Babcock & Wilcox

IThe business

Babcock & Wilcox is one of the oldest names in power equipment. Founded in 1867, it designs and builds boilers, the large steel vessels that turn fuel and water into steam, and it sells the parts, upgrades and field service that keep existing plants running.³⁴ It also sells pollution-control equipment and has a small hydrogen technology line called BrightLoop. It has about 1,650 employees, plants in the United States, Canada and Mexico, and in 2025 71% of its revenue came from customers in the United States and 87% from the United States and Canada together.³⁴ The head office is in Akron, Ohio. The year is the calendar year.

For most of the last ten years it was shrinking. Revenue as filed fell from $1,578M in 2016 to $588M in 2025 as it sold or shut business units, among them its European renewable arm, SPIG, Diamond Power, ASH and solar, and it lost money in nine of the ten years C (calculated).⁹⁵³ Shareholders’ equity went from +$552M at the end of 2016 to −$132M at the end of 2025.⁵ The 2025 10-K says earlier conditions raised substantial doubt about whether it could continue as a going concern, and that asset sales and a refinancing in 2025 relieved them.³ I start from that, because it is the business the market was pricing a year ago.

The change came in November 2025, when it signed a limited notice to proceed with Applied Digital, an AI data-center developer, and on 26 February 2026 a $2.4B contract with Base Electron, a company tied to Applied Digital, to design and install four 300 MW gas-fired plants.¹⁰² Backlog went from $424M at the end of 2025 to $2.57B at 30 June, and revenue for the second quarter was $320M against $139M a year earlier C (calculated).³²

How it earns

The company reports three kinds of revenue. Parts are the steady one: replacement components and service for boilers already running, $242M in 2025, 41% of revenue, and up 4% in the first half of 2026 C (calculated).³² Projects are engineered equipment such as new boilers and environmental systems. Construction is the installation work. Over the twelve months to June they were $346M and $242M. Together they more than doubled, from $166M in the first half of 2025 to $408M in the first half of 2026, and Base Electron added $132M of that C (calculated).² Take Base Electron out and the second quarter’s revenue was $219M, still 58% above a year earlier C (calculated).² So the older business is growing too, from a low base, helped by the same demand for power.

What it earns on each is the question that matters. Adjusted EBITDA was 7.4% of revenue in 2025 and 6.8% in the second quarter C (calculated).³² Cost of operations took 85% of second-quarter revenue against 70% a year earlier, so gross margin fell from 29.9% to 14.6% as the low-markup Base Electron work came in C (calculated).² The company also says construction productivity has been lower than expected because of a nationwide shortage of skilled labour.² More revenue at a thinner margin is not worse by itself. It does mean the revenue figures alone say little about what reaches owners.

The Base Electron contract

One contract now decides the story. I reproduce what the filings say, in the table below. It is a design-and-build contract. B&W gets a fixed fee of $434.8M for its own scope, and for the rest, mostly equipment and labour bought from others, it passes the cost to the customer and adds a markup.⁶ The markup is not disclosed. In the contract filed with the SEC the percentage is blacked out, as are the caps on damages and the termination terms.⁶ I can’t see the three things a lender would ask for first.

The second thing to see is who pays. Base Electron is not a utility. Applied Digital describes it as a variable interest entity that does not have enough equity at risk to finance itself, owned and managed in part by certain Applied Digital officers and directors acting in their own names, two of whom are Base Electron’s chief executive and president, with Applied Digital holding about 10% for $2.0M and lending it $58.5M on a demand note.¹¹ On 4 October Applied Digital signed a 15-year power purchase agreement with Base Electron for a 1,200 MW plant in North Dakota, with fixed monthly payments, and Base Electron may end it if financial close has not occurred by 31 March 2027.¹¹ B&W’s own investor slides describe a 1.2 GW project near Center, North Dakota, built from four 300 MW units.⁴ Applied Digital’s 10-Q describes a 1,200 MW natural gas-fired plant in Center, North Dakota, to be developed by Base Electron; B&W’s contract is with Base Electron for a 1.2 GW project near Center. The filings do not call them one project in so many words, and I treat them as one. If so, the plant’s buyer is a data-center company whose own cash was $2.9B at 31 August, and the plant’s owner is a young special-purpose company. Applied Digital’s 10-Q does not say how the plant will be paid for, and the right to end the power agreement if financial close has not occurred by 31 March 2027 implies, without stating, that it had not closed on 4 October. I can’t see anything filed since.¹¹

The third thing is that B&W has been paid partly in the customer’s own currency, in a sense. Applied Digital got 0.5M B&W shares for $2.0M and warrants for 10.46M shares at $4.11 a share, running seven years, and 5.23M of the warrants were later assigned to Base Electron.² B&W books them as a liability, $136.9M at 30 June when the shares were $14.10, and as an asset of $68.4M that it writes off against revenue over the life of the contract.² In plain words, part of the price of winning the contract was 7% of the company, at an exercise price near today’s share price.

