NSO Capital NSO Capital Consulting Ideas All ideas
02 Power & Grid Active

Solaris Energy Infrastructure

NYSE: SEI

A power plant on wheels, rented by the month. The fastest route to energising an AI data centre while the grid queue runs five to eight years. The business is real and contracted, my work says the shares are worth meaningfully more, and the debt behind the ramp is the risk that governs the entry.

Noel Misthopoulos
Access full research paper →
First published26 Sep 2026
Price at publication$70.97, close 25 Sep
Time horizon12 to 24 months
StanceConstructive, accumulate on weakness
Data as ofQ2 2026 results, 5 Aug 2026
In plain terms

An AI training campus needs the electricity of a mid-sized city, and it needs it now. Connecting a load that size to the grid takes years. Solaris brings the power plant to the site instead: it owns gas turbines on trailers, installs them behind the meter, and rents megawatts by the month on multi-year contracts. The customer supplies the gas and carries no fuel risk. Speed is the product.

$219.4M
Q2 2026 revenue, up 47% year on year, with adjusted EBITDA of $108.3M
2.2 GW
Contracted with three investment-grade counterparties on 7 to 18 year terms
$800–960M
Annualised EBITDA implied by the Q1 2027 guide, against $244M earned in all of 2025
~5.6x
Net-debt-to-EBITDA at its projected 2026 peak: the risk, and the reason I wait

Source: Solaris Energy Infrastructure Q2 2026 results and 8-K (5 Aug 2026); Q2 2026 Form 10-Q; guidance raise and Q1 2027 guide (8 Sep 2026); convertible and senior-note filings (May 2025, Sep 2026).

01

Why I chose it

1. It sells the one thing AI cannot wait for: power, now

Connecting several hundred megawatts to the grid in ERCOT or PJM takes years; large-load queues are measured in five to eight. Solaris installs turbines behind the meter and energises a site in months. That speed, plus turbine lead times of 18 to 36 months that keep new supply scarce, is the moat. A hyperscaler that has already bought billions in GPUs cannot leave them idle waiting for a grid connection.

In plain termsThe chips are ready but the electricity is not. Solaris is the fastest way to plug them in.

2. The demand is contracted, not hoped for

About 2.2 GW is under long-term contract with three investment-grade technology counterparties on 7 to 18 year terms. Guidance has been raised in every quarter for six consecutive quarters; the Q1 2027 adjusted-EBITDA guide of $200 to 240 million implies an $800 to 960 million annualised run-rate, against $244 million earned in all of 2025. Contracts are paid per megawatt per month whether or not the turbines run.

In plain termsThe profit expected next year is roughly four times last year's, and most of it is already signed under contract.

3. A cash-cow legacy funds the growth engine

The Power Solutions segment was 72% of Q2 2026 revenue and 89% of segment EBITDA and is the growth engine. The legacy Solaris Logistics business, last-mile frac-sand rental, is cyclical and shrinking but still generates more than $20 million of free cash flow a quarter that helps fund the power build. I treat logistics as a cash source, not a valuation driver.

In plain termsAn older, steady business throws off cash that helps pay for the new one.

4. The capital structure is the differentiated angle

This is the section most coverage skips and the one my lens is built on. Gross debt reaches about $3.8 billion once the $1.25 billion 7.000% notes due 2032 close on 1 October, with a roughly $2.2 billion maturity tower already built for 2031. Net leverage peaks near 5.6x at year-end 2026 and falls toward 3x only when the 2027 run-rate arrives. The equity's fate is tied to a dated demand ramp by a debt structure that cannot wait. That is precisely the setup my research is designed to read.

In plain termsThe company borrowed heavily to build ahead of demand. If the demand lands on schedule the debt shrinks quickly; if it slips, the debt becomes the story.

02

What would break it

Each signal below is observable in public data. This is a leveraged business tied to a dated ramp, so a single missed milestone matters more here than for a debt-free name.

Signal What I would see Where it shows first
Energisation slips A key site misses its dated start. Each two-quarter slip is worth roughly $150 million of EBITDA and adds to net debt. The demand ramp is what de-levers the balance sheet, so timing is the whole thesis. Q3 print language on "energised"; JV lease note in the 10-Q
A customer builds its own power The largest counterparty is an xAI affiliate, and xAI has already bought its own 1.1 GW mobile turbine fleet. Contracted megawatts are safe for the term; renewals and new megawatts are not. Any xAI-affiliate award to a third party; JV distributions
Leverage covenant pressure Interest cover is about 2.1x this year's EBITDA against a 3.0x covenant tested on trailing figures. If the ramp slips two quarters, trailing cover approaches the covenant. Severe if triggered. Secondary bond prices; rating outlooks; any RCF amendment
The credit market disagrees The 2031 notes trading below par while the equity re-rates would mean the two markets disagree, and the credit market is usually the one to trust. Secondary levels on the 6.375% 2031s and 7.000% 2032s
Permitting / litigation A Clean Air Act suit and a state removal order affect temporary turbines at the Southaven site; the permanent 1.2 GW plant there holds an uncontested permit, but headline risk is live. Mississippi court rulings; MDEQ compliance filings
The re-rating never comes My upside depends on the multiple moving from the guide-year level toward the contracted run-rate as the 2027 dates print. If the market keeps valuing the guide year rather than the run-rate, the shares can stall even with the business on plan. Multiple on forward EBITDA versus peers

Known noise, not a thesis breaker: short interest is about 25% of the float and realised volatility is high, with three single-day moves above 10% in eight weeks. Part of the short position is convertible hedging rather than a directional bet. Expect the price to move violently in both directions around each print.

