Solaris Energy Infrastructure

IThe business

Solaris is two businesses under one roof, and one is now much bigger than the other. Solaris Power Solutions rents out and runs power plants, mostly gas turbines, for data centres, oil and gas producers and industrial sites that cannot wait for the grid.³ Solaris Logistics Solutions is the company it used to be: it designs and rents the equipment that stores and moves sand on frac sites, and charges for the crews and trucking around it.³ The head office is in Houston. The year is the calendar year.

The company listed in May 2017 as an oilfield-equipment business.³ The change came in September 2024, when it bought Mobile Energy Rentals for $323.1M and with it a fleet of mobile turbines and a long-standing relationship with a turbine supplier.¹² Power Solutions went from $39M of revenue in 2024 to $334M in 2025 and $495M over the twelve months to June 2026 C (calculated).³²⁴ In the second quarter it was 72% of revenue and 80% of segment Adjusted EBITDA.²

How it earns

A power customer pays a fixed monthly fee for a plant, not for the electricity. The customer buys the gas and carries the fuel and power price risk.⁷ Solaris earned $158.3M of revenue and $96.4M of Adjusted EBITDA from the segment in the second quarter, on an average of 950 MW earning revenue.⁴ That is about $667k of revenue and $406k of Adjusted EBITDA a megawatt a year C (calculated). Over the next three years it plans to build that fleet to more than 3,300 MW, at about $5B of total capital, which is roughly $1.5M a megawatt C (calculated).⁷ If the fleet earns what the second quarter earned, each dollar spent returns about 27 cents a year before tax C (calculated). That is the whole case in one line, and I come back to how far it can be trusted.

The contracts

Four contracts now carry the story. Three were disclosed in the September offering information; the fourth was signed on 29 October.⁷⁸ I reproduce the terms as the company states them.

Three of these are with affiliates of investment-grade technology companies, and that is the main reason the debt is priced as it is: the company issued $1.3B of 6.375% notes in May and $1.25B of 7.000% notes in October, and S&P, Moody’s and Fitch rate it BB-, Ba3 and BB, all below investment grade.²⁶⁴ The Hatchbo agreement shows what the contract protection is really worth. The customer may end it for convenience on 30 days’ notice, and pays half of the remaining rent.⁹ That is a real payment, and it is also an exit. Half of the rent is not the same thing as all of it, and it does not repay the debt that built the plant on the day it is paid.

Concentration is the other thing to see. In the second quarter one customer was 63% of revenue and 68% of receivables; in 2025 one customer was 88% of Power Solutions revenue.²³ The 10-Q names the Stateline partner as MZX Tech LLC and says no more about it, and it names no other customer.² I can’t see who the customers are, and I can’t judge their credit beyond the company’s own description.

How it got here

The growth has come by purchase and by stock. In March 2026 it paid for Genco Power Solutions, a gas-turbine rental business, at a cost of $483.2M that included 4,182,772 new shares.²¹³ In the same week it bought the contract rights to 30 Baker Hughes NovaLT16 turbines from another developer, paying $66.9M to that developer and $64.3M to Baker Hughes for overdue milestone invoices, with $364.9M still owed to Baker Hughes, plus up to $130.2M more to the developer as units arrive.² The turbines come between September 2026 and September 2029.² In July it bought Global Energy Services Alliance (GESA), a power service firm, for about $55M and 2,880,682 shares.¹⁴ On 1 September it bought Omega Foundation Services, a civil and electrical construction firm, for 3,599,199 shares and about $77M in cash.¹⁵ The company said in September that it had 2,100 employees, up from 468 at the end of 2025.⁷³

That is a very different business from the one that finished 2024. It was a small equipment lessor. It is now a developer, builder and operator of power plants, a financer of them, and still a seller of sand logistics. Each step is reasonable alone. Together they are a lot to run at once, and the company says so in its own risk factors: its labour-intensive construction and service lines carry fixed-price and schedule risk it has not had before.⁷

The old business

Logistics Solutions earned $88M of Adjusted EBITDA over the twelve months to June, about a quarter of the two segments’ total C (calculated).⁵⁴ It tracks drilling and completion activity, and when drilling stops it stops: group revenue fell 57% in 2020 C (calculated).¹⁶ Its customers are oil and gas producers and service companies, and in the second quarter its revenue fell 17% on the year as last-mile trucking declined, even though the number of fully used systems rose from 94 to 106.² The company expects it to be under 10% of segment earnings once the new fleet is running.⁷ It still matters: it earns about a quarter of segment Adjusted EBITDA, which helps pay for the rest of the build and the dividend of about $30M a year C (calculated).