How it got here

The cleanup was real. In 2025 B&W sold Diamond Power for $177M, its ASH unit for $29M and Vølund for $15M, abandoned the solar unit, and swapped most of its 2026 notes into $129.5M of 8.75% second-lien notes due 2030, an exchange it accounts for as a troubled debt restructuring.³² In February 2026 B. Riley’s guarantee of the bank facility was cancelled and the Axos facility was extended to January 2028.¹² In July it called the last $61.4M of 6.50% notes, due December 2026, for redemption on 13 August.¹³⁷ It has sold $259.8M of new stock in the first half of 2026 alone.² The result is a balance sheet that no longer has a maturity before 2028. I come back to what it paid.

IIThe moat

What keeps a rival from earning what B&W earns? I look in three places. The first is real, the second is smaller than it looks, and the third is not visible in the numbers.

The installed base. B&W says its installed base is more than 400 GW of generating capacity around the world.⁴ A plant that runs on a B&W boiler needs parts and upgrades for decades, and B&W holds the drawings. That is the parts line: $242M of revenue in 2025 and a steady 4% growth in the first half of 2026 C (calculated).³² It is the best part of the business. It is not locked in. The 10-K says B&W also sells parts for the installed base of other manufacturers, which suggests others can supply parts for its base too, and the company names fifteen competitors, including GE Vernova, Mitsubishi Power, Doosan and Fluor.³

The right to bid. A project needs bonding and letters of credit. B&W had $246M of surety bonds and $39M of domestic letters of credit outstanding at 30 June.² Not every contractor can post those, and the 10-K lists the ability to post them as a risk if liquidity runs short.³ With cash above debt the right to bid is secure for now. I count it as an entry ticket, not a barrier to others.

Pricing power. This is where a moat would show, and I can’t find it. The 10-K lists the things customers choose on: price, technical capability, quality, timeliness, breadth of product and “willingness to accept contract risks.”³ A business with pricing power is not usually chosen for being willing to take risk. The Base Electron markup is blacked out. Gross margin fell from 29.9% to 14.6% year on year in the second quarter as that work came in, which says the new revenue earns less per dollar than the old C (calculated).² A moat would show up the other way round.

What would show one is a second AI-power customer on better terms than the first, or margins that rise as the volume rises. Neither has happened yet.

IIIThe competitive advantage

The advantage B&W claims is speed. Its slides say a boiler and steam-turbine plant can be deployed years faster than a traditional gas-turbine project, because it does not wait for gas turbines, and that a plant can start as a steam plant and add a gas turbine later.⁴ It has also secured Siemens Energy steam turbines for a further 1 GW, for fast delivery to a next data-center project that it has not named.⁷ The 10-Q does not mention Siemens or any purchase commitment, so I can’t see what B&W has paid or owes for those turbines, or whether a customer is attached.² I take the claim seriously and note one thing against it: the first plant is due in 2030, four years after the contract was signed.² A faster build than that is possible, and nothing in the filings shows it yet.

The second advantage is the repaired balance sheet. At 30 June B&W had $308.6M of cash and $74.2M of restricted cash against $239.8M of secured debt and bonds, which the company says leaves cash above debt by about $143M, before the 6.50% notes were paid in August.²⁴ At the end of 2025 total debt was $321.1M, including $84.8M of 6.50% notes due in December 2026 that the 10-K flagged as a refinancing risk.³ It is an advantage over B&W’s own past, not necessarily over competitors.

The third is where the demand is. Management says power demand from data centers, utilities and industrial customers is lifting parts and service as well as new builds, and the company’s pipeline of identified opportunities is over $14B across three years.⁷ The pipeline counts things that have not been decided, and the company says it may not generate the margins of past work.⁷ Backlog without Base Electron fell from $424M at the end of 2025 to about $301M at 30 June C (calculated).³² I read that as a reminder that most of the growth to date is one contract.

