NET Power Trades Its Carbon Capture Story for Wärtsilä Engines — and a Texas Site It Doesn't Own
Q3 2026: Entropy, PCC and 45Q vanish from the script, $108M of equipment lands on the balance sheet for an unsecured site, the customer is an unnamed frontier AI lab, and the call runs without a single analyst question.
NPWR · Earnings Call · 2026-11-13
The thesis that left without saying goodbye
For three consecutive quarters, NET Power's story was not a power plant. It was the Entropy team, the post combustion capture bolt-on, the PCC piece as a cheap, bankable retrofit onto gas turbines, a 45Q stack, and Oxy's EOR sink. Read the third-quarter 2026 prepared remarks and then read them again: Entropy is not mentioned. Neither is PCC, post-combustion capture, carbon capture, 45Q, or Oxy. Project Permian survives only as an afterthought — a "second project in parallel." The words that replaced them are gas power generation in its bluntest form: reciprocating engines, gas turbines, batteries, and a new self-description, "powered land developer." That is the story. In twelve weeks NET Power stopped being a carbon-capture technology company with a power project attached and became a developer of islanded power for data centers — a business whose competitive advantage is speed and whose product is a standardized block of engines. Management is explicit that this is a product, not a project: “we treat our islanded design as a product, not as a one-off project” — Marc Horstman, President and Chief Operating Officer · 2026-11-13.The customer is a name they won't say and a site they don't own
That single paragraph is the whole quarter. The initial project has a prospective customer — an unnamed frontier lab — and a cost reimbursement agreement capped near $180 million, a figure roughly 2.7x the company's entire $67 million market capitalization. It has no site, no definitive energy services agreement, and per the CFO's own framing, no committed financing. The reimbursement cap is the tell: it is a number large enough to sound like a contract and conditional enough to never be one.We are developing our initial project for a prospective customer who is a leading developer of frontier artificial intelligence models at a site in Texas that we have not yet secured, and that is separate from Project Permian... the customer must reimburse our non-refundable cost of approved expense items, up to approximately $180 million and subject to conditions, unless we elect to keep the related equipment.
Twelve months of runway, or one more engine order
The balance-sheet line is where the pivot gets expensive. Property, plant and equipment went from roughly $33 million at year-end to $141 million, of which $118 million is construction-in-progress on ten Wärtsilä reciprocating engines plus an EPC contract, and $23 million is deposits on two gas turbine packages. Against a $67 million market cap, the company has now committed more capital to physical equipment in nine months than the equity market says the whole enterprise is worth. The capex line tells the same story from the filings — capital spending peaked at $29 million in a single quarter, and the cumulative shift since year-end dwarfs quarterly research spend, which has already collapsed to $4 million from a $27 million run rate. Research and development fell to $4M And the cash burned to pay for it. On the August call the CEO described the cushion as “us having $300 million on the balance sheet affords us the patience” — Daniel Rice, Chief Executive Officer · 2026-08-14. Ninety days later: approximately $183 million, no debt, and a liquidity forecast that “assumes that we will not issue full notice to proceed under our EPC agreement or make additional significant equipment or construction commitments without committed financing” — Lee Shuman, Chief Financial Officer · 2026-11-13. Read that carefully — the twelve-month sufficiency claim depends on spending nothing more on the thing they just spent $141 million on. Cash runway reads as comfortable at 12.4 quarters, but it is a function of halting commitments, not of earnings. There are no earnings: operating loss was $210 million last quarter on essentially nothing but development spend.Pivoting into a de-rating
The irony is timing. NET Power has chosen to become a data-center power developer at precisely the moment the market is dismantling that trade. Across the last thirty days the AI data centers complex is 13 advancers against 57 decliners; Frontier AI lab exposure is down 26%, data-center revenue baskets down 15%. At AMAT this week the phrase attached to data centers was a data centers "slowdown in data center buildout"; at ONTO it was concerns about data center capital intensity. That is the demand curve NET Power's ten engines are pointed at. It is also the same demand curve the company insists is still vertical — data center demand sits atop its own keyword table, and the pitch leans hard on interconnection queue avoidance. Fair enough as an argument. But the keyword that carried this narrative for a year, speed to power, has quietly gone cold: it peaked as the company's single hottest theme in early 2026 and now registers barely off zero. The market has heard the phrase; it has stopped paying for it.What isn't there
Two absences matter as much as the pivot. First, the Q&A is gone. For a company that spent four prior calls fielding detailed questions — on equity checks, on $475 to $575 million project costs, on whether the technology could be retrofitted — this transcript goes from prepared remarks straight to the operator's close.That was the March number for a plant that no longer anchors the story. In August the promise was “hopefully, in the next couple of months, we'll have more to share” — Daniel Rice, Chief Executive Officer · 2026-08-14. Three months later there is still no definitive agreement, and this time no one was allowed to ask about it. Second, the equity is voting. NPWR has fallen 51.7% in ninety trading days, with a 58% decline concentrated in the last seven weeks — a 62% drawdown from the August $1.99 peak — on top of a 92% loss since 2021 and a 95% drawdown from the $16.51 high of September 2023. This is a sub-$70 million stub-priced option on a strategy that has now been rewritten twice. Danny Rice closed the call with “We have the team, we have the equipment, and we have a prospective customer.” — Danny Rice, Chief Executive Officer · 2026-11-13 True, and in that order of certainty. What he no longer has is a captured-CO2 story, a secured site, a signed contract, a Q&A session, or the $400 million he said a year ago he'd exit 2025 with — “we're going to exit the year with around $390 to $400 million of cash” — Danny Rice, Chief Executive Officer · 2025-11-14. The equipment is real. Everything that turns equipment into a business is still described in the conditional tense.we're sitting from a range of, call it, the upper 400s to the lower or to the upper 500. So call it, 475 to 575 is what we're looking for as a range.