ABB Ltd (Nasdaq Stockholm: ABB.ST) — A Real Turnaround, Now Buying at the Top of Its Own Cycle

Report date: 18 July 2026 · Price: SEK 945.20 (17 Jul 2026) · ADR (ABBNY): USD 98.15 · Market cap: ~USD 178bn (SEK ~1,716bn) · Enterprise value: ~USD 180bn Domicile: Zürich, Switzerland · Reporting currency: USD · Primary listing: SIX Swiss (ABBN); also Nasdaq Stockholm (ABB.ST), OTC ADR (ABBNY) · ISIN: CH0012221716 Sector: Industrials — Electrical Equipment & Automation · CEO: Morten Wierod (since Aug 2024) · CFO: Christian Nilsson (since Feb 2026) · Employees: ~111,000 FX used throughout: USD/SEK 9.6489, USD/CHF 0.8075 (17 Jul 2026)

Coverage status: new coverage.

⚡ Claude’s Take

This block is the author’s own subjective opinion, offered as general information and not as investment advice. The analysis in sections 1–15 below deliberately takes no position, contains no recommendation, and carries no price target.

Verdict: HOLD — a genuinely good business at a price that already pays for the good news. Accumulate below ~SEK 750–800 (ADR ~$78–83). Not a short.

ABB has done the hard thing and done it well. A sprawling, chronically underperforming conglomerate that nearly died in 2002 and drifted for a decade has, since 2019, been cut back to three businesses, decentralised, and made to earn its cost of capital — return on invested capital has gone from roughly 6% to roughly 17% on total capital employed (24–25% net of its cash pile) against a WACC near 8.5%. That is not financial engineering; it shows up in gross margin (33.0% to 41.1% in three years), in free cash flow (96% of net income), and in a share count down 14.8%. Stock-based compensation is 0.29% of revenue. Management even strips gains out of its own adjusted profit measure. On the evidence, this is one of the better-run industrials in the world, and the FY2025 numbers are real.

The problem is entirely price and timing, and it is compounded by a change in management’s own behaviour. Three things are true at once. First, the record is narrower than the headline. Q2 2026 orders of $12.0bn (+28%) are carried by one segment (Electrification +58%), one end-market (data centres, at triple-digit growth) and one geography (USA +62% — but base orders only ~+30%, so roughly half the US surge is lumpy large bookings); Automation orders fell 14% and its backlog was flat. Second, the reported profit improvement is substantially not operating. A $377m real-estate gain sits inside H1 Operational EBITA; strip it and H1 EPS grew ~2.4%, not the headline ~14%, while statutory operating margin actually fell 100bp to 16.7% even as the adjusted margin rose 90bp to 20.2%. A ~$2.3–2.4bn book gain on the Robotics close will flatter FY2026 further. Third, and most important, the allocator has changed character. The ABB that sold Power Grids, Dodge and Accelleron and bought back 219m shares at ~$30 is now paying ~19.5x EBITDA and a ~60% premium for Rotork — ~3.5% first-year return on the purchase price, against an 8.5% cost of capital — having told the market nine months earlier that the Robotics proceeds would be split between organic investment, M&A, dividends and buybacks. The buyback decelerated 35% quarter-on-quarter to fund it. The incentive plan explains the drift: ROCE is 10% of the annual bonus and zero in the long-term plan, where 50% (rising to 60%) sits on EPS — and a deal funded with low-yielding cash is EPS-accretive whether or not it earns its cost of capital.

The framing is quality-compounder-at-the-wrong-price, not falling knife and not crowded momentum. The factor evidence is unusual and worth stating precisely: beta 0.47, alpha 0.26, and the stock’s largest factor loading — a basket the model labels “Robotics & AI” (+0.52) — returned −10.8% over the twelve months in which ABB rose ~50%. Quality carries a zero loading. So ABB’s return has been earned idiosyncratically, by delivery, not by being carried by the market or by a hot factor. That is the most favourable version of “the market has started to agree” — and the most fragile, because it requires the delivery to continue. On 16 July the stock fell 5.4% on 2.6x normal volume into a record, guidance-raising quarter. That is the first evidence in a year that the bar is now rising faster than the delivery.

Two further cautions. The single best moat test available produced a negative result: triple-digit order growth in a capacity-constrained market yielded only ~2% pricing, with Wierod conceding that “many of the Hyperscalers are very large customers, and they have also leverage in a price negotiation.” A genuine bottleneck prices far above 2%. And the industry’s supply response — >$2.5bn of announced Western capacity, ABB’s own capex/D&A through 1.0x for the first time in seven years — lands in 2027–28, precisely when the data-centre order book converts. The supply response and the demand risk are correlated, which is exactly Marathon’s late-boom signature.

On valuation: ~25x EV/Operational EBITA (26.5x ex-gain), ~37x earnings, a 2.9% free-cash-flow yield, and P/S and P/B above the top of their own twelve-year ranges. Eleven of twelve comparable peers sit at or near record own-history valuations and every one drew a HOLD-or-worse conclusion; ABB’s ~17% ROIC is genuinely top-quartile in that cohort, but relative cheapness inside a uniformly re-rated cohort is not a margin of safety.

Tag: “The turnaround worked — now they’re paying boom prices for the encore.” Conviction: medium. Flips bullish if gross margin expands again while book-to-bill holds above ~1.1 — that would evidence genuine pricing power rather than volume leverage, and would make the current multiple defensible. Flips bearish if book-to-bill drops below 1.0 for two consecutive quarters, or if Rotork synergies are finally quantified and prove thin while another premium acquisition follows.

📈 Stock Price Action — Five-Year Event Map

Over five years ABB has run from SEK 233.20 (23 Jun 2022) to an all-time-high close of SEK 1,055.50 on 22 June 2026 — up 353% — and now trades at SEK 945.20, −10.5% off that high, inside a 52-week closing range of SEK 619.20–1,055.50. The twelve-month return is +49.9% in SEK. Two measurement cautions carried from the price work: FactorsToday’s rs_12m of +66.7% is struck from the day before a +9.6% result gap and overstates the move; and the 16 July reaction was −5.4% close-to-close (−7.8% open-to-close, which is not the convention, and press accounts range from −2.8% on the ADR to “about 3%”). For a USD-based holder the five-year annualised return is ~+26% against a far larger SEK move — the wedge is currency, and a Stockholm-listed, Swiss-domiciled, USD-reporting company hands its holders a three-currency problem before any operating result.

1–2. The 2022 low was made on rates and European multiple compression, not on ABB’s results; the recovery tracked delivery of the margin programme rather than any single catalyst. 3. The quietest leg of the series — no gap above 6%, an uneventful CEO transition. 4. The April-2025 tariff episode produced the only drawdown deeper than 25% in five years and was recovered inside four months, which is itself evidence the market treats tariffs as a cost rather than an access problem. 5. The 17 July 2025 session is the largest single up-day of the period. The October SoftBank agreement converted an execution story (a complex spin-off) into a cash event — a reversal in shareholders’ favour. 6. The Q4 2025 print delivered orders, EPS, dividend and buyback simultaneously; the March give-back has no company-specific cause on the record and is left unattributed rather than rationalised. 7. April 2026 was the strongest month of the five years, and the move began before the 16 April print — so the first leg is sector/macro, not results. 8. The 16 July 2026 session is the anomaly: the tape sold a record-orders, margin-expanding, guidance-raising quarter on 2.6x volume, alongside the largest acquisition in ABB’s history at a ~60% premium, with year-one EPS dilution and a management refusal to guide 2027. The move is factual; the attribution to the deal rather than the quarter is interpretation — but it is supported by the sell-side reaction (Deutsche Bank downgraded to Hold on the day; Vontobel, Bernstein and ZKB all criticised the price).