IVManagement

Kenneth Young, 62, has been chief executive since November 2018 and chairman since September 2020. Before that he was president of B. Riley Financial and ran its principal-investments arm.⁸ Cameron Frymyer has been chief financial officer since January 2025.⁸ The general counsel, John Dziewisz, left on 31 May 2026, and a former executive officer, Jimmy Morgan, ceased to be one in 2025.⁸² Mr. Dziewisz, after leaving, filed notices on 10 and 12 August to sell 5,963 and 410,862 shares, about $4.0M in all.²⁴ Mr. Young is credited in the proxy with cutting senior debt by more than $200M, buying businesses, and launching the BrightLoop line.⁸ The first of those is visible in the filings. The 2025 sales of Diamond Power, ASH and Vølund carried base prices of $177M, $29M and $15M, $221M in all C (calculated).³

How they are paid

Mr. Young’s 2025 pay in the summary table was $1.92M, with $3.55M “actually paid” once the stock rise is counted.⁸ Until 20 September 2024 he was not on B&W’s payroll at all. B. Riley paid his salary, and B&W paid a B. Riley affiliate $62,500 a month for his services.⁸ The annual cash bonus for 2025 rested on a single measure, Adjusted EBITDA; the long-term awards are restricted stock units and performance units.⁸ A 2022 grant of stock units vested on 5 March 2026, when the shares closed at $13.29, above the $12 goal.²

That design pays people to grow Adjusted EBITDA and to make the share price go up. Adjusted EBITDA leaves out stock pay, pension cost, restructuring and, from 2026, the cost of the customer warrants.⁷² It does not count cash left after pension payments, the capital the business needs to stand still, or the number of shares. A manager paid this way has a reason to win work and to fund it with shares, because each dollar of revenue adds to Adjusted EBITDA before it adds a dollar of cash. I do not say that is what is happening. I say it is what the pay rewards, and I watch it next to the owner cash numbers below.

What they own, and what the largest holder does

Mr. Young owned 1.72M shares, 1.3% of the company, at 23 March, worth about $7.3M at $4.27, and all officers and directors together owned 2.8% C (calculated).⁸ He has bought in the open market on three days since November 2025: 3,175 shares near $7.14 on 12 November, 7,000 at $15.145 on 18 March and 7,000 at $9.65 on 12 August, about $196k in all C (calculated).¹⁴ The chief financial officer bought about $39k. Other officer filings are tax withholding on vesting; I found no open-market sale by a current officer since November 2025.¹⁴ Those purchases are small next to the stock he holds, and they are a fact, not a forecast.

The largest holder is B. Riley, through BRC Group Holdings: 27.4M shares, 20.4%, at 23 March, and about 19% at 30 June.⁸² It has the right to nominate a director and to buy shares first in any new offering, and it earns fees from the company. It receives 3% of the gross on shares sold through the at-the-market programme, $1.3M in the first half. It led the underwriting of the May offering, which closed on 18 May, and the discount was $11.8M. Under an advisory agreement it earns 3% of any qualified debt financing plus $0.5M, and was paid $5.0M in the first half. It guaranteed the bank facility for about $3.0M a year until the guarantee was suspended in June 2025 and cancelled in February 2026.² B. Riley’s last reported sale of B&W stock was 1,155,382 shares at $9.00 on 11 February; I found no later Form 4 or Schedule 13D amendment from it through 30 October.¹⁴ The chief executive’s former employer is the company’s largest holder and its adviser, and has the right to name a director. I think that deserves saying plainly, and not as a charge of anything.

What they did with capital

Over the ten years to 2025, operating cash flow was −$1,058M and net income −$1,638M C (calculated).⁵ In 2015 and 2016, under earlier management, the company spent $25.4M and $78.4M buying back its own stock, then lost most of its equity.⁵ Since 2021 it has sold $559M of new stock, in 2021, 2024, 2025 and the first half of 2026 C (calculated).⁵² The share count rose from 54.5M at the end of 2020 to 149.0M at the start of August 2026, up 173%, and the share price over the same stretch is up 22%, from $3.51 to $4.27 C (calculated).⁵²¹ A holder from the end of 2020 has seen the share count nearly triple while the price rose 22%. The most recent sale, the May offering, was priced at $18.50 a share, 4.3 times today’s price.¹⁵ I do not blame management for the share price. I note that the company raised money when the stock was high, which is when any company would want to, and that anyone who bought at $18.50 is now down 77% C (calculated).

Two more things show how money is handled. In the second quarter the company paid $5.0M in cash to settle part of its directors’ vested stock units for their tax, at about $21.22 a share.²¹⁴ That is routine, and it happened the day after the company sold stock to the public at $18.50. A $50M buyback was announced on 13 July, when the shares closed at $10.93, to start after the second-quarter report.¹³¹ I can’t see whether it has bought anything since, because the third-quarter report is not out. The pay-for-performance table in the proxy shows $100 invested at the start of the period worth $178.63 at the end of 2025, against $282.63 for the proxy’s peer group.⁸

VFinancials

The history is awkward because the company keeps changing shape. The table below gives each year as it was filed, so revenue was $1,578M in 2016 and $588M in 2025 on a business that sold off more than half of itself in between. On today’s perimeter, revenue was $587M in 2023, $581M in 2024 and $588M in 2025, and Adjusted EBITDA $5M, $21M and $44M C (calculated).³ Over the twelve months to June 2026, revenue was $834M and Adjusted EBITDA $64M.⁴

The improvement is real. Operating income went from a $20M loss in 2023 to a $21M profit in 2025, and the second quarter of 2026 earned $11.8M against $7.0M a year earlier.³² Adjusted EBITDA margin has moved from 1% to 7% in three years C (calculated). What has not changed is the ability to turn it into cash.