IIThe moat

What keeps a rival from earning what Solaris earns? I look in three places, and the honest result is that the first is real, the second lasts only as long as the contracts and the third is thin.

Scarcity of turbines. The binding limit on behind-the-meter power today is how fast anyone can get gas turbines, switchgear and people to put them together. Solaris has firm orders to reach more than 3,300 MW by the end of 2029, including 30 NovaLT16 delivery slots it bought from another developer.⁷² A slot is only worth something while turbines are scarce. It does not stop the makers selling to others, and Solaris does not make the turbines. The slot purchase also shows the other side: the contract had overdue milestone invoices and suspension costs when Solaris took it, and Solaris paid Baker Hughes $64.3M to clear them.²

Switching costs once a plant is running. A customer with a power plant on its site, run and maintained by Solaris, cannot move it easily. A Hatchbo or Customer C customer that leaves pays 50% of the rent left, and a Stateline customer pays 100% in the first three years and 50% after; the October terms are not disclosed.⁷⁸ The 10-K lists a risk that Solaris may struggle to find another lessor for the Stateline turbines if the rental ends early.³ The switching cost protects the rent for a period. It does not set the price when the contract ends.

Intellectual property. Logistics holds eleven US patents that expire between 2032 and 2044.³ The patent that a rival, Masaba, sued over was held unpatentable by the patent office in January and the time to appeal passed on 14 July.² That removes a risk. It also reminds me how little of the company rests on patents. The power fleet rests on none.

What would show a moat is a price that holds when the first contracts end. Nothing in the filings tests that yet; the oldest big power contract began in 2026. The competitive set is also hard to see. The 10-K says the main competitor is the electricity grid, that most distributed-power companies are private, and that only a few have the same large-scale capability.³ Big technology customers could also build their own plants, and I can’t rule out that more of them will.

IIIThe competitive advantage

The advantage Solaris claims is time and completeness. A customer who needs power in a year can sign with one company that supplies the turbines, the electrical equipment, the batteries, the site work, the installation and the operation.⁷ Three purchases built that: HVMVLV in August 2025 for electrical control and distribution, GESA in July 2026 for installation and long-term service, and Omega in September 2026 for site and substation construction.³¹⁴¹⁵ GESA also brought operation and maintenance contracts on more than 1 GW of installed plant.⁷

The second advantage is cost and access to capital. It has gone to the debt markets four times since May 2025: $155M of convertible notes, $747.5M of convertible notes, $1.3B of senior notes and, in October, $1.25B of senior notes, which were upsized from $1.0B.²⁶⁷ That kind of access is worth something to a customer who needs a counterparty that will finish. It is also a bill: annual interest on the debt in place after October is about $215M before any is capitalised, against $329M of Adjusted EBITDA over the past twelve months C (calculated).²⁶⁴

The third is the contracts themselves. Management says the capital it puts to work earns returns that stay as scope grows: more equipment per project, same return profile.⁷ I can check that only in part. The second-quarter segment margin was 61% on Adjusted EBITDA, and cost of revenue was 36% of Power Solutions revenue against 37% a year earlier C (calculated).² But Stateline and the others have not started paying, so the best measure so far is a fleet of older contracts, with a single customer.

One more advantage is who runs it. Zartler led the water business Aris until Western Midstream bought it in October 2025, and Brock and the finance chief came from it.³¹⁷ They know how to raise money and build infrastructure on a schedule. That is useful, and it is a reason this page does not treat the company as a hopeful newcomer. It is also a team that has not yet run a power plant fleet of this size.

