NextEra, the largest IPP in the US. We expect Base Electron’s cost of capital to be significantly higher than these competitors, further challenging the economic viability of the proposed plant.

5. The B&W and Base Electron Deal Is a Tangle of Related Parties, Reciprocal Transactions, and Vaporous Guarantees

Base Electron appears to be controlled by B. Riley and Applied Digital insiders

Base Electron is a Nevada corporation that shared an address with B. Riley in Los Angeles when incorporated last December. By March, the record had been updated to reflect a new address in Dallas, which Base Electron now shares with Applied Digital.

Image: Base Electron Officers. Source: SilverFlume: Nevada Business Entity Search ¹⁰⁹

There has been no attempt to hide the relationship. While Base Electron may be a nominally separate company and is off the Applied Digital balance sheet, its board substantially overlaps with the Applied Digital management team and B. Riley affiliates: The President of Base Electron is Wes Cummins, CEO of Applied Digital, and other officers include Jason Zhang, President of Applied Digital; David Rench, former CFO of Applied Digital, and Richard Nottenburg, a director of Applied Digital. Base Electron’s VP of Finance¹¹⁰ is Zach Cummins, formerly of B. Riley.¹¹¹ Bryant Riley himself was listed as a Base Electron director until March 2026.¹¹²

Applied Digital’s guarantees and immediate paper gains

Since Base Electron is a new company with no apparent ability to pay for a $2.4 billion power plant, Applied Digital has guaranteed Base Electron’s obligations on the contract, now characterized as a “Full Notice to Proceed.”¹¹³

Yet what wasn’t disclosed until B&W’s 2025 10-K¹¹⁴ is that in connection with the deal, B&W issued equity and warrants to Applied Digital, some of which were later assigned to Base Electron itself. Those warrants all immediately vested when the “Definitive Agreement” was executed on Feb 26, 2026. By temporarily providing a “guarantee” for up to $50m or $100m, Applied Digital, together with Base Electron, to which approximately half of the warrants were later assigned,¹¹⁵ now appear to have a potential gain of over $100 million on their own B&W shares and warrants.

Note 16 to B&W 2025 10-K¹¹⁶

Also buried in Exhibit 10.5 to the B&W 2026 Q1 10-Q is an amendment to the Registration Rights Agreement stating that the shares issued to Applied Digital and Base Electron would be registered by May 29, 2026. Just eight days before this deadline, Applied Digital and Base Electron filed an S-3 to register their 10,900,000 shares of B&W and potentially monetize their deep in-the-money warrants struck at $4.11.¹¹⁷

Exhibit 10.5, B&W Q1 2026 10-Q¹¹⁸

The Applied Digital guarantee may be terminated¹¹⁹ upon the listing of Base Electron equity; Base Electron Raising at least $50m, or payment of a penalty of $50m (by Aug 1, 2026) or $100m after that.

B&W quietly raised the estimated project cost

In the initial November announcement, B&W described a 1 GW project valued over $1.5 billion. By March 4 2026, the deal had become a $2.4 billion project for 1.2 GW of new generation capacity. The 60% increase was never explained, and when asked on the March 16, 2026 conference call discussing the 2025 annual report, CEO Kenneth Young characterized the change as “kind of normal evolution.”¹²⁰

Applied Digital doesn’t need and hasn’t yet committed to offtake

Despite the initial announcement of the deal as delivering “one gigawatt of power for an Applied Digital AI Factory,” the entire structure of this deal appears designed to keep the project far off Applied Digital’s balance sheet and limit Applied Digital’s true commitment. In the most recent Applied Digital earnings call in April, Wes Cummins goes further to explicitly disclaim the idea that the power is specifically for an Applied Digital data center:

“But I think it’s worth reiterating. The Base Electron business model is actually adding grid power. It’s not building on-site generation specifically for the Applied Digital data centers, but it’s strategically adding it in places on the grid in North Dakota that will definitely feed the Applied Digital sites, but it’s meant to make the grid overall better and more resilient and be of benefit to all of the stakeholders and the ratepayers in the state and not just putting it on site to generate electricity for Applied Digital.”