What an owner could really take out

Very little, so far. Operating cash flow less capital spending was −$1,170M over the ten years to 2025 and −$58M over the twelve months to June C (calculated).⁵² In ten years the company has had one profitable year, 2021, on net income of $31M, and in that year operating cash flow was −$111M.⁵

The first half of 2026 looks better on its face: operating cash flow was +$0.4M. But it includes a $100.3M rise in accounts payable, money owed to suppliers, which I think comes mostly from the Base Electron build although the 10-Q does not say, and a $60.9M rise in receivables, money owed by customers.² Without the payables rise, operating cash flow would have been about −$100M C (calculated). Payables of $169.6M at 30 June against $69.2M at December are a loan from suppliers that falls due as the plant is built. If Base Electron pays late while B&W pays suppliers on time, the gap comes out of B&W’s cash.² That is the first thing I would watch.

Quality of earnings

Adjusted EBITDA is not owner cash. For the twelve months to June it was $63.8M against GAAP operating income of $25.7M C (calculated).⁴ The $38M between them is depreciation of $10M, stock pay of $16M, restructuring of $3M, financial advice of $4M and a few smaller items.⁴ Net loss from continuing operations was $75.1M, of which $64.3M is a non-cash loss because the warrant liability rose with the share price, and the price has since fallen.⁴ I would not call the warrant charge a real loss for owners, but I would count the warrants themselves as a cost, and I do in my valuation.

Three other items sit between Adjusted EBITDA and cash. The pension plans were $174M underfunded and cost $25.2M in cash in 2025 against $12.2M the year before.³ Interest cost $37.5M in 2025 and $9.5M in the first half of 2026, as the debt was cut.³² Preferred dividends take $14.9M a year.³ Add stock pay and the capital the business needs to stand still, and a company that reports $44M to $92M of Adjusted EBITDA is likely to leave something like $0 to $40M a year before interest and preferred dividends, even if it does well. My base case below shows $6M in 2027 and $42M in 2030.

The company also reports material weaknesses in internal control in three areas as of 31 December 2025, and says they had not been remediated at 30 June 2026. Its chief executive and chief financial officer concluded that controls were not effective but that the statements present fairly.³² I take the numbers as filed and keep that in mind when I look at revenue recognised over time on cost-to-cost estimates, which is 97% of revenue.³

The balance sheet

The claims ahead of the shares are listed in the table. At 30 June there was $308.6M of cash and $74.2M restricted, against $239.8M of secured debt and bonds on the company’s own count, and $191.7M of preferred stock.²⁴ After the August redemption of the 6.50% notes, face debt including leases is about $171M, with the preferred stock on top. I set the claims out so that one number stands beside the market value of the shares: preferred stock, debt and leases of $424M at 30 June, before $309M of cash C (calculated).²

Cash is the cushion. On my base path it stays above $200M; on my bear path it is down to about $11M by 2030 after interest and preferred dividends E (estimate). The company has a $50M buyback authorised, which would use some of it, and the third-quarter report, expected on 9 November, will show whether any was spent.¹³¹ B&W says it can fund the next twelve months from its operating plan and the Axos facility, and may raise capital beyond that.²

The share price is the other side of the table. It was $165.90 at the end of 2016, $3.90 at the end of 2018 and $0.26 on 21 April 2025, weeks before the exchange of its notes that it books as a troubled debt restructuring.¹² It reached $21.85 on 15 May 2026 and has since fallen 80%, to $4.27. Part of the fall is the group. Since 30 June, Applied Digital is down 52%, BWX Technologies 33%, Bloom Energy 33%, Vertiv 31% and GE Vernova 27%, on Nasdaq closes C (calculated).¹⁶ B&W is down 70% over the same four months.¹ I found no filing, insider sale or holder filing since 12 August that explains the extra drop, and the last 8-K is a preferred dividend notice on 8 September.¹⁷ I will not write a story to fill that gap.

VIValuation

First, how this goes badly wrong

Before any numbers, I ask what would most likely ruin the business or the investment, and work back. Here there are three.