Long-run context, and it matters. On the split- and dividend-adjusted series ABB’s dot-com-era peak was SEK 120.14 (26 May 2000) and its trough SEK 4.29 (23 Oct 2002) — a −96.4% decline during the Combustion Engineering asbestos and liquidity crisis that nearly took the company into insolvency. This is not a business with a serene record of compounding. It is one that has been to the brink, and today’s rating embeds no memory of it.

1. Executive Summary

ABB is a Zürich-domiciled, USD-reporting electrical-equipment and automation group, listed primarily on SIX and secondarily on Nasdaq Stockholm, with ~111,000 employees and ~$35.8bn of trailing revenue. Following the agreed $5.375bn sale of Robotics & Discrete Automation to SoftBank — reported as discontinued operations since Q4 2025 — it operates three business areas: Electrification (~52% of revenue, 23.5% Operational EBITA margin), Motion (~25%, 19.4%) and Automation (~24%, 14.0%). Electrification generates 52% of revenue but 65% of segment profit; the 950bp margin spread across the three is evidence they are three different businesses, not one franchise.

The turnaround is genuine and the numbers support it. Since 2019 ABB has divested Power Grids, Dodge, Accelleron, Power Conversion and now Robotics, decentralised into ~20 accountable divisions, and lifted gross margin from 31.8% (FY2019) to 41.1% (FY2025). Return on invested capital has risen from ~6.6% to ~17% on total capital employed — or ~24–25% netting the ~$8bn cash pile, against ABB’s own disclosed ROCE of 25.3%. All three measures clear an ~8.5% WACC comfortably; the conclusion is robust to the definition, though the magnitude is not, and the disclosed figure is the most flattering of the three. Free cash flow is 96% of net income, stock-based compensation is 0.29% of revenue, net debt is 0.3x EBITDA, and the share count is down 14.8% since 2019.

But the current results are flattered, and the demand behind them is narrower than the headline. Q2 2026 delivered record orders of $12.0bn (+28% comparable), revenue +12%, a 20.2% Operational EBITA margin and a $30.0bn backlog. Beneath that: Electrification orders +58% while Automation orders fell 14% with a flat backlog; the Americas +52% and USA +62% against Germany −2% and Brazil −26%; data centres at triple-digit growth while utility orders were merely “stable” and residential “muted.” US base orders grew only ~30% against the +62% headline, so roughly half the surge is large, lumpy, non-repeating bookings. Separately, a $377m real-estate gain sits inside H1 Operational EBITA: ex-gain, H1 margin is ~19.8% rather than 21.8% and H1 EPS growth is ~2.4% rather than ~14%. Statutory operating margin fell 100bp to 16.7% in Q2 even as the adjusted margin rose 90bp — a 190bp divergence in direction, not just in level.

The moat is real but segmental, and the best available test of it failed. ABB holds strong positions across a loose oligopoly, but the top five vendors hold only ~20–25% of low voltage — too fragmented for industry-wide barriers in Greenwald’s sense. The genuine barriers (UL/NRTL certification with no IEC reciprocity for assemblies, utility and EPC qualification, code spec-in, installed-base service, distribution density) are real and durable, but they are shared equally by every incumbent, stabilising the oligopoly’s shape without conferring relative advantage. The decisive test: triple-digit order growth in a capacity-constrained market produced only ~2% pricing, with the CEO conceding hyperscaler negotiating leverage. A genuine bottleneck prices well above 2%.

Capital allocation has changed character, and the incentive plan explains why. The record through 2024 is strong — including a genuinely well-timed ~219m shares repurchased at ~$30 — but is dominated by one enormous error: Power Grids was sold to Hitachi at ~1.1x sales for $11bn in 2018–20, and Hitachi Energy is plausibly worth $38–51bn today, one year before the grid supercycle ABB now cites as its own demand driver. Since August 2024 the direction has reversed from disciplined seller to premium buyer: ~$5.5bn for Rotork at 5.3x sales and ~19.5x EBITDA with a ~60% premium, delivering a ~3.5% first-year return on the purchase price against an ~8.5% WACC, with synergies unquantified. In October 2025 management said the Robotics proceeds would be balanced across organic investment, M&A, dividends and buybacks; by July 2026 the entire ~$4.8bn was earmarked for Rotork and the buyback had decelerated 35% quarter-on-quarter. ROCE carries a 10% weight in the annual bonus and zero in the long-term plan, where 50% — rising to 60% in 2026 — sits on EPS. CEO bonus outcomes have not fallen below target in five years.

Valuation embeds continuation. At ~$180bn EV the stock trades at ~25.1x Operational EBITA (26.5x ex-gain), ~27.9x statutory EBIT, ~5.0x sales, ~37x earnings and a 2.9% FCF yield. Against its own twelve-year history, price/sales (~5.4x vs a 1.11–4.07x range) and price/book (~7.9x vs 1.80–5.94x) are above the top of the prior range, not merely at it. A comparative review of twelve peers found eleven at or near record own-history valuations, with a median cohort ROIC of ~13–15% — the sector re-rating has been a multiple event, not a returns event. ABB’s returns are genuinely top-quartile within it; its multiple is too.

The central question this report leaves open is whether ~20% Operational EBITA margins and a 28% ROCE are the new structural base or a cycle peak. ABB’s own Capital Markets Day band is 18–22% and its ROCE target is >20% — its own framework says current results are above normal. With >$2.5bn of Western capacity landing in 2027–28, ABB’s capex/D&A through 1.0x for the first time in seven years, and lead times not yet shortening, the supply response arrives in the same window as the data-centre order book converts. The risks are correlated: the cycle that produced the record orders is the cycle that set the currency ABB is spending on Rotork.

2. Business Overview

2.1 What ABB is, after a decade of subtraction

ABB Ltd is the product of a 1988 merger between Sweden’s ASEA and Switzerland’s BBC Brown Boveri, and for most of the subsequent thirty years it was a sprawling electrical conglomerate whose complexity destroyed value. The company today is defined as much by what it has sold as by what it retains. Since 2018 it has disposed of Power Grids (to Hitachi), Solar Inverters (Fimer), Mechanical Power Transmission/Dodge (RBC Bearings), Turbochargers (Accelleron, spun off), Power Conversion (AcBel) and now Robotics & Discrete Automation (SoftBank). Revenue from continuing operations is ~$35.8bn on a trailing basis against a group that once consolidated ~$10bn of additional grid revenue alone.

What remains is three business areas, reorganised at Q4 2025 when Robotics moved to discontinued operations and Machine Automation was folded into the renamed Automation business area. (Note: pre-Q4-2025 segment history is not directly comparable; ABB restated only FY2024.)

Segment percentages of profit are calculated before Corporate & Other and therefore sum to segment level, not group.

The concentration is the first thing an investor should internalise: Electrification is the company. Its 23.5% margin against Automation’s 14.0% is a 950bp spread, and the two businesses face entirely different competitive sets, customers and cycles. Valuing ABB as a single franchise obscures that roughly two-thirds of its profit comes from one business currently enjoying the most favourable demand conditions in its history.

2.2 How it makes money, and the quality of that revenue

Route to market differs sharply by segment, and this matters for moat analysis. Electrification is predominantly a channel business — sold through electrical distributors, panel builders, OEMs and contractors, with products specified into buildings and plants by consulting engineers. Motion sells through a mix of distribution (standard drives and motors) and direct/OEM (large machines, traction). Automation is overwhelmingly direct and project-based — DCS installations are sold to plant owners and EPCs, with multi-decade installed-base relationships. These are three different economic models: Electrification’s advantage is density and breadth across a fragmented channel; Automation’s is installed-base lock-in; Motion’s is closest to a scale-manufacturing cost game.