IVManagement

Bill Zartler, 60, is chairman and co-chief executive. He ran the predecessor company from 2014, has been chief executive since July 2018 and, until its sale to Western Midstream in October 2025, was also chairman of the water business Aris.³ Amanda Brock, 65, joined as co-chief executive in October 2025 from the top job at Aris, and the chief financial officer, Stephan Tompsett, 49, joined in February 2026 from the same company.³¹⁷ The finance chief before him, Kyle Ramachandran, is now president.¹⁰ This is a team that has worked together, and it was assembled in four months while the company was signing the largest contracts in its history.

How they are paid

Zartler took $13.9M for 2025, of which $12.6M was stock awards and $0.45M salary; Brock took $7.0M for three months, nearly all stock.¹⁰ The stock awards include the extra expense of changing earlier performance awards so they can be paid in cash.¹⁰ The performance awards pay out on total shareholder return, from 0 to 200% of target; the grants made in the second quarter of 2026 are weighted 90% to absolute return and 10% to return against a peer group.¹⁰² For 2025 the awards vested at 200% on both measures because the share price had risen so far.¹⁰

That design pays people to make the share price go up and to do it quickly. It does not pay for return on the capital spent, for cash left after the turbines are bought, or for the share count. A manager paid this way has a reason to build and to buy, and to buy with shares, because each acquired megawatt adds to the story before it adds to cash. I do not say that is what they are doing. I say it is what the pay rewards, and that it is worth watching next to the numbers below. The company does bar hedging and pledging of its stock, and has a clawback for restated results.¹⁰

What they own, and what holders sold

Zartler owns 1,314,883 Class A shares and 4,240,315 Class B units, together 5.56M, worth about $393M at $70.70 C (calculated).¹⁰ That is about 7% of the economics and 7.6% of the votes.¹⁰ He has made no open-market sale since October 2025.¹¹ Brock bought 3,669 shares at $40.89 in December, a small sum.¹¹ The offering information says management retains about 20% ownership; I can’t reproduce that from the proxy, where the largest holders are Yorktown, a fund whose managers include director Howard Keenan, Zartler, and KTR Management, owned by John Tuma.⁷¹⁰

The largest holders did sell. KTR sold 2,114,783 shares at $50.15 on 4 November 2025 and 2,000,000 at $70.75 on 30 April 2026. A fund tied to director Howard Keenan sold 2,000,000 at $74.50 on 6 May. Together that is about $397M C (calculated).¹¹ Smaller sales by the chief administrative officer (47,011 at $44.62), the chief legal officer (25,492 at $74.84) and one director (56,841 at $72.11) look routine. A different director, A. J. Teague, bought 3,900 shares at $64.21 on 8 September.¹¹ I treat the large sales as pre-IPO holders reducing a position, not as a signal, but they are the shares that supplied the market while the company was also issuing its own.

What they did with capital

Over nine years since the listing, owner cash flow has been negative in six years and the total is −$621M C (calculated).¹⁶ The 2017 listing raised $44.7M of new equity; in 2024 a share sale raised $160.9M.¹⁶ Between 2017 and 2025 the company paid regular dividends and, under the 2023 programme, bought back 4,272,127 shares at an average of $8.09, a good result in hindsight.² It is not a result that carries forward: the share count has since gone the other way. Class A and B shares together were 44.1M at the end of 2023, 67.1M at the end of 2024, 68.5M at the end of 2025 and about 80.2M after the Omega deal C (calculated).¹²³²¹⁵ That is 82% more in under three years. Three purchases in 2026 alone added 10.7M shares: Genco 4.18M, GESA 2.88M and Omega 3.60M.²¹⁴¹⁵ The price paid per share in the GESA deal was $73.42, which is the price in June when the shares were near their high C (calculated).²

The tax receivable agreement is a smaller, older item. It entitles the pre-listing owners to 85% of certain tax savings, and the liability was $109.2M at 30 June.² Zartler’s own company, Solaris Energy Management, also rents office space to Solaris for about $0.5M for the first half of 2026.² The sums are small. They are the kind of thing I would want to see stay small as the company grows.