The problem is that without a firm, long-term commitment by somebody to purchase the power at a fixed price, Base Electron is not going to be able to raise the $2.4 billion needed to ensure that the project gets built and to ensure that B&W gets paid. Investors could be forgiven for presuming that because Applied Digital and data centers are mentioned, Applied Digital will be taking the power or that the project is “behind-the-meter.” But Applied Digital appears to have neither the will nor the balance sheet to pay for a $2.4 billion power plant by itself, and already has power secured for its data centers currently under construction.

Who is going to want to pay for this power or pay to build the plant? Hearing the story, one energy executive reacted:

“I very much doubt this is a replicable model. If ... B. Riley [hadn’t] restructured this complicated commercial agreement for power takeoff on their own, they would not have sold this. I mean, no customer would have gone and said, hey, can you put four boilers in there for me.”¹²¹

But there isn’t even a commercial agreement for offtake yet, and B. Riley hasn’t sold the Base Electron project – all that has been sold so far is B&W stock. As a purported grid asset, without guaranteed off-take or a firm PPA, Base Electron’s plan deserves to be judged like any other grid generation asset. Unfortunately for

B&W, when evaluated using the technical and economic criteria used to plan new generation capacity across the US for the last 30 years, its story soon runs out of steam.

6. Current B&W valuation cannot be justified, with or without Base Electron

B&W’s core business is low-margin (under 10% EBITDA) sticky recurring revenue that includes parts and maintenance for thermal plants, environmental controls, and fuel conversions. The problem is that their core customer, coal power plants, are in secular decline leading to B&W’s history of revenue volatility and business challenges. Low-margin sticky recurring revenue businesses in secularly declining industries typically merit single digit EBITDA multiples.

B&W has recently discussed the value of its coal-to-gas conversion business, but we are skeptical of prospects for significant growth in fuel switching. While B&W is a leader in this area, the market opportunity appears to be limited to a relatively small number of the oldest, highest-emission coal plants that haven’t yet been converted or upgraded. Indeed, according to an article written by an expert at B&W itself, “it would appear that the coal-to-gas conversion market has peaked and is now on the decline”.¹²² And that was in 2015.

An engineer who has worked on coal-to-gas projects agreed: “it doesn’t change the capacity at all. You got the same capacity, you just increase the fuel cost and decrease the capacity factor because it’s economic fewer hours of the year, but the trade-off is the emissions are improved quite a bit [...] There are still projects being considered, I know, but it’s not like, let’s say, during a non-Trump administration, viewed coal-to-gas as being a way to lengthen [operating life] because [otherwise] you’re dead.”

“The projects were staring at a cliff, if you will, of being able to continue to operate if they didn’t convert to gas. That seems off the table in the current administration, so it has to rely more on the economic drivers. It doesn’t seem like the economic drivers are compelling.” While data centers have generally increased electricity demand, they might only drive the viability of coal-to-gas conversion if “it fits what they’re trying to do.”

While AI load growth has temporarily extended the operating life of some coal plants, this has primarily benefited B&W’s parts and aftermarket business rather than generating new coal-to-gas conversion projects. B&W’s most recent large conversion contract, AES Indiana, was awarded before the AI buildout and concludes in Q1 2027,¹²³ with no announced replacement projects of comparable scale.

Using management’s 2026 EBITDA guidance of $80-100M (which includes some Base Electron revenue), and considering B&W’s pension and warrant liabilities plus the new equity offering, results in an estimated value of $1.20-$3.60 per share at 6x-10x EBITDA for the existing business, in our opinion.

Best case, we believe that Base Electron will be a one-hit wonder for B&W.

Even if the Base Electron project does materialize, there is no modern history of B&W or anyone else building power plants like this, nor any public evidence of unaffiliated customers seeking to build these plants. For all of the reasons we have explained earlier, we do not expect the plants to be of interest to established power producers. Furthermore, the market for large gas turbines will eventually normalize, eliminating the CCGT supply bottleneck argument for relying on less-efficient boilers.

In our view, the market is making a critical error in assuming an EBITDA run-rate based on expected 2028 earnings from what is a speculative one-time project. We believe that best case, it is only appropriate to value B&W on a sum-of-the-parts of its core recurring business plus a one-time contribution from Base Electron.

