ABB's Best Q3 Ever, Priced by a Market That Stopped Believing in AI Data Centers
The quarter that beat everything except the share price
ABB reported on 20 October and, by its own accounting, delivered the best third quarter it has ever produced. Orders came in just above $12 billion, up 40% on a comparable basis, with book-to-bill of 1.24 and an order backlog swelling to roughly $32 billion — “almost equivalent to 2025 full-year revenues” — Morten Wierod, CEO · 2026-10-20. Operational EBITA rose 24% to $2 billion and the group margin of 20.8% was, in the CEO's words, “the strongest Q3 margin ever for ABB” — Morten Wierod, CEO · 2026-10-20. Return on capital employed stood at 29.9%.
And yet earnings per share fell 2% to $0.65. Two items did it: a $115 million non-cash unrealised FX hedge tied to the pending Rotork acquisition, and a weaker result in discontinued operations as the robotics separation grinds toward completion. That gap — record operating performance, shrinking per-share earnings — is the tension underneath everything else in this call, and it explains why management spent so much of the hour reassuring analysts rather than celebrating.
The segment detail is unusually lopsided. Electrification revenue of $5.4 billion grew 21% comparably, with margin up 110 basis points to 25.6%. Motion grew 9% to $2.3 billion but margin fell 100 basis points to 19.1%, dragged by the still-lossmaking Gamesa Electric acquisition and inefficiencies in high power. Automation was flat on orders at $2.1 billion and grew revenue 5%, yet managed to lift margin 90 basis points to 14.8%. Free cash flow of $1.6 billion took the year-to-date figure to $3.7 billion, with the CFO steering toward something that “starts with a five” — Christian Nilsson, CFO · 2026-10-20 for the full year.
The data center story has quietly been rewritten
Data centers have driven ABB's narrative for at least six quarters — the term appears in this company's keyword history back in 2023 and never left. So the triple-digit order growth in the Electrification segment is not new news. What is new is the argument management now builds on top of it.
This quarter ABB launched Infinitus, a direct-current portfolio spanning source-to-rack, and with it a new vocabulary: DC architecture arrived as the single highest-ranked keyword in the company's deck, and gray space — the electrical rooms outside the white space of the rack hall — appeared alongside it. The claim is that in a target DC design, as much as 75% of electrical power infrastructure spend attaches to the gray space versus roughly 55% in a legacy AC build. Since ABB is historically stronger in gray space than white, a technology shift toward centralised high-density DC would expand its addressable scope without requiring it to win a rack-level fight.
We are very strong today on the gray side of the data center, not that strong on the white side… more of that investment is moving from the white space to the gray space, and with our leading technology here on the 800 volt DC and that DC architecture, we believe that this is a great opportunity for us to capture more of that growth.
That is a genuine evolution in the pitch. Twelve months ago, the same CEO was telling investors the opposite about scope: “we don't see any large benefits of being this one-stop shop… you need to be best-in-class in the different packages that are being awarded” — Morten Wierod, CEO · 2025-10-16. The shift from "packages" to "architecture" is the real content of this quarter. It is also still mostly promissory: solid-state transformers are orderable in the second half of 2027, and as recently as July the company conceded “we have zero order in the backlog for 800 volt DC data centers. Nobody has.” — Morten Wierod, Chief Executive Officer · 2026-07-16
The analyst pushback was pointed. Will Mackie of Kepler Cheuvreux asked directly what risk controls guard against pre-ordering and cancellations; Morten Wierod pointed to advance payments, cancellation policies, and the fact that these orders are engineered for specific permitted sites with grid connections and funding in place. James Moore of Rothschild asked whether hyperscaler returns can support the capex path at all — a question management answered with a circular-sounding observation about “deeper pockets than most kind of standard industrial customers” — Morten Wierod, CEO · 2026-10-20 and the coming contest between decarbonising industry and powering data centres for scarce grid capacity. Meanwhile the company's own hedge is explicit: “it is a tall order to expect triple-digit growth in each quarter going forward” — Christian Nilsson, CFO · 2026-10-20.
Japan: the deal machine restarts, and the balance sheet is the tell
The second genuinely new theme is geographic. ABB will take a 51% controlling stake in a joint venture with Mitsubishi Electric's low-voltage power distribution business — roughly 2,800 employees, about $640 million of fiscal 2025 revenue at a low double-digit EBITA margin, valued at an enterprise value near $890 million, or 1.4x sales and 11.1x EBITA, with a contractual path to full ownership and closing expected in the second half of 2027. Management's target is at least a 20% margin five years out.
The strategic logic is access, not cost. Japan is one of the world's largest electrification markets and one ABB has barely penetrated; the JV makes it a local leader overnight and, in management's framing, a door-opener into Malaysia, Thailand, Vietnam and Indonesia. It is being run explicitly on the GE Industrial Solutions playbook. Note also the appearance of management capacity at the top of the quarter's keyword list — an analyst question, not a management boast, and a fair one given Rotork, Mitsubishi and a robotics separation running simultaneously across different divisions.
The financial capacity is not in doubt. Net debt to EBITDA sits near 0.2x, and the CFO volunteered that moving toward 2x within the current rating would “free up, call it $13 billion-$14 billion” — Christian Nilsson, CFO · 2026-10-20. For a company that spent years shrinking, the re-rating question is no longer whether ABB can do deals but what it will overpay for.
What management stopped talking about — and what the tape is saying
Two themes that dominated prior calls have quietly exited the script. E-mobility, once a several-hundred-million-dollar annual drag discussed at length, now merits a single clause about lower losses. The robotics divestment, a running subject for over a year, is reduced to a target completion date. The portfolio surgery is essentially done; what remains is a pure-play electrification and automation company with a very large cheque book. Guidance was nudged up accordingly: low-teens comparable growth for 2026 and operational EBITA margin improvement of at least 270 basis points, helped by a roughly CHF 240 million real estate sale gain booked in Q4.
The awkward part is the tape. Across the market's AI infrastructure complex, the 30- and 90-day windows show broad, fresh declines: AI data center exposure is down about 5.5% with six advancers against 47 decliners, data center capacity names are off 8%, high performance compute down 7.6%, co-packaged optics down 5.2%. The 30-day advancer list is dominated by biotech, oil and gas and utilities — a rotation away from the very end market ABB says shows "no signs of demand softening."
Nor is data centres a differentiating theme among reporters. AQ.ST, KEY, OVH.PA, PLD and FNB all cited data centers this week. Meanwhile the market's editor-curated top keywords for the quarter are dominated by bank net-interest-margin vocabulary — retail deposits, loan beta, deposit spread — with no data centre term in the top 75 at all. ABB is riding a sector wave that the broader market has stopped paying a premium for.
Which is the honest read of this report: operationally excellent, strategically coherent, and exposed to a single customer segment whose equity market has already begun to de-rate. The lead time discipline and the large orders converting into backlog give ABB two years of visibility — roughly 40% of the $32 billion backlog converts in 2027. The question for the next four quarters is not whether the orders are real. It is whether the architecture thesis arrives before the market's patience with AI capex runs out.