VFinancials

Solaris listed in May 2017, so its filings give nine full years and the twelve months to June 2026. That is the history that exists. For most of it the company was a profitable oilfield equipment lessor: revenue reached $242M in 2019 with a 45% operating margin, fell to $103M in 2020, and took until 2022 to pass $300M C (calculated).¹⁶ Then it changed.

Revenue has grown from $313M in 2024 to $762M over the past twelve months, 2.4 times, almost all of it from power. Operating income has risen from $53M to $185M. The margin, 24% over the twelve months, is lower than in 2018 and 2019 because the new assets are heavily depreciated and the company has hired at speed: selling and administrative cost was $30.0M in the second quarter against $14.9M a year earlier.⁴ I would not read the margin as a ceiling or a floor yet. Most of the 3,300 MW is not earning.

What an owner could really take out

Nothing yet. Over the twelve months to June, net income before the minority share was $79M. Operating cash flow was $425M, capital spending $1,153M and stock pay $30M, so owner cash flow was −$758M C (calculated).¹⁶² That is $837M worse than reported net income. The gap is the build. Operating cash flow itself is flattered: in the first half of 2026 it included $181M from deferred revenue, which is customer money received in advance.² Without it the owner cash outflow for the twelve months would be nearer $940M C (calculated). Deferred revenue was $192M at 30 June against $6M at the end of 2025.²

Over nine years, owner cash flow has been negative in six and the total is −$621M C (calculated). I would expect that to turn once the fleet is built, because maintenance spending on a running turbine is much smaller than the cost of buying it. I can’t verify how small. The company calls it “relatively low”, and I assume $35,000 a megawatt a year in my base, which is about 8% of the Adjusted EBITDA a megawatt earns E (estimate).⁷

Quality of earnings

Adjusted EBITDA excludes stock pay, which was $29.7M over the twelve months, and is added back before comparing with debt C (calculated).⁴ The company’s adjusted pro forma earnings per share was $0.39 in the second quarter against a GAAP diluted $0.26.⁴ The gap is the loss on repaying earlier loans, $14.8M, and the swap of Class B units into shares. Interest is changing the picture. In the first half, $17.4M of $39.1M of interest cost was capitalised into the cost of turbines under construction, which makes today’s profit look better than it will once the plants run and the interest goes through the income statement.² In the second quarter alone interest cost was $28.7M against interest expense of $16.9M.² The new October notes add about $87.5M a year.⁶

The customer side matters too. Receivables doubled in six months to $210M, with one customer 68% of the total.² That fits fast growth and big projects. It also means an owner depends on how one buyer pays.

The balance sheet

At 30 June, consolidated debt and convertible notes were $2.49B on the balance sheet and cash $0.82B, with another $0.07B restricted.² Since then the company has raised $1.25B more, so debt and convertible notes at face are about $3.8B against $0.3B at the end of 2024.⁶¹² Against that, equipment held for lease was $2.49B net, including $1.14B of turbines and other items still under construction.² The next table lists what is owed.

Three things about these debts. First, the covenants. The revolver limits net debt to 5.25 times EBITDA (5.5 times for four quarters after a large acquisition), secured net debt to 3.5 times, and requires EBITDA of at least 3 times cash interest, from the quarter ending 30 September.² Net debt counts at most $50M of cash, so cash raised ahead of spending does not help the test.¹⁸ EBITDA, in the agreement, can be increased for contracts that are not yet earning, by the share of their budgeted capital already spent, and the credit falls by 25% for each 90 days a projected increase is late.¹⁸ So the test is workable on schedule and harder if the build slips. Second, the interest. Cash interest on the notes and the two convertibles is about $180M a year at the October level, plus about $35M on Stateline C (calculated).²⁶ Against $329M of twelve-month Adjusted EBITDA, that is thin; against the $433M annualised second quarter, it is 2 times. Third, the dates. Four maturities fall between 2030 and 2032 and total $3.45B; the plan has to turn the fleet into cash or refinance it by then.

Liquidity at 30 June was $824M of cash and $575M of unused revolver, to which October added $1.23B. Against that stand $1.48B of turbine commitments and about $1.0B of remaining 2026 spending after June.² The company says its sources cover the next twelve months and beyond.² I take it at its word for now, and I note that my base case needs about $1.1B more in the rest of 2026 and $2.5B in 2027 to 2029 E (estimate).