While the headline number of $2.4 billion may sound like a lot, B&W’s EBITDA margin will likely be 5-10% for an EPC project (setting aside that B&W has historically lost money on big construction projects).¹²⁴ B&W’s most recent comparable greenfield power plant project, Dry Fork Station, was completed in 2011 — a 422 MW coal-fired unit roughly one-third the size of the proposed Base Electron facility.¹²⁵

A 10% best-case EBITDA margin on $2.4 billion in revenue results in total project EBITDA of approximately $240 million spread evenly across 2027-2029. Minimal further margin upside exists above 10%, as B&W’s EPC contract is fixed price for $435 million of $2.4 billion with the residual as cost-plus. Alternatively, if Applied Digital were to terminate its guarantee of the contract, B&W might earn a cancellation fee of up to $100m, also worth less than $1.00 per share.

The current $19 B&W stock price implies a 24x multiple on the new growth business and new Base Electron project every two years.

7. Conclusion & Policy Considerations

Sunlight believes that there are many lower-cost, lower-emissions pathways to support rapid growth of data center loads. Adding generation capacity and achieving speed-to-power are different problems, and the solutions only partly overlap. Those solutions vary with the size and type of load, available natural resources, reliability requirements, and project duration.

Technology evaluation must balance capital cost, operational risks, execution risks, and financial risks, as well as environmental externalities; the longer-duration and bigger the asset, the more carefully such concerns should be weighed. Data center developers, utilities, regulators, and investors should cooperate on a buildout that keeps electricity costs low for all users, and it would be negligent to re-carbonize the US grid with obsolete technology when better alternatives are available.

Growth in load demand and the capital now available to meet that demand ought to be viewed as an opportunity to improve the existing grid by judiciously deploying technologies that will increase efficiency, improve utilization of existing infrastructure, while minimizing overall emissions and avoiding higher rates.

Unfortunately for B&W, greenfield gas steam plants present a uniquely inefficient, inflexible, and expensive approach no longer suitable for base load generation. Building a new power plant — thermal or otherwise — is not a short-term prospect. While a handful of hyperscalers may be willing and able to independently sponsor development of co-located and even islanded utility-scale generation, those are the players with the capital, patience, and desire to thoughtfully and responsibly deploy efficient resources, including renewables and battery storage¹²⁶.

Even if a data center developer’s goal is instead to greedily solve for speed-to-power, boilers are outmatched by faster, cheaper, and more efficient ways to increase capacity: demand flexibility, interconnection reform, and transmission upgrades can all more quickly increase utilization and unlock stranded capacity. And while many projects are electing to deploy gas-fired generators behind-the-meter as bridging solutions, there are more responsible, modular, and capital-efficient ways to do so which can ultimately support the grid, without locking in the long-term costs and risks of B&W’s boilers.

Demand flexibility

Data center electricity demand can become more flexible by shifting computer workloads across facilities, reducing operations, or relying on backup power or on-site storage¹²⁷. The available capacity this unlocks is substantial: research from the Nicholas Institute at Duke University finds that the grid could reliably integrate 76 to 126 gigawatts of new data center demand with no additional need for capacity expansion if those loads accepted modest curtailments of just 0.25 to 1 percent of annual hours.¹²⁸ Google demonstrated demand response by reducing the power demand associated with machine learning workloads during grid stress events in a pilot with Omaha Public Power District, and has since extended the program to partnerships with Indiana Michigan Power and Tennessee Valley Authority.¹²⁹ Its carbon-intelligent computing platform reportedly shifts non-urgent tasks to data centers in other regions in response to forecasted grid stress events or high availability of renewable energy.¹³⁰

Admirably, Applied Digital already recognizes the benefits of being a flexible load. In a story describing Applied Digital’s expansion, the North Dakota Monitor reported that “the data center will have diesel generators to supply power if a storm or spike in power demand forces the co-op to divert power elsewhere. ‘They know that if we ask them to curtail their load that they’re going to have to do so,’ Matthys [the utility spokesperson] said. He described the data center as ‘first off and last on’ if there is a disruption.”¹³¹

Interconnection reform

Engineering and economic considerations will drive most new generation and most new loads to be grid connected. Keeping generation on the grid ensures a market for its energy and mitigates stranded asset risk. Powering new loads on the grid allows those loads to benefit from reliability, market pricing, and grid stability and inertia.