Recurring and service revenue is the key quality marker, ABB discloses it clearly, and the disclosure is the most important single fact in this section. In its ASC 606 revenue disaggregation ABB splits “Products” from “Services and other” by business area. The result inverts the intuitive story:

Service percentages are calculated; the underlying dollar figures are disclosed. Group service share was 16.5% (FY2024), 16.2% (FY2023) and 18.0% (FY2022), and is independently corroborated by ABB’s Integrated Report 2023 strategy graphic (“84% Products and solutions / 16% Services”).

Read the first and last columns together. Where ABB has genuine installed-base lock-in — Automation, where DCS platforms run 20–30 years and generate spares, upgrades and remote monitoring, and where service is 37.6% of revenue — it earns its lowest margin, 14.0%. Where it earns its highest margin, 23.5% in Electrification, service attach is just 7.3%, and it has been stuck in a 6.5–7.3% band for four consecutive years (7.0% FY2022, 6.5% FY2023, 7.0% FY2024, 7.3% FY2025). The inescapable conclusion is that ABB’s best margins are not being generated by recurring aftermarket annuity economics. They are being generated by equipment volume, mix and price in an exceptionally strong market. Anyone importing a “~16% service business” quality argument into Electrification is importing a process-automation number into a product business.

(An important qualification developed in section 4.1: Automation’s low margin does not mean low returns. Because it is asset-light — trade working capital of ~3.3% of revenue and ~87% direct sales — Process Automation standalone earns a 33.9% return on capital employed, above the group’s 25.3%. The margin ranking and the returns ranking point in opposite directions, and the returns ranking is the more meaningful one.)

Two fair qualifications. Electrification service is growing faster than any other business area off that low base — FY2025 service orders +16% comparable and service revenues +12% comparable, against group service revenue of +7% — and orders outpacing revenues is consistent with attach rising. Management is clearly pushing it: the Rotork rationale leans explicitly on “service expansion.” But four years of data show the ratio barely moving, and shifting a 7.3% attach rate materially is a multi-year project. [Note: the margin earned on service versus new equipment is not disclosed by ABB — nor by Schneider, Eaton or Hubbell. Any assumption that service carries materially higher margin is an assumption, not a fact.]

For context, Schneider Electric discloses Field Services at 11% of FY2025 revenue (targeting Software & Services at ~25% of group by 2030), while Eaton discloses no aftermarket or service revenue for its Electrical segments at all — though it does break out Aerospace aftermarket at 36.6% of that segment — and Hubbell discloses none (“aftermarket” appears zero times in its FY2025 10-K). That Eaton itemises aftermarket precisely where it is commercially central and declines to for Electrical is evidence that low service attach is a characteristic of the electrical-equipment industry, not a failing unique to ABB. (Caution: Schneider’s “Field Services” is a management business-model split and ABB’s “Services and other” is a US-GAAP performance-obligation disaggregation; they are not strictly comparable.)

Revenue is a mix of short-cycle and long-cycle that the market currently treats as more visible than it is. The $30.0bn backlog against ~$35.8bn of trailing revenue is roughly ten months of cover — real, but far thinner than, say, GE Vernova’s multi-year book. A book-to-bill below 1.0 would begin eroding revenue within three to four quarters. Electrification in particular carries a substantial short-cycle product component sold through distribution, which is precisely the revenue that destocks fastest: Rockwell’s shares fell ~8% peak-to-trough on a mere destocking cycle in which organic growth was +1%, all price, with volumes down 2%.

2.3 Geographic and end-market footprint

ABB manufactures on a local-for-local basis, which is central to its tariff resilience. Wierod, Q1 2025: “In the United States we cover as much as 75 to 80% of our sales with local production and we invest to increase this number. In Europe and China, we have already reached an even higher local value chain.” Rosengren had earlier put China and Europe self-sufficiency at ~95%. This footprint has kept the tariff impact “limited to the tens of millions” — a genuine structural advantage over import-dependent competitors, and one that has required real capital: ~$230m of announced US capacity in 2025 alone on top of ~$500m across 2022–24, plus ~$200m into European medium-voltage capacity announced May 2026.

End-market exposure has become materially more concentrated over the last eighteen months, and in the wrong direction for diversification. Data-centre demand is growing at triple-digit rates and drove Electrification orders past $7bn in a quarter for the first time, while process end-markets (pulp & paper, chemicals, mining) are described as “muted” or “soft” and residential is weak, particularly in China. ABB does not disclose data-centre revenue or orders as a percentage of group. Asked point-blank for the figure by James Moore (Rothschild) on the Q2 2026 call, CFO Nilsson declined to give it, offering only the bound that “if we exclude the data center segment, Electrification orders still increased by double digits.” That is a real disclosure gap on the single most important concentration question facing the company.

Verdict. ABB is a substantially better business than it was five years ago, and the improvement is structural rather than presentational — the portfolio is coherent, the operating model is genuinely decentralised, and the economics have followed. But it is not one business; it is three, with a 950bp margin spread, and roughly two-thirds of its profit now sits in the segment most exposed to a single, undisclosed, exceptionally hot end-market. The revenue is of good but not outstanding quality: a real backlog, but only ~10 months of cover; a meaningful short-cycle component sold through distribution; and — the finding that should most trouble a quality investor — a profit pool whose highest-margin component carries the lowest service attach in the group. ABB’s profitability today is an equipment story, not an annuity story, and equipment stories are cyclical.

3. Industry Dynamics

3.1 Structure: a loose oligopoly, not a protected one

ABB competes in three related but distinct industries, and the temptation to describe “electrical equipment” as a single consolidated oligopoly should be resisted. The top five global vendors hold more than 40% of switchgear but only ~20–25% of low voltage. Greenwald’s screen is unambiguous on what that means: if you cannot count the meaningful competitors on one hand, there are no industry-wide barriers to entry. ABB is strong-but-contested in every segment and dominant in none — a materially weaker structural position than, say, Emerson in DCS or Rockwell in North American discrete automation.

The competitive set by segment:

The right way to characterise this is Greenwald’s shared advantages among a few — a repeated pricing prisoner’s dilemma in which incumbents collectively enjoy real barriers against outsiders but compete vigorously against each other. That regime produces decent, defensible returns for good operators. It does not produce monopoly economics, and it does not by itself justify a monopoly multiple.

3.2 The barriers are real, durable — and shared equally

The single most robust barrier is regulatory and it is worth stating precisely, because it is frequently asserted loosely and is in fact demonstrable. In the United States, OSHA 29 CFR 1910.303(a) requires that electrical equipment be “approved”; 29 CFR 1910.399 resolves “approved” to certification by a Nationally Recognized Testing Laboratory; and 29 CFR 1910.7 provides that only OSHA may confer NRTL status. The NEC independently requires listed equipment be installed per its listing. The practical consequence is that enforcement runs against the end user — installing unlisted gear exposes the plant owner to citation and gives the authority having jurisdiction grounds to refuse occupancy. No specifying engineer, contractor or insurer will accept that risk.

Critically, IEC certification confers no credit. IEC 61439, the international standard for low-voltage assemblies, is absent from OSHA’s recognised test-standard list, as is UL 61439 — while UL 67 (panelboards), UL 489 (moulded-case breakers), UL 891 (switchboards), UL 845 (motor control centres) and UL 1558 (LV switchgear) are all present. The harmonisation pathway has been used extensively for components (UL 60947 and others) but never for the assemblies that constitute the bulk of electrification revenue. A European or Asian manufacturer holding IEC 61439 certification holds nothing in the US and must test from scratch, then submit to ongoing factory surveillance. OSHA is explicit that CE marking — a self-declaration regime — “is unrelated to the requirements for product safety in the United States.”