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 build is financed and then the demand changes. About $3.6B is still to be spent between now and 2029, on top of $3.8B of debt and convertible notes already raised E (estimate).⁷⁶² Most of it is on contracts with large customers, which is a good reason to spend. But Hatchbo may end its contract on 30 days’ notice for half the remaining rent, and Customer C’s contract has the same 50%.⁹⁷ If AI demand slows or a customer decides to build elsewhere, Solaris would be paid a fraction of the rent while it still owes the full cost of the plant. The company has said it can cancel purchase orders, at a cost of 5% to 90% of the price, and $365M to Baker Hughes cannot be cancelled.²

The first contracts end and the price falls. The business earns about $406k of Adjusted EBITDA a megawatt on cost of about $1.5M. Returns like that attract rivals, and the turbines, which anyone can buy, are the same. After the first terms, the fleet may re-let at much lower prices. In my base the price falls to 75% of today’s; in the bear, 70%. The largest single number I can’t know is what a 2037 megawatt of behind-the-meter power will rent for.

The price. The business can work and the investment still fail. The shares are $70.70, 15% below the high of June, and 18 times the 2020 low.¹ A buyer pays for the whole plan to be built, and for more of it, and takes all the debt ahead. Of the three, I think a poor result from a good business at a high price is the most likely.

So the questions for everything below are: how much of the 3,300 MW is truly built and paid for as signed, what each megawatt earns after the first terms, and how much of all that is already in the price.

What an owner gets today

Owner cash over the past year was −$758M, against net income before the minority share of $79M C (calculated).² In my base the fleet turns that around in 2028, and owner cash reaches about $1.0B a year by 2030, before interest, against Adjusted EBITDA of $1.3B after the Stateline partner’s share E (estimate). About 77% of that reaches the owner. The rest goes to stock pay of about $45M a year, maintenance of about $115M, and cash tax that begins at about $30M and rises past $120M E (estimate). Interest on $3.8B of debt, about $215M a year, comes out of that before shareholders see anything, and the company pays $0.48 a share in dividends.²⁶

What I can’t judge

Four inputs matter most and can’t be known from outside. What the October 900 MW and every later contract pay, since no price is disclosed.⁸ What the megawatts earn after the first terms. Who the customers are and how strong their credit is, since the filings do not name them beyond “investment-grade”. And what it costs to build a plant at this scale, because the company has little history of doing so: its fleet has grown from about 230 MW in 2024 to 950 MW in the second quarter.³⁴ For the first two I show a range below rather than a false precise number, and for the last I add 5% to cost in the bear.

How I value it

Owner cash for each year from the second half of 2026 to 2045, then a tail that assumes the fleet runs for another 10, 15 or 20 years in the bear, base and bull while its cash falls 4%, 2.5% or 1% a year. Owner cash is Adjusted EBITDA less stock pay, maintenance capital, growth capital, cash tax and a small working-capital charge, before interest. Net debt is then subtracted. The fleet is built from megawatts earning revenue and Adjusted EBITDA a megawatt, which I take from the second-quarter segment figures ($406k a megawatt a year) and from management’s own illustration of $820M on about 1,850 net megawatts ($443k); I use $400k, $420k and $450k.⁴⁷ Growth capital follows the company’s plan: about $1.1B for the rest of 2026, then $1.35B, $0.8B and $0.35B, which is my estimate and includes the October contract’s extra equipment.⁷ Stateline’s partner takes 49.9% of its earnings, which I deduct, and I credit 49.9% of its debt.² The 2030 convertible converts into shares when my value is above $26.39; the 2031 convertible stays as debt, with the capped call covering dilution up to $88.00.² Values are per share on 86.9M shares.