The customer cannot pay for the plant. Base Electron is 88% of the $2.57B backlog and 31% of second-quarter revenue, and B&W was owed $73.9M by it at 30 June C (calculated).² Its buyer is a young company with $2.0M of Applied Digital equity behind it, a $58.5M demand note, and a power agreement that it can end if the money is not raised by 31 March 2027.¹¹ If financing fails or is delayed, B&W is still owed for work done under the contract, but it loses the margin it expected, the plant stops being the story, and any unpaid receivable becomes a loss. Applied Digital is also down 52% since 30 June, though it reported $2.9B of cash at 31 August.¹⁶¹¹

The work earns less than I think. The markup is blacked out. The fixed-fee scope, $434.8M, carries delay and performance damages, and B&W says labour is short and productivity lower than expected.⁶² Gross margin on all revenue fell to 14.6% in the second quarter from 29.9%.² In my model one point of margin on everything moves the base value by about $0.70 a share, about half of the value itself C (calculated).

The shares are the wrong claim at the wrong price. The stock has almost tripled in number since 2020, from 54.5M to 149.0M, and there are 10.46M warrants on top.⁵² If the contracts need cash before they pay, new shares may have to be sold much cheaper than the $18.50 of May. A good business can still be a poor purchase. Here I worry more about the price and the claims ahead of the shares than about the plant itself.

So the questions for everything below are: does Base Electron get financed, what does the work really earn, and how much of a much larger company is already in the price.

What an owner gets today

Owner cash over the past twelve months was −$58M against a net loss of $18M C (calculated).² In my base case it is $6M in 2027 and $42M by 2030, against Adjusted EBITDA of $114M. About 37 cents of each Adjusted EBITDA dollar reaches the owner E (estimate). The rest goes to stock pay of about $13M, pension cash of $25M, capital spending of about $25M, cash tax and other cash costs.²³ Interest of about $13M and preferred dividends of $15M come out of that before the shares see anything.³² The company pays no common dividend.

What I can’t judge

Four things matter most and I can’t know them from the filings. First, the markup on the Base Electron variable charges, which is blacked out, and so what the next contract would pay.⁶ Second, whether Base Electron gets financed, and how strong its owners are beyond what Applied Digital states.¹¹ Third, what B&W has committed to Siemens Energy for the further 1 GW of steam turbines, since the 10-Q does not disclose it and the release says only that B&W has secured them.²⁷ Fourth, how many such contracts follow, which is the biggest number in the model. For these I show a range, not a false precise number. In the bear I add $70M of credit loss and overruns, and in the bull I give B&W three times my base volume.

How I value it

I take owner cash for each year from the second half of 2026 to 2036, plus a value at 2036 that grows 0%, 1.5% or 2.5% a year in the bear, base and bull, all discounted to the price date. Owner cash is Adjusted EBITDA less stock pay, pension cash, other cash costs, capital spending, cash tax and a working-capital charge of 6% of added revenue, before interest. Adjusted EBITDA is built from the older business ($720M, $650M or $780M of 2027 revenue at a margin of 9%, 7.4% or 10.5%), Base Electron revenue on the schedule in the backlog note at 4.5%, 3.0% or 6.5%, new AI-power contracts at $2.0B a GW ($2.2B in the bull) spread over five years from the award, and a small service stream of 1.0% to 2.0% a year of project revenue booked two or more years earlier at 20% to 28%.²³ Cash tax uses about $640M of US federal tax losses, which I estimate from the $135.8M deferred tax asset at 21% and which the company carries a full allowance against, offset against 80% of taxable income a year, plus $6M a year of foreign and state tax.³ Claims are subtracted at face: debt and leases, the preferred stock at $25 a share, and the customer warrants as a call option on the value per share with 100% volatility and 6.3 years, the company’s own inputs.² Stock pay is a cost in owner cash, so I do not also add shares for future grants. Values are per share on 150.5M shares, which includes unvested stock units.

The discount rate. The ten-year Treasury yield was 4.94% at the price date and Damodaran’s implied equity risk premium 4.20%.¹⁸¹⁹ B&W’s beta against the S&P 500 on month-end returns is 0.88 over three years, 1.19 over five and 1.93 over ten, which gives a cost of equity of 8.6% to 13.0% C (calculated).¹²⁰ The three windows disagree so widely that none of them tells me much, and the shares rose 83 times in 13 months and then fell 80% in five, which beta does not describe. At market weights, with the notes at 8.75% and the preferred at 7.75%, and a beta of 1.4, the blended cost of capital is about 9.8% C (calculated). I use 13%, three points above that, because the company’s filings say going-concern conditions existed when it restructured its notes in May 2025, because there are material weaknesses in its controls, and because 88% of the backlog is one customer.³² That premium is large on purpose. At 10% the base value is $2.30, and at 16% it is $0.90, so the rate does not change the conclusion.