Layered on top, Build America Buy America (2 CFR Part 184) imposes a 55% domestic-content threshold on manufactured products in federally funded infrastructure, and the separate FAR Buy American schedule escalates from 65% (2024–28) to 75% from 2029 — mechanically raising the required depth of US manufacturing footprint over time and progressively closing the import-subassembly-and-finish workaround. Incumbents with existing US plants gain a widening advantage without doing anything. ABB’s 75–80% US local content and ~$730m of announced US capacity since 2022 position it well here.

But note what this barrier does and does not do. It protects the incumbent group against new entrants — and it is the best explanation for why there has been no successful Western entrant of scale in a generation, which is Greenwald’s strongest empirical test. It confers no relative advantage between ABB, Schneider, Siemens, Eaton and Hubbell, all of whom clear it equally. It stabilises the oligopoly’s shape; it does not determine who wins inside it.

3.3 Demand: the bull case is real and independently corroborated

The demand backdrop is the strongest in the industry’s modern history and this should not be understated. Global grid investment is rising ~20% to roughly $550bn in 2026 from ~$400bn, and needs to rise a further 50% by 2030; more than 2,500 GW sits stalled in interconnection queues. Hyperscaler capital expenditure is running from >$400bn in 2025 toward ~$725bn in 2026. Battery storage is growing ~35% to ~$110bn. These figures come from the IEA’s Electricity 2026 and Energy and AI work and from Dell’Oro, and are independently corroborated by the filings of eight peer manufacturers. The premise of the bull case — that electrification demand is in a multi-year structural upcycle — is sound.

3.4 The crux: how much of +28% is structural?

This is the question on which the investment case turns, and the honest answer is that the headline materially overstates the breadth of the demand. Five pieces of evidence, all from ABB’s own disclosure:

1. One segment. Electrification orders +58%. Automation orders fell 14%, and its backlog was flat quarter-on-quarter ($10.4bn → $10.5bn) — no backlog build at all in a “record” quarter. (Part of the Automation decline is a comp artifact against a ~$600m single prior-year booking, which should be acknowledged; the flat backlog is not.)
2. One end-market. Data centres at triple-digit growth, while utility orders were merely “stable” and residential “muted.” The grid/utility pillar most central to the structural narrative is not what is producing the growth.
3. One geography. Americas +52%, USA +62% — against Europe +12%, AMEA +12%, Germany −2%, Korea −8%, Brazil −26%. This is substantially a US data-centre capital-expenditure event wearing a global-supercycle costume.
4. Large-order driven. ABB discloses that US base orders grew ~30% against the +62% headline — so roughly half the US surge is large projects. Q4 2025 quantified the same effect: ~$1.2bn of a $10.3bn quarter (~12%) came from a handful of individual $100m+ projects, representing ~21% of Automation’s segment orders. Large orders are lumpy, non-repeating, and in some cases cancellable.
5. Price is only ~2%, and gross margin is falling. Volume is doing the work, which is higher-quality than growth driven purely by price. But group gross margin fell 50bp to 40.0% in Q2 2026 and Electrification’s fell 140bp, on price/cost gap, input costs and tariffs. If ABB cannot expand gross margin at a 1.28 book-to-bill with data centres growing at triple digits, it is difficult to construct the conditions under which it will.

[ASSUMPTION, clearly labelled: roughly half to two-thirds of the order surge is structural; the balance is AI-capex spike, large-order timing and shortage-driven forward ordering. ABB’s non-disclosure of data-centre share makes this impossible to pin down precisely, and reasonable analysts could bracket it differently.]

3.5 The capital cycle: late boom, and the supply response is already funded

Applying Marathon’s supply-side framework produces the most uncomfortable finding in this report, because every precondition for a classic capital-cycle top is present and documented:

• Returns are high and rising — the precondition that attracts capital. ABB’s ROIC has gone from ~3–6% (2019–20) to ~17% (2025); FY2025 incremental margin was 37%.
• Capital is responding, including ABB’s own. ABB’s capex has run 762 → 694 → 820 → 831 → 770 → 799 → $1,001m in 2025 (+25%), with ~$1,000m guided for 2026, while D&A has declined from $961m to $813m. The capex/D&A ratio has moved from ~0.79–0.92x to ~1.2–1.3x — above 1.0x for the first time in seven years. This is Marathon’s single most-cited negative signal, and ABB is now emitting it.
• Industry-wide capacity is committed. More than $2.5bn of announced Western expansion: Eaton’s $1bn programme with transformer capacity tripled since late 2023 plus a $340m South Carolina plant (2027); Schneider’s $140m US switchgear investment within $700m+ of US supply-chain spend to 2027; Siemens Energy’s $150m Charlotte transformer plant (2027); ABB’s own $230m US and $200m European commitments. Essentially all of it lands 2027–28.
• Assets are being bought at rising prices. Rotork at ~19.5x EBITDA; Specialtrasfo, bought into transformers at the point of maximum shortage; Gamesa Electric, Siemens Wiring Accessories, Sensorfact, Brightloop. Private equity is reported entering the space.
• Lead times are not yet shortening — the decisive tell, and it currently says “not yet.” MV switchgear ~44 weeks, LV ~54 weeks, LV draw-out 70–80 weeks, pad-mount transformers 110–130 weeks, large power/GSU transformers 128–144 weeks. Industry surveys expect no significant improvement in 2026 and only modest improvement in 2027.

The structural insight is the correlation. Because the build lag is two to three years, capacity committed in 2024–26 arrives in 2027–28 — the same window in which the data-centre order book converts and in which any hyperscaler digestion would occur. The supply response and the demand risk are not independent events; they are scheduled to arrive together. Marathon’s warning is precisely that the lag phase feels safest when it is most dangerous: prices stay elevated, reinforcing the bull narrative, exactly while the capacity that ends the shortage is being poured. The 50–140bp gross-margin erosion already visible at peak demand is the first crack in that wall.

3.6 China, tariffs and regulation

Tariffs have materially de-escalated and are a cost problem, not an access problem. The November 2025 US–Switzerland framework cut the tariff ceiling on Swiss goods from 39% to 15%, retroactive to 14 November, alongside a $200bn Swiss US-investment pledge in which ABB was named. Combined with 75–80% US local content, ABB has kept the tariff impact to “the tens of millions” and has passed cost through in price. Swiss domicile is largely a red herring for a business that manufactures locally in each major region.

China is a structural headwind rather than a growth market — but ABB disputes the specifics, and the disagreement is worth surfacing rather than resolving falsely. External market data supports genuine price pressure: Chinese vendors (Chint, Delixi, Inovance) are reported to undercut by 25–30% in low-voltage drives, the global LV drives market is growing only ~4.7% annually, and structural overcapacity and price wars are well documented. Against that, ABB’s management is on record contradicting the read-across to its own business. Asked point-blank by Jonathan Mounsey (Exane BNP) on the Q3 2025 call whether Chinese competitors were moving into Europe, Wierod did not answer the question, and confined his admission of “increased competition” to robotics — the business ABB has just sold — adding: “When you’re looking at the electrification, automation, there are also, of course, local competition. We have not — but we have not seen the same.” CFO Ihamuotila twice claimed Motion is growing in China on localised drive products with “a very competitive position there also locally,” and by Q2 2026 Wierod stated that “the price decline that we saw earlier, at least a year back in China, that has reversed.” China order growth has been +13%, +2%, −12% and +10% across the last four disclosed quarters, on ~14% of revenue.