The discount rate. The ten-year Treasury yield was 4.94% at the price date and Damodaran’s implied equity risk premium 4.20%.¹⁹²⁰ Solaris has traded for only part of the period I would like, and its beta on month-end returns against the S&P 500 since March 2020 is 1.51, 1.40 over five years and 1.69 over three, which gives a cost of equity of 10.8% to 12.0% C (calculated).¹²¹ The notes cost 6.4% and 7.0% before tax, so at market weights the blended cost is about 9.0% C (calculated). I use 10.5% on the cash before interest, a point and a half above that blend, because beta does not capture the weight of a few customers and a construction programme.

Bear

The three big plants are built, but a year late and 5% over cost, so 2,900 MW is earning by 2029 and 3,100 MW by 2031. Each earns $400k a year rather than the $420k I use in the base. The fleet re-lets after the first terms at 70% of today’s price. Services and construction earn $80M. On these numbers net debt reaches 6.3 times EBITDA at the end of 2027, above the 5.25 times the revolver allows, so the company would need a waiver or new equity. The $5 includes neither. The covenant counts at most $50M of cash, which would push the ratio above my 6.3 times, and credit for contracts not yet earning would pull it down. I show the equity route below.

Base

Every contract is built as signed, including the October 900 MW at the same economics. The fleet reaches 3,300 MW earning revenue in 2030 at $420k a megawatt, a little under the $443k management’s own build implies. Services earn $125M by 2029 and sand logistics $85M. After the first terms end, from 2035, the price drops to 75% of today’s by 2040. Owner cash is first positive in 2028 and about $1.0B in 2030.

Bull

The company keeps winning. The fleet doubles again to 6,000 MW by 2033 at $450k a megawatt and a cost of about $1.5M each, maintenance is lower, and re-letting holds 95% of today’s price. Owner cash reaches $2.1B in 2033.

If the bear needs new equity

To bring bear net debt back to 5.25 times its 2027 EBITDA of $648M, the company would have to take about $0.7B off net debt E (estimate). Shares sold above their value do the holders no harm. Shares sold below it do, and a company in breach is likely to find itself there. The table raises $0.7B at three prices, after 5% costs, on the bear’s 81.1M shares. A lender could take a fee and a higher rate instead, and I cannot see which. At $3 the new holders would own three quarters of the company.

What the price says

Turn it round. At $70.70 the market values the company at about $6.1B for the shares and $7.8B with my estimate of net debt C (calculated). My base is worth $4.5B on the same measure, after taking out the $3.6B still to be spent E (estimate). Something has to fill a gap of $3.4B.

The simplest reading is that the price expects a company about three times the one in the plan: some 9,418 MW rather than 3,300 MW, at my base returns C (calculated). Or the same fleet earning 47% more on each megawatt. Management’s own illustration of $1,000M of Adjusted EBITDA by 2029 is below my base of $1.3B for that year, both net of the Stateline partner, because mine includes the October contract and the megawatts not yet under long-term contract at the time.⁷ I am not being harsh on the fleet. I am harsh on the price paid for it, and on how long the payback is.

How much the answers move

The surprise is how much one assumption matters and how little the discount rate does. A 10% change in what each megawatt earns moves the base by about $8 a share. A one-point change in the discount rate moves it by about $6. By contrast, every 1,500 megawatts of extra growth at the base returns adds about $9 or $10. The price is not a bet on a rate or on a margin. It is a bet that the 3,300 MW is the first stage of something much larger, and that every later stage earns what the first does.

The margin of safety

I set it from the bear case: the price at which my $5 bear value is three quarters of what is paid, so $6. At $6 a share the market value would be about $0.5B, against net debt of $1.7B and $3.6B still to spend E (estimate). That tells you how thin the equity is in a bad case. The stock last closed at or below that price in November 2020.¹

The price against the range

The price is 122% above my base and 19% below my bull. At $70.70 my base cash flows return 5.9% a year and the bull 11.7%, and the bear 2.1% E (estimate). Weighting the cases 25%, 50% and 25% gives $39 E (estimate). The market may be right and I may be wrong. The bull case is not an extreme one: the company grew from about 230 MW in 2024 to 950 MW in the second quarter, and has orders that take it above 3,300 MW. What I do not know is how long it can keep finding customers who pay the same price, and I would rather say so than move the numbers until they meet the price.