Total US interconnection queue capacity stood at approximately 2,300 gigawatts in 2024, driven in part by speculative applications. In 2023, FERC introduced reforms aimed at addressing this, requiring interconnection customers to demonstrate site control, pay increasing financial deposits and withdrawal penalties.¹³² On October 23, 2025, the Secretary of Energy further directed FERC to accelerate interconnection of large loads, proposing that customers be able to file joint, co-located load and generation interconnection requests directly with FERC, in addition to reducing study times and grid upgrade costs.¹³³

Such reforms, while overdue, are beginning to bear fruit. As discussed earlier, in North Dakota MISO and SPP have created expedited queues for both generation resources and large loads, and these programs are already being used to accelerate interconnection studies for qualified projects prepared to demonstrate site control and financial commitment.

Grid utilization

Much of the US grid generation capacity is substantially unused for most of the year because the grid is engineered to reliably serve short, infrequent periods of peak demand. The Brattle Group’s March 2026 report, “The Untapped Grid,” found that a representative mid-sized US investor-owned utility operates at only approximately 43% generation capacity utilization.¹³⁴ The same report estimates that improving system utilization by 10% could reduce electricity rates by approximately 3.4% relative to current levels, and that better utilization of the existing grid could save US consumers more than $100 billion over the next decade.¹³⁵

A utility that met new data center load growth through a combination of flexible tariffs and distributed energy resources rather than new infrastructure investment would reduce average customer rates by 4.8% compared to the status quo approach of simply building new poles and wires.¹³⁶ Technologies such as dynamic line ratings and advanced conductors can also unlock substantially more throughput from existing rights-of-way at a fraction of the cost of new construction.¹³⁷

In North Dakota, Applied Digital itself has demonstrated the value of accessing stranded renewable capacity when it secured more than 400 megawatts of stranded wind power for its Ellendale site.¹³⁸ Transmission system upgrades financed by users such as Applied Digital seeking to add large loads are also an effective way to improve capacity utilization without imposing costs on rate payers¹³⁹.

Behind-the-Meter Power

Demand flexibility, interconnection reform, and transmission upgrades to improve grid utilization do depend, to varying degrees, on coordination between regulators, utilities and electricity customers.

What a data center developer can do unilaterally, immediately, and without waiting for policy reform, grid upgrades, or utility approval is deploy its own behind-the-meter power, often in partnership with companies specialized in providing power-as-a-service or microgrids. The market has already demonstrated this at scale through many projects deploying fleets of modular generating assets like gas reciprocating engines and aeroderivative turbines.¹⁴⁰

The small frame turbines, aeroderivatives, and reciprocating engines used in these projects are highly flexible; they are as or more efficient than gas steam plants and can ramp quickly and operate at part load. Modular generators can be added incrementally with demand growth, and multiple generators can support reliability needs. Although far from optimal from an efficiency or emissions perspective, negative impacts will be mitigated by the fact that most such deployments are expected to be temporary bridging solutions. Once grid generation and transmission does catch up, we expect that nearly all data centers will run on grid-supplied power, and modular gas gensets will likely be repurposed or used to provide backup or even provide peaking capacity back to the grid. Such solutions can also be deployed very quickly: some reciprocating engine solutions can be delivered in 12 to 18 months, and refurbished aeroderivative and modular turbines can offer an even faster route for operators unwilling to wait for new OEM lead times of three to five years.¹⁴¹

The “Stampede to Gas”, as a recent IEEFA report¹⁴² has dubbed the present situation, may ultimately expose energy users to gas pricing risks and volatility. Wind and solar generation supported by large-scale battery deployment should be preferred as a sustainable and increasingly economical path. Yet there is little doubt that some amount of new gas-fueled generation will continue to be necessary to ensure reliable dispatchable capacity, and policy and market conditions remain favorable for gas in the US.

But if new gas generation assets are to be built and deployed, that should be done efficiently with technologies suited to task. For baseload capacity the grid will rely on for decades, that means investing in the most efficient combined-cycle plants possible, not rushing to build obsolete plants with less flexibility and higher emissions. Achieving genuine speed-to-power will likely continue to rely on modular turbines and engines, for which environmental concerns are at least partly mitigated by their temporary role as a bridge to eventual grid connections.

B&W’s boilers offer none of the advantages of speed, modularity, efficiency, or scalability of bridging power, and instead ask investors – and possibly ratepayers – to underwrite a 30-to-40-year power plant that most likely will still take years to build.