[OPEN QUESTION, and an important one: ABB’s Motion margin fell ~130bp in Q2 2026, and it is tempting to attribute that to Chinese competition. ABB attributes it instead to Gamesa Electric operating at a loss (~70bp) and to lower Traction profitability on delayed volumes — neither of which is China. The industry price data is real; management’s denial is on the record; neither should be treated as settled. This report declines to assert a causal link the company explicitly denies, while noting that a management denial is not evidence either.]

Verdict: a good-but-not-great industry, enjoying exceptional cyclical conditions that are widely being mistaken for structural ones, with a supply side deteriorating in real time. The barriers are genuine — NRTL certification without IEC reciprocity, utility and EPC qualification, code specification, distribution density — and they explain the absence of any successful Western entrant of scale in a generation. Good operators genuinely earn above their cost of capital through a cycle, which is unusual and valuable, and ABB is one of the better operators in it. But the top five hold only ~20–25% of low voltage, so ABB’s advantages are segmental and local rather than a company-wide franchise; more than $2.5bn of capacity is landing in 2027–28; capex/D&A has crossed 1.0x; and the favourable conditions are narrow, with two of three segments showing zero FY2025 margin expansion. This is not an industry with barriers strong enough to defend both a ~17% ROIC and a ~25x EBITA multiple against the capital now being committed to it. The analytical base case should be Marathon’s default — that returns earned in a supply-constrained oligopoly mean-revert once supply normalises — with the burden of proof resting on the bull to name what structurally prevents it.

4. Competitive Position

4.1 Naming the moat, segment by segment

Greenwald’s discipline requires naming the mechanism or conceding there is none. Taking each business area in turn:

Electrification — economies of scale plus distribution density, with genuine but shallow customer captivity. This is ABB’s profit engine and its moat is real but narrower than the margin implies. The advantages are: breadth of portfolio (Wierod’s own defence, and a good one — “You need a very large offering to be a relevant player in that field… built up over even tens of years and with hundreds of millions of dollar in R&D investment year after year”); density in a fragmented distribution channel, where distributors resist carrying too many brands because it burdens their inventory and capital efficiency; and specification into projects by consulting engineers, which creates a soft form of lock-in at the design stage. Against that: the top five hold only ~20–25% of low voltage; service attach is just 7.3%, so there is little aftermarket annuity anchoring the customer; and the segment’s gross margin fell 140bp in the strongest demand quarter in its history. Verdict: a real scale-and-distribution advantage, but not customer captivity in any strong sense. This is a well-run share of a fragmented market, not a franchise.

Motion — cost and scale, the weakest position of the three. ABB claims #1/#2 globally in drives and industrial motors, and localisation in China has evidently worked. But drives and standard motors are the most technically commoditisable products ABB makes, with the largest and most capable low-cost competitor set. Service attach is 15.4%. Margins were flat in FY2025 at 19.4% and fell ~130bp in Q2 2026. Verdict: a scale cost advantage, genuine but eroding at the margin, in the segment most exposed to structural price competition.

Automation — the only genuine customer captivity in the group, and the segment whose economics are most easily misread. DCS is the textbook high-switching-cost business: a control system is embedded in a plant for 20–30 years, the cost of a rip-and-replace vastly exceeds the cost of the system, and the incumbent captures spares, upgrades, cybersecurity and remote monitoring. The lock-in here is documented, not asserted. ABB’s System 800xA ships orderable migration products for controllers dating to the 1980s product generations — MOD 300, Advant Master, INFI 90, Symphony Harmony/Melody, DCI System Six — and ABB’s published lifecycle policy guarantees a minimum of ten years’ support after any product leaves active sale, with twelve months’ notice of last-buy. The installed base is 35,000 DCS systems across more than 100 countries and roughly 100 million connected I/O points, built over 40 years, with service at 37.6% of segment revenue (and ~44% for Process Automation excluding the newly folded-in Machine Automation).

It would be easy — and wrong — to conclude from the 14.0% margin that this moat produces poor economics. ABB’s Capital Markets Day disclosure shows Process Automation standalone earning a return on capital employed of 33.9%, materially above the group’s 25.3%, on trade working capital of just 3.3% of revenue and ~87% direct sales. The combined Automation business area, diluted by Machine Automation, still earns 22.1%. This is a low-margin, high-return, asset-light service business — the moat is attached to a genuinely strong financial outcome, it simply shows up in capital efficiency rather than in margin. Verdict: the strongest and best-documented moat mechanism in the group, earning excellent returns on the capital it employs. The legitimate concerns here are different ones: orders fell 14% in Q2 2026, the backlog is flat, and ABB is number three or four in DCS behind Emerson and Honeywell — capturing the lock-in on its own base but not setting the industry’s pricing.

4.2 The decisive test: pricing power under maximum stress

The best available test of a moat is what happens to price when demand overwhelms supply. The industry has spent two years in acute shortage — lead times of 44 to 144 weeks — and ABB has just reported triple-digit data-centre order growth against a 1.28 book-to-bill. Under those conditions, a business with genuine pricing power should be extracting substantial price.

ABB extracted approximately 2%.

And when Karri Rinta (SB1 Markets) asked directly why price was such a modest driver, Wierod’s answer was candid and, for a moat analysis, damaging:

“many of the Hyperscalers are very large customers, and they have also leverage in a price negotiation.”

This is the single most important piece of competitive-position evidence in the report. A genuine bottleneck — a supplier that customers cannot go around — prices well above 2% when lead times run past a year. That ABB does not is prima facie evidence that hyperscaler buyer power offsets ABB’s supply position. The customers driving the boom are among the largest, most sophisticated and most concentrated buyers in the world economy; they multi-source deliberately; and they are perfectly capable of qualifying Schneider, Eaton, Siemens and Vertiv against ABB on every order. The same conclusion is corroborated by the gross-margin evidence: group gross margin down 50bp and Electrification down 140bp at the cycle’s peak.

This does not mean ABB has no moat. It means the moat protects ABB against entrants — via certification, qualification and distribution — but affords little protection against its established peers, and none at all against the negotiating leverage of its largest customers. That is precisely the “shared advantages among a few” regime, and it caps the through-cycle margin that a rational analyst should underwrite.

4.3 Market-share stability and the financial-outcome test

Greenwald’s share-stability test is the cleanest empirical check on a moat, and here the evidence is mixed and honestly incomplete. The strongest positive: no Western entrant of scale has successfully entered this industry in a generation — the incumbent set of ABB, Schneider, Siemens, Eaton, Legrand, Emerson, Honeywell and Rockwell has been stable for decades, which is powerful evidence that the collective barriers work. The strongest negative: within that set, share moves. ABB itself lost ground through the 2010s under a conglomerate structure and has been regaining it since 2020; Chinese vendors have taken material share in robotics on the government’s own five-year-plan design; and in low voltage the fragmentation is such that segment-level share data is not reliable enough to support strong claims. [OPEN QUESTION: granular multi-year segment share data by geography was not obtainable within this engagement and would be the highest-value additional evidence.]

The financial-outcome test (section 9 of our standards: if a moat claim cannot be tied to a financial outcome that would deteriorate without it, it is not a moat) produces a clear result. Automation’s 37.6% service attach is tied to a financial outcome — it is why that revenue is stickier and more predictable than Electrification’s. Electrification’s scale and distribution density are tied to an outcome — they are the most plausible explanation for a 23.5% margin against Automation’s 14.0%. But the “installed base annuity” claim frequently made about ABB’s electrification franchise fails the test outright: with 7.3% service attach and four years of no movement, there is no annuity there to deteriorate.

