Innio Books $8 Billion of Engines Into a Tape That Just Fell Out of Love With AI Data Centers

A record $1 billion quarter, 19 GW of backlog and slot reservations into 2031 — defended, not celebrated, against a newly credit-anxious market

INIO · Earnings Call · 2026-10-27

The Order Book Outran the Tape

Innio's third quarter was, by the only measure management cares about, an inflection: “INEO delivered a record quarter, exceeding $1 billion in revenue for the first time” — Olaf Berlien, Chief Executive Officer · 2026-10-27. Equipment order intake has already cleared $6 billion year-to-date, and the equipment backlog reached $8 billion — 2.4x last year's level — with “a last-12-month book-to-bill ratio of 3.3x” — Dennis Schulze, Chief Financial Officer · 2026-10-27. The data center is no longer one vertical among many; it is the axis on which the whole story now turns. What makes this quarter structurally different from a simple cyclical upswing is the reservation mechanism. Slot reservations — signed, capacity-holding agreements that convert to firm orders six to eighteen months before delivery — give management a line of sight well beyond the normal order book. Olaf Berlien was explicit: “I do not see any slowdown for the next years. I am talking about with customer about projects in 2029, and already with 2031” — Olaf Berlien, Chief Executive Officer · 2026-10-27. That is a strong pipeline claim of unusual duration for a capital-goods maker, and it implies revenue visibility that the 19 GW backlog-and-reservation figure only begins to quantify.

What Management Started Defending Instead of Celebrating

Here is the tension worth paying for. Innio's own curated keyword set for this quarter is dominated not by growth language but by risk language. NeoCloud providers and credit risk both surface in the top tier of the company's themes — and both appear exclusively in the Q&A, in three separate answers, unprompted by anything in the prepared remarks. That is the signature of a topic investors are pressing on, not a topic the company is selling. Berlien's rebuttal was categorical: “we do not have orders with NeoCloud” — Olaf Berlien, Chief Executive Officer · 2026-10-27. Martin, who runs global data center sales, supplied the mechanical reason the exposure does not exist — and it is the single most informative passage on the call:

We are working with hyperscalers collocators, energy-as-a-service companies... We usually do not work with land developers... the NeoClouds very often work with LAN developers and reason we are not working with them is because they are very far away from the end customer And when they order the equipment, there is normally no back-to-back.
Martin · 2026-10-27

That distinction — counterparty credit quality as a function of distance from the end user — is precisely the fault line the public market has been repricing. The global tape shows the same theme in reverse: the AI data center basket of roughly 54 names is up about 54% over twelve months with 43 advancers, yet it has given back roughly 6% in both the trailing 90 and 30 days, with 48 decliners against 6 gainers. The AI data center complex is de-rating while Innio reports its best order quarter ever. Either the market is wrong about the marginal developer and right about the incumbent engine supplier, or Innio's backlog is marked at a discount it does not yet admit.

Water, Permits, and the Argument Nobody Saw Coming

The most company-unique moment on the call came when Ben Kallo of Baird asked about new state-level rules around data centers and Texas permitting. Management reported zero cancellations, one project delayed a week and since resolved, with the customer base concentrated in Texas, Virginia, Ohio and New York/New Jersey. Then the pivot: “The concern about water usage, to be honest, I like because it favors our technology.” — Martin · 2026-10-27 That is a rare thing — a regulatory constraint that is accretive. Water usage is a genuinely new keyword for this company this quarter, and it inverts the usual framing where environmental rules are a cost. Reciprocating-engine gensets consume materially less water than the open-cycle turbines they displace at these sites, which turns a permitting headline into a share-gain argument against gas-turbine incumbents — the classic "are you displacing anyone out in the field" question answered in the affirmative. Global peers confirm the demand but not the pricing power. Johnson Controls flagged a critical vertical in data centers and a cold plate market; Bloom Energy talked up 800V DC architectures; MaxLinear described its AI data center business growing off a multiyear cycle; Nucor cited data center construction activity. But Bloom also described electricity demand as flat. That dispersion — pervasive mentions of data center demand, no consensus on the electricity curve — is exactly where a niche engine supplier with a permitting advantage wants to sit.

Services: The Payoff Is Real, and It Is Late

The service business grew above 20% this year on a strengthening installed base, and service margin expanded from 29% to 31.1%. But Dennis Schulze deliberately damped the extrapolation: high single-digit growth for the next couple of years, with the real step-up arriving in the early 2030s as major overhauls become a larger share of the mix. The reason is arithmetic — Innio is adding a lot of young fleet, and young engines do not need overhauls. The economics behind that future step-up were quantified with unusual precision. Service intensity across the 19 GW of backlog and reservations runs “north of $50 thousand per megawatt per year” — Dennis Schulze, Chief Financial Officer · 2026-10-27, with data center projects at north of $70 thousand. Carl added the crucial caveat that the number is today's value, not inflated, with contracts indexed as inflation arrives. So the Service intensity embedded in the existing book is a nominal floor, not a ceiling — and it is a base that was never sold at a discount. Worth noting what the company chose to disclose on Page 24: a slide that did not exist before, mapping the long-term service agreements — ten years, up to fifteen, majority including the major overhaul — uniformly across hyperscaler, energy-as-a-service and rental-fleet customers. Investors asked for the bridge between the backlog and the annuity. Management drew it.

The Read

Innio is not riding a wave it discovered yesterday; it is riding one it has been selling into for several quarters, and the tape has stopped paying for the theme in the abstract. The company's answer is specificity: 3.3x book-to-bill, reservations into 2031, zero NeoCloud exposure, a water advantage against turbines, and a service annuity that is contractually indexed and back-end loaded. The risk is equally specific — a backlog that investors now insist on credit-underwriting themselves, and a services inflection that lives in the 2030s while the multiple lives today. A record quarter delivered into a skeptical tape is the most interesting kind of quarter there is.