Bear

Base Electron does not reach financial close on time and the plant is built a year late and at a thinner markup, with a $40M receivable loss in 2027 and $30M of overruns on the fixed-fee scope in 2029. B&W wins one more contract of about 300 MW in 2031 and nothing else. The older business holds at 2025 margins, 7.4%, and grows 2% a year. Owner cash is negative in most years to 2036 and the unrestricted cash falls to about $11M by 2030 after interest and preferred dividends, so the company would need new money on poor terms. The model value is −$1.47, which I show as $0, because shares cannot be worth less than nothing.

Base

Base Electron is financed and built as signed, finishing in 2030 at a 4.5% Adjusted EBITDA margin. B&W wins 0.6 GW in 2028 and then 0.3 GW a year, 3.0 GW in all by 2036, at the same margin. The older business earns 9% on revenue that grows 4% a year from $720M, a little below what I read the 2026 target as implying. Owner cash is $6M in 2027, $42M in 2030 and $64M in 2036, and the preferred dividends and interest of about $28M a year come out of it.

Bull

B&W wins about 1.2 GW in each of 2027 and 2028, then 1 GW, 0.8 GW and 0.6 GW a year, 7.8 GW in all, at $2.2B a GW and a 6.5% margin. Its older business earns 10.5% and grows 5%. Owner cash reaches $143M in 2030 and $152M in 2036, net claims turn into net cash by 2030, and the company can fund the next project without new shares.

What the price says

Turn it round. At $4.27 the market values the shares at about $643M, and with my estimates of the warrants and the claims ahead of the shares, the whole business at $813M C (calculated). My base is worth $357M on the same measure E (estimate). Something has to fill a gap of $456M.

The simplest reading is that the market is paying for several more Base Electrons, about 13 of them, at the margin I assume. The other way to read it is that margins on this work are higher than I think. With my bull margins, a business that earns 6.5% on pass-through revenue and 10.5% on the older work, the price needs about 4.2 GW, close to what I have in the base. So the debate is mostly about margin, not volume. The markup is blacked out, which is why I can’t settle it. The company’s own pipeline of over $14B over three years is a count of opportunities, and at $2.0B a GW it is about six contracts the size of Base Electron, and not all of it is AI plants.⁷⁶

Where the base value comes from

My base is worth $1.41 a share. The older business alone, with no Base Electron and no new contracts, is worth $0.12. Base Electron on its present terms adds $0.51. The 3.0 GW of further contracts adds $0.78. In other words, nearly everything I can support depends on work that is not yet financed, and more than half depends on work that has not been won.

How much the answers move

Margin matters more than the discount rate. A point of margin on everything moves the base by about $0.71 a share. A point of discount rate moves it by about $0.20. Every 3 GW of further contracts adds roughly $0.75. The price is a bet on both, margin and volume, and on a customer base that has so far shown up once.

The margin of safety

I set it 35% under my base value, $0.92, because the pension gap, the preferred stock and the notes are paid before the shares, which is the case where I take the largest discount, and because one customer drives the business. My bear value is $0, so a bear-based rule would give nothing, and I do not want to pretend that a price near zero is a measured level. At $0.92 the market value of the shares would be about $138M, against $424M of debt, leases and preferred stock. The stock last closed below $0.92 on 30 July 2025, three months after the company exchanged its notes.¹ There is no margin of safety at $4.27.

The price against the range

The price is 203% above my base and 29% below my bull. At $4.27 my base cash flows return 6.9% a year and the bull 16.3%, and the bear loses the whole investment E (estimate). Weighting the cases 25%, 50% and 25% gives $2.21 E (estimate). The market may be right and I may be wrong. The bull case is not extreme: the company has gone from $424M of backlog to $2.57B in six months and has turbines for a further gigawatt. The stock fell 80% from May, and the pipeline is real in the sense that the company is in talks. What I do not know is what the work pays and whether a second customer like Base Electron exists, and I would rather say so than move the numbers until they meet the price.

Against the alternative

A dollar here could go elsewhere. A ten-year Treasury pays 4.94%.¹⁸ My base return from today’s price is −2.6% a year to year 10, because the price sits well above my value. The nearest comparison on this site is GE Vernova, which the 10-K names as a competitor, and whose price sits 34% above my base value, with a base return of 5.8% a year. Vertiv, another seller to the same power buildout, sits 89% above base with a base return of 3.7%. B&W has the weakest return of the three on my numbers, and it pays no common dividend.

Quality and price, separately

How good is the business? The older business is a modest one: a 160-year-old brand with a parts business of about $250M a year and thin margins, which was restructuring its notes in 2025 and has improved for three years. The new business is unproven: one contract, with a thinly financed customer, at a markup I can’t see. How good is the price? Above my base on every set of numbers I can defend, and it already assumes either margins or volume that the company has yet to show. This looks like an improving business at a price that already includes the improvement and a good deal more.