4.4 The aftermarket is contested, not captive — the strongest counter-evidence

A standard defence of electrical-equipment franchises is that the installed base yields a captive service annuity. The best evidence available says otherwise, and it comes from a competitor’s own marketing — a disclosure against interest.

Schneider Electric publishes that its service representatives “are qualified to work on ALL brands of equipment. In fact, over half of our work is on competitive brands!” — accompanied by a printed motor-control-centre cross-reference table listing the gear it will retrofit, built by Eaton/Westinghouse, Siemens/Allis, GE, Allen-Bradley, ITE, Furnas, Sylvania and Square D. Eaton runs the mirror image, retrofilling its Magnum DS cassettes into third-party enclosures and publishing a competitor cross-reference tool. If installed-base ownership conferred durable aftermarket capture, no OEM would volunteer that the majority of its service revenue is poached from rivals. Retrofit and retrofill are a contested, cross-brand land-grab in which the majors openly target each other’s bases.

US regulation reinforces the point with an asymmetry that runs against the OEMs where it matters most. The 2020 National Electrical Code prohibits reconditioning of moulded-case circuit breakers, panelboards, receptacles, GFCI/AFCI devices and fuses — the commodity, low-ASP end — while expressly permitting reconditioning of power circuit breakers, switchboards and switchgear. NEMA’s own policy concedes the same list. The code protects the wrong end of the mix: the entire high-value installed base is open to independent reconditioners, an ecosystem that includes consolidated roll-ups such as Group CBS. A further carve-out: NETA accreditation requires that a firm be “not associated with competing service or manufacturing interests,” which makes an OEM’s own field-service arm structurally ineligible wherever an owner’s specification demands a NETA Accredited Company — common in mission-critical and data-centre work.

Against this, specification position is a genuine and documented barrier — but an oligopoly-level one. Published institutional design standards name acceptable manufacturers explicitly: the University of Georgia’s Division 26 standard specifies for switchgear “Schneider/Square D, Eaton/Cutler-Hammer, and ABB/General Electric (ABB/GE). No exceptions.” Northwestern’s switchboard standard names three manufacturers and adds a single-source clause requiring board, breakers and accessories from one of them — an all-or-nothing package award with no component-level entry. But two facts limit what this is worth: every such list contains the same three or four names, so it excludes entrants without differentiating among incumbents; and federal procurement (UFGS 26 24 16.00 40, governing USACE, NAVFAC, AFCEC and NASA work) names no manufacturer at all and is purely performance-based, so a large, brand-blind, price-competitive segment exists by law.

Verdict: a genuinely strong competitive position, but a segmental and shared one rather than a durable company-wide franchise. ABB is a top-three player in a well-defended oligopoly and one of its better operators; the certification, qualification and distribution barriers are real and keep entrants out. But it is dominant in nothing, its highest-margin business has almost no aftermarket anchor, its strongest lock-in mechanism sits in its weakest-performing segment, and — decisively — it could not convert the tightest supply conditions in its history into more than ~2% of price against buyers who hold negotiating leverage over it. The moat is real. It is not wide, and it is not the moat the current multiple implies.

5. Growth History and Forward Opportunities

5.1 The historical record, correctly stated

Multi-year growth figures for ABB are a minefield and most published series are wrong, because the company has restated its perimeter repeatedly. Robotics moved to discontinued operations at Q4 2025 and ABB restated only FY2024 — so FY2023 and earlier are total-group figures while FY2024–25 are continuing operations, and FY2019–20 additionally exclude Power Grids. The apparent 5.1% revenue “decline” from FY2023 to FY2024 is entirely an artifact of this break and reflects no underlying deterioration.

*Continuing operations (ex-Robotics). Columns to FY2023 are total-group and are directionally comparable only.

The honest summary: underlying organic growth over the period has been mid-single-digit, and the transformation has been overwhelmingly a margin story rather than a growth story. Revenue on a comparable perimeter has compounded at roughly 5–6%; operating margin has more than doubled. An investor paying ~37x earnings is paying primarily for the margin achievement and for an assumption about the durability of the current order surge — not for a demonstrated high-growth franchise.

5.2 The current order surge — quality assessment

Q2 2026 orders of $12,042m (+30% reported, +28% comparable) with a book-to-bill of 1.27 and a backlog of $30,007m (+27%) are, on their face, exceptional. Section 3.4 sets out why the composition is narrower than the headline: Electrification +58% against Automation −14%; Americas +52% and USA +62% against Germany −2% and Brazil −26%; data centres at triple-digit growth against “stable” utility and “muted” residential; and US base orders up only ~30% against a +62% headline, implying roughly half the surge is large, lumpy bookings.

Two further quality markers deserve weight. First, growth is volume-led, not price-led (~2% price), which is genuinely higher quality than the alternative — several comparable peers grew on price alone with volumes declining. Second, and offsetting it, the CEO himself declined to extrapolate, unprompted:

“I cannot promise you that we will continue to make that kind of records… You will see some variation in the order intake.”

That is creditable candour and should be read as management guidance about the shape of what follows, not as boilerplate.

5.3 Forward opportunity set

The genuine opportunities, ranked by how well-evidenced they are:

1. Data-centre electrification — the largest and best-evidenced, with hyperscaler capex heading toward ~$725bn in 2026 and ABB positioned across MV/LV distribution, UPS (including the HiPerGuard 34.5kV medium-voltage UPS) and synchronous condensers. The constraint is not demand; it is that ABB cannot price it above ~2% and does not disclose its size.
2. Grid and utility investment — global grid capex rising ~20% to ~$550bn and needing +50% by 2030, with >2,500 GW in interconnection queues. Well-evidenced structurally, but conspicuously not what is currently driving ABB’s orders (“stable”).
3. Service attach expansion in Electrification — from 7.3%, growing at +12% comparable revenue and +16% orders. This is the highest-quality available growth because it would improve durability as well as level, and it is the explicit strategic rationale for Rotork. It is also unproven: four years of data show the ratio essentially flat.
4. Rotork and flow-control adjacency — adds ~3% to group revenue and ~12% to Automation, at a 24.6% operating margin, with claimed revenue and service synergies. Quality of the asset is high; price paid is the issue (see section 7).
5. Electrification of industry and buildings, reshoring, BESS — real but diffuse, and largely already inside the numbers.

Against these, two genuine drags: E-mobility, which has gone from a $445m annual loss (FY2024) to $205m (FY2025) and is being sold down piecemeal but remains loss-making; and Automation, where orders are falling and the backlog is flat.

Verdict: high-quality growth in composition, low-quality growth in breadth. The volume-led, price-light character of the surge is a positive marker, and the underlying demand drivers are structurally real and independently corroborated. But this is not broad-based growth: it is one segment, one end-market and one geography, roughly half of it in lumpy large orders, with the CEO explicitly warning against extrapolation and two of three business areas showing no margin expansion. Underlying compound growth on a stable perimeter has been mid-single-digit, and nothing in the forward set yet justifies underwriting materially more than that through a cycle.

6. Financial Quality

6.1 The transformation is real — and the measurement of it requires care

ABB’s financial improvement since 2019 is the strongest fact in the bull case, and it survives scrutiny. But three measurement traps must be cleared first, because most published figures on this company fall into at least one.

Trap 1 — the perimeter break. As set out in section 5.1, FY2023 and earlier are total-group; FY2024–25 are continuing operations. Growth rates spanning that boundary are meaningless without adjustment.

Trap 2 — third-party data is materially wrong on this name. Reconciling ROIC.ai’s aggregated figures to ABB’s own reporting produced five material discrepancies, and the company wins every one:

Balance-sheet line items do tie to the filing; the income statement, ratios and EV do not. Anyone screening ABB on aggregated data is working with a ~35% understatement of enterprise value and an EV/EBITDA multiple that is actually EV/EBIT.