Disclaimer

By accessing this report through our website, you have agreed that you have read, understand, and agree to the following Terms of Use, which are substantively set forth herein, and you represent that you have sufficient investment sophistication to critically assess the information, analysis and opinions presented herein.

Sunlight Partners Research LLC (“Sunlight Research”) has prepared this report. As discussed in more detail below, Sunlight Research produces research reports based on its own journalistic due diligence and investigation into publicly traded companies. Sunlight Research is under common control and affiliated with Sunlight Partners Pty Ltd (“Sunlight Fund”). Sunlight Fund is a hedge fund that invests in securities and makes other types of investments on behalf of its investor clients, thereby seeking to obtain a return on its clients’ investments. Sunlight Research and Sunlight Fund, respectively, means those entities and their respective affiliates and related parties, including, but not limited to any principals, officers, directors, employees, members, clients, investors, consultants and agents.

As of the date of publication of this report, Sunlight Fund holds a short position in the securities of Babcock and Wilcox Enterprises Inc. and/or in derivatives or other financial instruments linked to those securities (the “Covered Issuer”). Accordingly, Sunlight Fund stands to realize substantial gains in the event that the price of the Covered Issuer’s securities declines. Following the publication of this report, Sunlight Fund may transact in the securities covered by the report, including reducing, covering, increasing, or otherwise changing positions at any time for any reason, without notice.

Sunlight Research employs an investment strategy commonly referred to as “research activism,” pursuant to which it conducts independent journalistic research and analysis of publicly traded companies based on publicly available information and its own investigation and due diligence. In general, Sunlight Research’s business model involves identifying issuers that it believes are misvalued or as to which the market’s consensus view is incomplete or inaccurate. It issues investigative research reports highlighting such issues for the general public and market participants. Sunlight Research is a for-profit journalistic organization, which funds its in-depth investigative reports by taking positions in the issuers about which it writes before publishing its reports. Sunlight Fund generally covers some or all of those positions following publication of Sunlight Research’s reports to realize gains from the change in price of the issuer’s securities. This business model involves significant financial risk for Sunlight Fund and its clients. To manage that risk, Sunlight Fund will close open positions as it deems prudent.

This report reflects the opinions and analysis of Sunlight Research as of the date of publication, based on its journalistic investigation. Sunlight Research is providing its opinion on and analysis of the securities of the Covered Issuer to contribute to the mix of overall information available to the general public, based on the facts detailed in the report. The information and views expressed in this report are based on publicly available information, field research, and other sources believed to be accurate and reliable and based on sources who are not insiders or connected persons of the stocks of the Covered Issuer or who may otherwise owe any fiduciary duty or duty of confidentiality to the Covered Issuer. The information presented herein is “as is,” without warranty of any kind, whether express or implied. All opinions are subject to change without notice for any reason. Investor strategies or trade ideas are intended for illustrative and educational purposes only and are not indicative of historical or future performance or the chances of success or any particular investment end or investment strategy.

Once a research report is released, market participants are expected to review and analyze the merits of the arguments in the report. Readers rely on the information in this report at their own risk. Readers should conduct their own independent investigation and analysis, review other publicly available information about the Covered Issuer not included in the report, and consult with their own professional advisors before making any investment decisions.

Following publication of the report, Sunlight Fund will continue transacting in the securities of the Covered Issuer for an indefinite period and may be long, short, or neutral at any time regardless of the initial views expressed in this report. Sunlight Fund may, at any time, for any reason, and without notice, increase, decrease, close, or otherwise change its positions, including before or after the publication of this report. Sunlight Fund undertakes no obligation to update or supplement this report or to disclose any such changes. Sunlight Fund’s trading activity may be conducted before, during, or after the publication of this report, and such activity may be consistent or inconsistent with the views expressed herein. By the time a reader reads this report, Sunlight Fund may be covering or have already covered its short position. A reader should not

make any investment decision based on its own view or interpretation of Sunlight Fund’s positioning in the Covered Issuer’s securities.