How the odds multiply

The case for the price needs several things to go right at once. My guesses are in the table. They are not independent, since a financed customer makes a second contract more likely, so the true chance of all five is higher than the product. Even so, five steps of 50% to 85% leave about 16% for all of them, and that gets only to my base value, not the price.

Where this could be in 5 and 10 years

Here is my model’s value at two later dates. Each is the value, at that date, of the owner cash that comes after it, less the claims ahead of the shares rolled forward, per today’s 150.5M shares and in that year’s dollars E (estimate). The warrants are valued as options at five years and treated as exercised at ten if they are in the money. The company pays no common dividend in my model, so owner cash paid out is nothing; what is not needed for interest and preferred dividends stays in the business and shows up in the value.

For the base to stand at five years, Base Electron must be financed by March 2027 and built on schedule, the older business must keep earning about 9%, and B&W must win about 0.6 GW more by 2028. At ten years it also needs roughly 0.3 GW a year of further wins, about 3.0 GW in all. The base value rises from $1.41 to $2.24 and $3.29 mostly because the early losses end and owner cash reaches about $42M by 2030, then because the pile of claims shrinks. The bear differs by a failed or delayed financing, thinner margins and no further wins. The bull differs by winning about 7.8 GW at higher margins. Three numbers drive it: margin on cost-plus work, the number of contracts after Base Electron, and the discount rate. The most likely way the base is wrong is that the next contract is smaller or later than I assume, or earns the lower margin of Base Electron, which would pull every year down.

These are estimates of value, not forecasts of the share price. The market price can sit above or below value for years.

VIIThe case for the price

What most people miss

Three things cut in B&W’s favour and rarely come up. First, the older business is growing too. Take Base Electron out and second-quarter revenue was $219M, up 58% from $139M a year earlier, and parts alone grew 4% in the half C (calculated).² Second, the balance sheet is cleaner than the share price suggests: the company counts cash above debt by about $143M at 30 June, the 6.50% notes that were its only 2026 maturity were called for redemption on 13 August, and B. Riley’s guarantee of the bank facility was cancelled in February.⁴¹³¹² Third, people who run the company have bought on the way down. Mr. Young bought 7,000 shares at $9.65 on 12 August, and director Philip Moeller 5,000 at $9.50 the next day.¹⁴ The sums are small, about $68k and $48k, and they are dated facts, not a forecast.

The bull case

In plain words: AI data centers need power faster than the grid or the gas-turbine makers can deliver, B&W’s boiler-and-steam-turbine plant is a proven design that can be built without a gas-turbine slot, and it has signed one $2.4B contract and secured turbines for another gigawatt.⁴⁷ In the August release Mr. Young said the first Base Electron project was ahead of expectations and on budget, that manufacturing of boilers and turbines was moving quickly, and that talks on other data-center projects were active.⁷ The company raised its 2026 Adjusted EBITDA target to a range of $80M to $105M, says its pipeline of identified opportunities is above $14B, and, citing a Bloomberg NEF report of July 2026, puts the market for AI data-center power it could address at about $190B for contracts signed in 2026 to 2028.⁷⁴ Insiders have been buyers at $7 to $15.¹⁴

On the holder side, one fact, with its date. Situational Awareness LP, the fund run by Leopold Aschenbrenner, held 2,027,451 shares of Babcock & Wilcox Enterprises common stock, valued at $28.6M, at 30 June 2026, according to its 13F-HR filed on 14 August.²¹ That is about 1.4% of the shares then outstanding C (calculated). A 13F shows long positions in US-listed stocks only, not shorts or holdings abroad, and says nothing about what a fund does next. I found no analyst price target or published investor model that I could open and cite; Nasdaq lists one analyst forecast for the quarter.¹

Against reality

Today B&W earns about $64M of Adjusted EBITDA a year and owes $424M of preferred stock, debt and leases, with a pension gap on top C (calculated).⁴² If Base Electron is financed and built as signed, the older business earns about 9% on a growing base and B&W wins about 3.0 GW more in ten years, then my base is possible: $1.41 a share. The price needs more. It needs about 15.5 GW of new contracts at my margins, or roughly 4.3 points more margin on everything, or my bull margins with about 4.2 GW C (calculated). That is the distance between today’s numbers and the case. The company has shown one contract and a margin it does not disclose.

What a 10x would take. Ten times $4.27 is $42.70, about $6.4B of share value. At 12 times Adjusted EBITDA, after about $134M of claims ahead of the shares, that needs about $547M a year, 4.8 times my base for 2030 and 2.1 times my bull for the same year C (calculated). My bull case is worth $6.01, about one-seventh of that. In my reading a company this size multiplying its share price tenfold in ten years is rare, and where it happens it has usually followed a collapse. This one has already done it once, rising 83 times from $0.26 in April 2025 to $21.85 in May 2026, and then fell 80%.¹ A second move of that size from a value near $0.6B is close to impossible on my numbers, and I would not build a case on it.