Trap 3 — the $377m real-estate gain, treated in section 6.3.

6.2 The core financial record

*Continuing operations. Pre-FY2024 columns are total-group and directionally comparable only.

The gross-margin move from 33.0% (FY2022) to 41.1% (FY2025) — 810bp in three years — is the single most important number in the record, and it is not an accounting artifact. It reflects portfolio pruning (exiting low-margin businesses), the decentralised operating model’s cost discipline, price/cost recovery through the inflation cycle, and mix shift toward Electrification. However, it stalled and reversed in the most recent quarter: TTM gross margin is 39.7% and Q2 2026 gross margin fell 50bp to 40.0%, with Electrification down 140bp. That reversal, occurring at peak demand, is the most important negative datapoint in the financial record.

6.3 Quality of earnings — two gains that are not earnings, and one honest adjustment

The Q1 2026 real-estate gain of $377m is booked inside Operational EBITA, ABB’s own headline profit metric. The effects:

• H1 2026 Operational EBITA margin: 21.8% reported vs ~19.8% excluding the gain. The ~205bp contribution of a property sale is essentially the entire reported ~200bp year-on-year improvement.
• H1 2026 EPS: $1.41 reported vs ~$1.26 ex-gain vs $1.23 in H1 2025 — +2.4% underlying against a headline +14%. (Tax assumed at 25.5%; ASSUMPTION.)
• In Q1 alone, reported margin rose 320bp of which 250bp was the gain and only 70bp underlying — the headline overstated real improvement by roughly 4.6x.

In fairness to ABB, the disclosure is good. The gain is booked in Corporate & Other, so no business-area margin was flattered; management volunteered the split in three separate places and wrote the exclusion into its own guidance language (“should improve year-on-year, even when excluding the real estate gain in the first quarter of 2026”). This is disclosed, not hidden. But it sits inside the metric on which management is bonused and on which the market values the company, and a great many published FY2026 margin and EPS figures will not strip it out.

A second, larger distortion is coming. The Robotics business is held for sale at a net carrying value of ~$2.45bn against ~$4.8bn of expected net proceeds, implying an estimated ~$2.3–2.4bn book gain on close in H2 2026, landing in discontinued operations and therefore inside reported net income and EPS. FY2026 reported EPS will look extraordinary. Both items must be stripped from any FY2026 P/E.

Against those, one genuinely creditable adjustment. ABB’s Operational EBITA add-back wedge narrowed from $837m (2.74% of revenue) in FY2024 to $267m (0.80%) in FY2025 — and in FY2025 management excluded a $210m unrealised derivative gain from Operational EBITA. Very few management teams strip out gains as well as charges, and it deserves credit. Nonetheless the add-backs are structurally recurring — acquisition amortisation ~$185m, restructuring ~$92m, “business transformation”/ABB Way ~$166m, acquisition/divestment expense ~$55m, together ~$498m or ~1.5% of revenue annually, with FY2026 guidance confirming they persist. A sustainable operating margin is therefore closer to ~17.5% than to the 19.0% headline.

The most under-discussed signal is the divergence in direction. In Q2 2026 Operational EBITA margin rose 90bp to 20.2% while income-from-operations margin fell 100bp to 16.7% — a 190bp divergence, driven by ~$130m of non-operational items including ~$100m for “unasserted legacy claims and remediations” (undescribed), a ~$30m equity fair-value adjustment and ~$60m of derivative expense. When the adjusted and statutory measures move in opposite directions, the adjusted measure deserves less weight, not more.

Cash conversion is clean and this is a genuine strength. Operating cash flow to net income ran 1.15x, 1.19x and 1.16x in 2023, 2024 and 2025. The one poor year, 2022 (0.52x), was a $1,602m inventory build since unwound. FY2025 free cash flow of $4,566m equals 96% of net income; capex is $1,001m (3.0% of revenue, 1.23x depreciation). There is no earnings-versus-cash divergence to worry about. (One correction to a common assumption: ABB does not run negative working capital — trade net working capital was $4,059m at December 2025, 13.0% of revenue, though improving ~200bp a year.)

Stock-based compensation is $97m — 0.29% of revenue and 2.1% of free cash flow, with a diluted-versus-basic gap of only 4m shares. Against the software and even the industrial cohort this is exceptionally low and is an unambiguous quality marker.

6.4 Returns on capital — state the range, not the flattering point

Three different return figures circulate for ABB and all three are arithmetically correct. The gap is entirely definitional:

NOPAT $4,523m throughout; equity $16,646m; net debt $1,683m.

The investment conclusion is robust to the choice: on any definition ABB earns well above an ~8.5% WACC, and the improvement from FY2019 — when the same measures gave 5.5–6.6% — is real. That is the point that matters and it should not be lost in the definitional weeds. But the magnitude is not robust: quoting the disclosed 28.4% ROCE unqualified overstates the economics by roughly eleven percentage points against the most conservative defensible measure. This matters beyond presentation, because ABB’s annual bonus pays on ROCE (10% weight) using the company’s own most generous definition — the FY2025 outcome was scored at 136.4% of target. A metric that is both self-defined and the most flattering of three available measures is a weak governance instrument.

Three honest qualifications to the improvement: it is flattered by six years of divestitures shrinking the invested-capital base; by a 14.8% reduction in share count; and, at the margin, by the gains discussed above.

6.5 Balance sheet

Net debt is $2,320m at June 2026 — 0.3x EBITDA, which is conservative by any standard. Defined-benefit pension plans are overfunded by $519m, producing a $55m non-operational credit; pension is a non-risk here. Pro forma for receiving ~$4.8bn of Robotics proceeds and paying ~$5.5bn for Rotork, net debt moves to roughly $3.0bn, ~0.4x EBITDA. The balance sheet remains strong and is not a constraint on anything management wants to do.

The one balance-sheet concern is intangible. Goodwill of $9,637m plus other intangibles of $1,119m totals $10,756m against total equity of $16,646m — 65% of equity. Tangible book value is just $5,331m, or $2.94 per share, against a share price near $97.77 — a price/tangible-book of ~33x. Rotork will add roughly $4.5–5bn of further goodwill and intangibles, taking tangible book to approximately zero or negative. This is presentational rather than economic today, and it is normal for a serial acquirer of asset-light franchises. But it means two things: P/B and P/tangible-book are near-meaningless for valuing ABB (valuation must rest on earnings, cash flow and EV/EBITA), and there is no asset backing beneath the equity if the cycle turns. ABB ran quantitative goodwill impairment tests on five divisions in 2025, including E-mobility, and its own report warns the assumptions “could lead to a material goodwill impairment charge.”

Verdict: economics genuinely improve with scale, and the quality of the balance sheet and cash conversion is high — but the reported profit trajectory currently overstates the underlying one. Cash conversion at 96% of net income, SBC at 0.29% of revenue, net debt at 0.3x EBITDA and returns comfortably above cost of capital are all real and all good. Against that: the gross-margin expansion has stalled and reversed at the cycle peak; the adjusted and statutory margins are moving in opposite directions; H1 EPS growth is ~2.4% rather than ~14% once a property sale is removed; a further ~$2.3–2.4bn disposal gain will distort FY2026; and the sustainable operating margin is ~17.5%, not the ~19–20% the headline implies. This is a high-quality business whose current reported numbers are of noticeably lower quality than its underlying economics.