Sunlight Fund may from time to time enter into financing, stock borrow, derivative, swap, risk participation, capital support, or other economic arrangements with third parties, including institutional counterparties and investment partners, relating to positions referenced in a report. As a result, persons other than Sunlight Fund may possess direct or indirect economic interests in positions discussed herein. Any such financing counterparties, lenders, capital providers, or other market participants may maintain independent investment views or trading positions, and Sunlight Fund and Sunlight Research undertake no obligation to provide any information or update regarding the activities or positions of any such persons.

This report is provided for informational purposes only and does not constitute investment, legal, tax, or other advice, nor does it constitute an offer to sell or a solicitation of an offer to buy any security or financial instrument. Sunlight is not acting as an investment advisor to any recipient of this report, and no fiduciary relationship is created by virtue of a reader’s review of this report. Sunlight Research is not registered as an investment advisor with the Securities and Exchange Commission. Sunlight Research’s reports articulate its reasons at the time of publication Sunlight Fund has taken its short position in the securities of a Covered Issuer.

Sunlight Research does not provide “price targets,” although it may express its opinion of what the security is worth. An opinion of the value of a security differs from a price target in that Sunlight Research does not purport to have any insight as to how the market as a whole might value a security. Sunlight Research is only expressing its own view of its value. Sunlight Fund thus does not hold a position until it reaches a certain price target, nor does it hold positions until they reach the price at which it has expressed a valuation opinion. There are numerous factors that enter into investment decisions aside from opinions of the value of the security, including without limitation, the borrow cost of a shorted security, the potential for a “short squeeze,” prudent risk sizing relative to capital and volatility, reduced information asymmetry, the opportunity cost of capital, client expectations, and the ability to hedge market risk. Again, a reader should assume that upon publication of any report, Sunlight Fund will, or has begun to, close a substantial portion—perhaps all—of its positions in the Covered Issuer’s securities.

Sunlight Research has relationships with third parties who have contributed to or assisted in the preparation of this report, including through research, analysis, or other services. Sunlight Research also has relied on various sources of information in generating this report. Any such relationships, including any compensation arrangements, are disclosed in this report where material. Readers should assume that such relationships create potential conflicts of interest.

To the fullest extent permitted by law, Sunlight Research shall not be liable for any claims, losses, costs, or damages of any kind, including direct, indirect, punitive, exemplary, incidental, special or consequential damages, arising out of or in any way connected with this report. This limitation of liability applies regardless of any negligence or gross negligence of Sunlight Research.

Readers agree that any dispute arising out of or relating to this report or the information contained herein shall be governed by the laws of the State of New York, without regard to its conflict-of-law principles. Readers further agree to submit to the exclusive jurisdiction of the state and federal courts located within the State of New York. To the fullest extent permitted by applicable law, any claim arising out of or relating to this report must be brought within one (1) year after the date on which the facts giving rise to the claim were discovered or reasonably could have been discovered; provided, however that nothing herein is intended to limit or waive any rights that cannot be waived under applicable securities laws.

The failure of Sunlight Research to enforce any right or provision in this report shall not constitute a waiver of such right or provision. If any provision of these terms is held to be invalid or unenforceable, the remaining provisions shall remain in full force and effect.

For any reader located in the United Kingdom, by accessing this report through our website, you confirm that you are accessing research and materials as or on behalf of: (a) an investment professional falling within Article 19 of the Financial Services and Markets Act 2000 (Financial Promotion) Order 2005 (the “FPO”); or (b) high net worth entity falling within Article 49 of the FPO (each a “Permitted Recipient”). In relation to the United Kingdom, the research and materials in this report and on our website are being issued only to, and are directed only at, persons who are Permitted Recipients and, without prejudice to any other restrictions or

warnings set out in these Terms of Use. Persons who are not Permitted Recipients must not act or rely on the information contained in any of the research or materials in this report and/or on our website.

This report is the property of Sunlight Research. Readers agree that they will not communicate the contents of this report and other materials on Sunlight Research’s website to any other person unless that person has read and agreed to be bound by these Terms of Use. By accessing, reviewing, reading, downloading, receiving, or relying on the information contained in this report, a reader is agreeing to be bound by these Terms of Use.

The information in this Report and on our website is copyrighted. A reader therefore agrees not to distribute this information in any manner other than by providing a link directly to the report on our website. If a reader has obtained this report through any other method, the reader may not use this research report for any purpose without agreeing to these Terms of Use.

Footnotes