Against the alternative

Every dollar here could go elsewhere. A ten-year Treasury pays 4.94%.¹⁹ My base return here is 5.9%. The nearest comparison on this site is Constellation Energy, which owns long-lived power plants that data centres also want, whose price sits 20% above my base value and whose base return is 7.7% a year. GE Vernova, which makes the turbines Solaris rents, sits 34% above my base and returns 5.8%. Solaris has the highest price against base of the three, and less history of paying owners. It does not have the Constellation balance sheet behind it.

Quality and price, separately

How good is the business? The idea is good: offer time and certainty to a customer who is short of both, on long contracts, with a high return on each plant. The proof is thin: only 950 of the 3,300 megawatts are earning, and the company has never run a fleet this size or a construction business at all. How good is the price? Above my base on every set of numbers I can defend, with a debt pile the equity must carry first. This looks like a promising business at a price that already assumes the promise is kept, and then some.

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 customers who pay on time also make it easier to borrow, so the true chance of all five is higher than the product. Even so, five 70% to 85% steps leave about 25% 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 own model’s value at two later dates. Each is the value, at that date, of the owner cash that comes after it, less the net debt I roll forward, per today’s 86.9M shares and in that year’s dollars E (estimate). Owner cash paid out is the dividend I assume stays at $0.48 a share; the rest pays down debt first.

For the base to hold at five years, three things must happen: the three big plants and the October contract earn from 2027 on as signed, the 3,300 MW is earning by 2030, and net debt peaks near 3.8 times EBITDA in 2027 and falls. At ten years it also needs the fleet to re-let at 75% of today’s price. The base value rises from $32 to $54 and $89 mostly because debt gets paid off, not because earnings grow. The bear differs by the price per megawatt, a year of delay and 5% extra cost. The bull differs by adding another 2,700 MW at the same returns. Three numbers drive it: the megawatts earning, what each earns, and the cost of the capital still to spend. The most likely way the base is wrong is that the October contract and later ones are signed at lower prices than the first three, 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 Solaris’s favour and rarely come up. First, customers are paying in advance. Deferred revenue was $192M at 30 June against $6M at the end of 2025, and $181M of it arrived in the first half.² People about to walk away do not usually prepay. Second, the Hatchbo contract has grown twice before a plant has started paying: from over 500 MW to about 530 MW, then to about 660 MW, and from up to 15 years to up to 18.⁹² Third, the lenders have been willing. The October notes were raised from $1.0B to $1.25B at 7.000%, and the revolver was enlarged to $850M the same day.⁶⁷

The bull case

In plain words: the grid cannot connect data centres fast enough, gas turbines are the quickest way to give them power, and Solaris is one of the few companies that can supply, build and run a whole plant. Zartler said in August that Solaris is building a proven power and infrastructure business, and Brock that the company continues to see “strong market demand for our services”.⁴ The company’s own illustration, in its September offering information, is $1,000M of annual Adjusted EBITDA by 2029 from the contracts then signed, net to Solaris, with net debt falling toward 3 times EBITDA, and the October 900 MW is on top.⁷⁸

On the holder side, one fact, with its date. Situational Awareness LP, the fund run by Leopold Aschenbrenner, held 1,129,621 Class A shares, worth $90.9M, at 30 June 2026, according to its 13F-HR filed on 14 August.²² That is about 1.5% of the 76.6M Class A and B shares outstanding on 3 August 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 named sell-side target or published investor model for Solaris that I could open, so I cite none.

Against reality

Today the fleet earns on 950 MW, and the second quarter produced $108M of Adjusted EBITDA; the third-quarter guide is lower, $90M to $105M.⁴ If the three big plants and the October contract are built within budget, and earn what the second quarter earned per megawatt, my base is possible: $32 a share. The price needs more. It needs about 6,118 MW beyond the 3,300 MW planned, earned at the same rate, or 47% more on every megawatt C (calculated). That is the distance between today’s numbers and the case.