Where I come down

Three things moved me: the older business is growing, the balance sheet is cleaner than a year ago, and the people running it have been buying. They did not move my numbers to the price, because the price needs several more Base Electrons or a markup I can’t see, and because nothing I can open shows that the first one has closed its financing. Financial close on the North Dakota plant, a disclosed markup, a second named customer and positive cash flow after payables normalise would raise my values most. What happens by 31 March 2027 matters more to my numbers than any quarterly report before it.

VIIIRisks, and what would change my view

The three ways this goes badly wrong are at the start of Valuation. These are the rest.

• One customer. Base Electron was 31% of second-quarter revenue and 43% of trade receivables, and its contract is 88% of backlog.² Nothing filed shows its financing closed. Applied Digital’s 10-Q of 7 October lets Base Electron end the power agreement if financial close has not occurred by 31 March 2027, and its owners include Applied Digital officers and directors in their own names.¹¹
• Terms I can’t see. The markup, the caps on damages and the termination terms are blacked out in the filed contract.⁶ A suspension would be paid for, but how much and for how long is not public.
• Fixed-price scope and labour. $434.8M of the contract is a fixed fee with delay and performance damages, and the company says skilled labour is short.⁶²
• Working capital. Operating cash flow in the first half was held up by a $100.3M rise in payables.² If customers pay more slowly than suppliers, B&W’s cash goes first.
• Pension. The plans were $174.3M underfunded at the end of 2025, and a PBGC waiver raises payments over five years; a $3.0M installment fell due by 15 September 2026.³¹²
• Preferred stock and notes. $191.7M of preferred stock at 7.75% is cumulative and perpetual, and $129.5M of notes at 8.75% fall due in 2030.² Both rank ahead of the shares.
• Dilution. The share count rose from 95.1M to 149.0M in 19 months. Warrants over 10.46M shares at $4.11 are outstanding, B. Riley has the right to buy first in new offerings, and a $200M at-the-market programme sold 4.4M shares in the first half.⁵²
• Controls. Material weaknesses in three areas of internal control were reported at the end of 2025 and had not been fixed by 30 June.³²
• Related parties. B. Riley is the largest holder, the underwriter and the adviser, and Applied Digital is both a warrant holder and the sponsor of the customer.²¹¹
• Tax losses. The company found an ownership change in July 2019 that limits part of its tax losses, and a further change is possible when shares are issued.³
• Demand for coal and gas equipment. The older business depends on utilities and industrial customers spending on boilers and parts, and the 10-K lists policy and fuel-price swings as risks.³

Geopolitical risk, and what it means for the value

B&W is less exposed to foreign politics than most of its peers. In 2025, 71% of revenue came from customers in the United States and 87% from the United States and Canada, and it builds in the United States, Canada and Mexico.³ Its export-control and sanctions risk factor names Russia, and its own market estimate leaves out China, India, North Korea and Russia.³⁴ The exposure is a supplier and a customer.

Turbines. Siemens Energy, a German company, supplies the steam turbines for Base Electron and for a further 1 GW.⁴⁷ The filings do not say where the units are made, what B&W has committed to pay, or whether that purchase can be cancelled. Siemens Energy reported a total backlog of €162B at 30 June, so a supplier this full can move a delivery date.²³ The first plant is due in 2030. Tariffs. The 10-K says tariffs have raised its costs.³ US steel and aluminium duties were reset on 1 June 2026: 25% by default on covered products entered from 8 June 2026 to 31 December 2027, with a reduced category for some fixed power equipment.²² I could not confirm whether boilers or steam turbines are in it. The $2.0B variable part passes costs on with a markup; the $434.8M fixed fee does not.⁶ After 2027 the rate is undecided.

The money. The only evidence on Base Electron’s financing is in Applied Digital’s 10-Q of 7 October. It lets Base Electron end the power agreement if financial close has not occurred by 31 March 2027, and it mentions a credit support agreement without terms I could find.¹¹ I can’t know whether the money arrives. Abroad. International sales were 29% of 2025 revenue, with no country other than Canada above 3%, and the 10-K names Europe and the Middle East only as places where fuel prices and climate rules shape demand.³ B&W sells no nuclear fuel or reactors, so nuclear policy reaches it only as competition for the same power orders.

None of this is in the price. At $4.27 the price is already 203% above my base, so it is not pricing these downsides; if it prices a geopolitical story, it is the other one. If the gas-turbine queue stays full and B&W wins twice my 3.0 GW, the base value rises to $2.14 E (estimate). Public today: the tariff rate to 2027 and Siemens Energy’s backlog. Not public: Base Electron’s financing, the markup and B&W’s terms with Siemens.