03

How I value it

Bear $75 Probability 25%

A key site and the JV lease slip two quarters and ~800 MW of ordered capacity stays unsigned, so the run-rate arrives late and the multiple stays where it is. 2027 EBITDA near $900M at ~10.5x.

Base · reference value $100 Probability 50%

The ramp lands on the announced calendar and the open capacity signs at framework terms, so the market prices the contracted run-rate rather than the guide year. 2027 EBITDA near $1,000M re-rating toward ~11.5x.

Bull $125 Probability 25%

Open capacity signs early at rising rates, the services roll-up proves accretive and a fourth hyperscaler signs, re-rating the shares to independent-power multiples. 2027 EBITDA near $1,100M at ~12.3x.

Method. Enterprise value on 2027 estimated EBITDA and year-end 2027 net debt, on ~86 million economic shares. Probability-weighted value about $100, roughly 40% above the price at publication.
Horizon. Twelve months from the date of first publication.
Sensitivity. Leverage amplifies everything: each 1.0x of multiple is about $11 per share, each $100M of 2027 EBITDA about $11, each $500M of net debt about $6.

Where the value comes from. The base case lands close to the analyst mean of about $97, but I reach it conservatively: I credit the ~800 MW of open capacity only in the bull case and value the services roll-up at zero until it reports as a segment. The whole return therefore rests on the market re-rating the shares from the guide-year multiple to the contracted run-rate multiple as the 2027 energisations print. If those dates slip, the re-rating slips with them. That is why my discipline sits in the entry and the checkpoints below, not in the target.

04

Entry discipline and dated checkpoints

The base value sits well above the current price, but this is a leveraged name tied to a dated ramp and it moves violently around each print. So I stage the entry rather than buy it all at spot: a starter is defensible here, with the bulk added into weakness or on confirmation.

Accumulate $58 – $63 The September gap and the 20/50-day moving-average confluence. Build in thirds.
Confirmation to add Q3 EBITDA ≥ $120M With the Customer C site energised and the Stateline lease commenced.
Invalidation Weekly close < $52 Loses the early-September base. A missed energisation date invalidates regardless of price.
1 Oct 2026 2032 notes close Use of proceeds and any change to the covenant package.
Early Nov 2026 [confirm] Q3 2026 results The first of the dated energisations that prove or refute the run-rate.
Q4 2026 / Q1 2027 Stateline lease, Hatchbo start 900 MW JV lease commencement; the Hatchbo rental begins 1 Jan 2027.
~28 Feb 2027 Q4 results, lock-up expiry First look at Q1 2027 actuals; ~3.6M Omega shares come off lock-up.
05

Glossary

Behind-the-meter
Power generated on the customer's own site, feeding the load directly without passing through the public grid.
Bridge power
Temporary on-site generation used until a permanent plant or grid connection is ready.
MW-month rental
Solaris's unit of revenue: a fixed fee per megawatt of capacity per month, paid whether or not the turbine runs.
Interconnection queue
The multi-year waiting line to connect a large new load or plant to the grid (ERCOT in Texas, PJM in the mid-Atlantic).
Adjusted EBITDA
Earnings before interest, tax, depreciation and amortisation, adjusted for one-off items. The profit the debt is sized against.
Net debt / EBITDA
Total debt minus cash, divided by annual EBITDA; the standard measure of how leveraged a company is.
High-yield (BB-)
A credit rating below investment grade; borrowing costs more and covenants matter more.
Covenant
A financial test in a loan agreement; breaching it can trigger default. Solaris's revolver tests leverage and interest cover.
Convertible note
Debt the holder can exchange for shares at a set conversion price.
Capped call
An option the issuer buys to offset share dilution from conversion, up to a cap price ($88.00 on the 2031 notes).
Non-recourse
Debt secured only on a specific project's assets; the lender cannot claim against the parent company.
IPP
Independent power producer: a company that generates electricity to sell under contract (Vistra, Constellation, Talen).
06 / Disclosures
Author. Noel Misthopoulos, Founder & Portfolio Manager, NSO Capital Asset Management.
First dissemination. 26 Sep 2026. Price referenced: $70.97, NYSE close on 25 Sep 2026.
Other interests. [NSO Capital has no advisory, banking or commercial relationship with the issuer and has not shared this idea with the issuer before publication.]
History. First publication. Any change to the view or reference value is dated and listed in the Ideas section.
Sources. Solaris Q2 2026 8-K and 10-Q; 8 Sep 2026 guidance raise; senior-note and convertible filings; company investor decks. Facts are distinguished from my interpretation throughout.
Not advice. General research for information only. Not personal investment advice, nor an offer or solicitation to buy or sell any security or interests in the ΣDX Fund.
← All ideas Read the cycle. Trade the cycle. Build with the cycle.