What a 10x would take. Ten times $70.70 is $707, about $61B of share value. At 20 times owner cash that needs about $3.1B a year, three times my base for 2033; at 12 times Adjusted EBITDA, after $1.7B of net debt, it needs about $5.3B a year, nearly 4 times my base for 2030 C (calculated). By 2036 that is a fleet of roughly 13,000 MW at today’s returns, four times the plan C (calculated). This stock rose 18 times from March 2020 to June 2026, when the company changed what it was.¹ A second change of that size from a value near $6B is close to impossible on my numbers, and I would not build a case on it.

Where I come down

Two things moved me: customers paying in advance and enlarging contracts, and lenders upsizing the notes. They did not move my numbers to the price, because the price needs a much bigger company, and because a contract that can be ended on 30 days’ notice for half the rent is a strong claim only while the customer is happy. A price disclosed for the October 900 MW, the first Stateline lease starting, and a second ten-year contract from a customer who is not one of these four would raise my values most, and I would also want the share count to stop rising.

VIIIRisks, and what would change my view

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

• A few customers. One customer was 63% of second-quarter revenue and 68% of receivables, and three contracts account for most of the planned fleet.²⁷ The filings do not name them. A change of mind, a credit event or a new plan by any one of them would show up quickly.
• Termination for convenience. The Hatchbo contract can end on 30 days’ notice for half the remaining rent.⁹ The Stateline venture has stronger terms in its first three years.⁷
• The credit agreement. The covenants are net leverage of 5.25 times, secured leverage of 3.5 times and interest cover of 3.0 times, and cash counts for at most $50M.²¹⁸ A delay lowers the credit allowed for contracts not yet earning by 25% each 90 days.¹⁸
• Construction and labour. The company now employs more than 2,100 people and does fixed-price and milestone-based work it had little of before.⁷ Cost overruns or accidents would hurt margins and customer trust.
• One turbine supplier. A limited number of key suppliers provide the equipment, and $364.9M is owed to Baker Hughes on a contract Solaris can end only if Baker Hughes defaults.³²
• Rules and permits. Customers’ use of the equipment is subject to air-quality rules that may require permits and emission standards.³
• Gas. For Customer C, Solaris will also manage natural-gas supply on a cost-plus basis.² That adds a counterparty exposure the earlier contracts did not have.
• Dilution. The share count rose from 44.1M to about 80.2M in under three years, and the company says it may issue more securities.²
• The old business. Sand logistics earns about a quarter of segment EBITDA and follows drilling.²
• Key people. The founder and the co-CEO lead a company whose revenue more than doubled in 2025 and which now has 2,100 staff; the 10-K lists reliance on a few key employees.³⁷

Geopolitical risk, and what it means for the value

Almost everything Solaris earns is American: US data-centre and oil and gas customers on US sites. I found no foreign-currency, sanctions or foreign-ownership risk factor in the 10-K.³ The exposure comes from three outside forces: where the turbines are made, how US power policy moves, and the oil price.

Turbines. Baker Hughes builds its NovaLT turbines in Florence, Italy, and is adding capacity there.²³ The 10-K warns that Power Solutions equipment may carry tariffs Solaris cannot pass on.³ Washington sets the rate, the filing says legal challenges are open, and I cannot give a rate. Much of the exposure is fixed: $1.48B of orders are priced and $365M to Baker Hughes cannot be cancelled.² What is open is later orders and delivery dates.

US power policy. Customers rent from Solaris because the grid is slow. Faster grid connection would cut demand and rents, and the 10-K lists more grid capacity as a risk.³ Tighter air permits for on-site gas plants would slow builds.³ The October 900 MW is with an existing investment-grade customer at an undisclosed price, so its return is a guess.⁸

Oil. Logistics follows US drilling and completion.³ Group revenue fell 57% in 2020.¹⁶ The 10-Q names Russia and Ukraine, the Middle East, Venezuela and a US and Iran conflict as things that could move oil markets.² The 10-K names the Permian Basin only for disposal-well earthquake rules, so I cannot say how much rides on it. Logistics is about a quarter of segment profit now and under 10% once the fleet runs.⁷

None of this is in the price as a risk. At $70.70 the market already assumes a fleet about three times the plan, so each row widens the gap. My base has no tariff cost, no delay and no oil downturn.