7. Capital Allocation

7.1 The deal record — a strong process undone by one enormous error

Power Grids is the decade’s defining error and it is testable, not rhetorical. ABB sold a business with ~$10bn of annualised revenue for $11bn enterprise value — approximately 1.1x sales — with the 80.1% closing in July 2020 and the residual 19.9% following in December 2022. Total gross proceeds were ~$10.8bn. Hitachi Energy today generates ~$19.8bn of revenue (+26% year-on-year) at a 12.9% adjusted EBITA margin with an order backlog near $50bn, and Hitachi targets 13–15% revenue CAGR to 2030 and a 16–20% margin. At 15–20x EBITA on implied FY2025 EBITA of ~$2.55bn, Hitachi Energy is plausibly worth $38–51bn today against the $11bn ABB accepted — value forgone of roughly $27–40bn, more than ABB’s entire market capitalisation at the time of sale. ABB sold the grid asset at 1.1x sales one year before the grid supercycle it now cites as a demand driver for its own Electrification business.

The mitigant is real and deserves weight. ABB redeployed the proceeds unusually well, repurchasing 218,686,689 shares for ~$6.6bn at an average of ~$30/share between July 2020 and 2022 — into a stock that has since roughly tripled. Those shares would be worth roughly three times what was paid. That is genuinely good buyback timing and recovers a meaningful fraction of the forgone value. It does not offset it. Net verdict: strongly value-destructive versus retaining the asset, partially rescued by excellent redeployment.

7.2 Returns of capital — well executed, with a currency and tax friction

The dividend is a single annual payment in CHF: 0.80 (2019), 0.80, 0.82, 0.84, 0.87, 0.90, 0.94 (2025) — five consecutive increases, but a ~2.7% CAGR against far faster EPS growth. The cash payout ratio has drifted down from ~47% to ~41%. This is deliberate design — a low-volatility dividend floor with the buyback as the swing instrument — and is defensible, but income-oriented holders should not expect the dividend to track earnings.

For a Nasdaq Stockholm holder there is a real, recurring friction. Swiss dividends carry 35% federal withholding tax at source; the Sweden–Switzerland treaty typically reduces the effective rate toward 15%, but the balance is recoverable only by filing a reclaim. Twenty percentage points of the gross dividend is withheld pending an administrative process, and a non-treaty or tax-exempt holder can suffer permanent leakage. Combined with the three-currency structure (USD reporting, SEK quotation, CHF dividend and cost base), the income component of return is meaningfully degraded relative to the headline ~1.2% yield.

Buybacks have been the better instrument and have been well executed: ~$1,258m (2023), $1,247m (2024) and $1,499m (2025), with the share count down from ~2,133m (2019) to 1,815m (Q2 2026), −14.8%. A new $2.0bn programme launched in February 2026. But its pace decelerated ~35% quarter-on-quarter in Q2 2026, to just $147m from $225m — cash is visibly being retained for Rotork.

7.3 Rotork — the decision that changed the story

Announced alongside Q2 2026 results on 16 July: 503 pence per share in cash, ~£4.14bn equity value, ~$5.5bn enterprise value, a ~60% premium to the three-month average and ~73% to the prior close, executed as a UK scheme of arrangement, expected to close H1 2027. It is the largest acquisition in ABB’s history, exceeding Baldor ($4.2bn, 2011).

Rotork is a genuinely good asset: FY2025 revenue £777.3m (+3.0%), adjusted operating margin 24.6%, ROCE 38.4%, service 24% of sales, and 8% average organic growth 2022–25. It is a quality upgrade on the Robotics business it replaces (12.1% margin).

The arithmetic is the problem, and it is arithmetic rather than opinion. Rotork’s ~$1.04bn of revenue at a 24.6% margin is ~$256m of operating profit. Taxed at ~22%, that is roughly $192m of NOPAT on $5.5bn of purchase-price capital — a ~3.5% first-year return on invested capital, against ABB’s ~8.5% WACC. Achieving the claimed reduction in the multiple from ~19.5x to “the mid-teens” requires ~$85–90m of incremental EBITDA; even fully delivered, post-synergy NOPAT of ~$258m is a ~4.7% return. Management declined twice to quantify the synergies, which are described as mainly revenue- and service-driven — the least reliable category.

Management’s two accretion claims are both true and both beside the point. Rotork’s 24.6% margin is above ABB’s 19.0%, so bolting it on raises the group margin percentage arithmetically. Funding it with cash yielding 2–3% and cheap debt against an asset yielding ~4.5% earnings-to-price makes it EPS-accretive. Neither fact means the capital earns its cost. This is precisely the distinction the incentive plan fails to make (see section 7.4).

The multiple arbitrage runs backwards. ABB is selling Robotics at roughly 17–19x EBITA and buying Rotork at ~19.5x EV/EBITDA — and because EBITDA exceeds EBITA, Rotork’s EV/EBITA is higher still, around 21–22x. ABB is recycling proceeds from a lower-margin asset into a higher-margin asset at a modestly higher multiple. That is a defensible portfolio quality upgrade. It is not the “sell high, buy low” discipline the prior decade’s record implied.

And the use-of-proceeds story changed. In October 2025 Wierod said the Robotics proceeds would go to “more organic investment… more focus on M&A… also increasing dividends and share buybacks. So we will find the right balance here over time.” By July 2026 the entire ~$4.8bn net was earmarked for Rotork and the buyback had decelerated 35%. The market noticed: Deutsche Bank downgraded ABB to Hold on the day, Vontobel called the multiple excessive, Bernstein called it expensive, and Berenberg rated Rotork Hold at 503p — exactly the offer price. On the call, Alessandro (Octavian) computed a 5–10% entry ROIC and said ABB would need to “double EBIT”; CFO Nilsson did not dispute the starting return, contesting only the required uplift.

Verdict on Rotork: not disciplined. A ~60% premium, 5.3x sales, ~19.5x EBITDA and unquantified synergies for an asset that grew 3% in its most recent year with a cautious outlook is the profile of a strategically motivated buyer with cash to deploy, not a returns-disciplined allocator. The asset is good; the price is full; the timing — at the top of ABB’s own cycle, paying with peak-cycle currency — is the concern.

7.4 Incentives: the instrument panel explains the drift

This is the most important governance finding in the report.

Annual incentive (AIP), target 100% of base salary, maximum 150%. FY2025 weights for the CEO: Group Operational EBITA margin 30%; Revenues 30%; FCF conversion 20%; ROCE 10%; sustainability 10%.

Long-term incentive (LTIP), CEO target 150% of base (rising to 200% from 2026), maximum 300%. Since 2022: Average EPS 50%, Relative TSR 30%, Sustainability 20%. There is no ROCE, no return-on-capital and no cash-flow measure whatsoever. From 2026 sustainability is removed and the weights become EPS 60% / relative TSR 40% — increasing the EPS weight further.

So return on capital carries a 10% weight in the annual bonus and a zero weight in the long-term plan, while 50% — soon 60% — of long-term pay rests on EPS. For a company that has just committed the largest cheque in its history to an asset earning ~3.5% on the purchase price, this is the wrong instrument panel. A cash-and-debt-funded acquisition of a high-margin asset is close to guaranteed to raise EPS and the group margin percentage — both heavily weighted — while diluting return on capital, which is barely weighted at all. The incentive structure does not merely fail to discourage the Rotork transaction; it actively rewards it.

Are the targets demanding? No. CEO annual bonus outcomes over five years were 145.0%, 120.0%, 149.0%, 120.7% and 121.8% of target — never once below target, averaging 131.3%. LTIP vesting across the last five grants averaged 149.8% of target, including one at the 200% maximum. The LTIP’s EPS threshold and maximum are set at merely −11.4% and +11.4% around target, a narrow band that converts modest outperformance into maximum vesting.

Three elements are genuinely well designed and deserve credit. The EPS measure explicitly neutralises buybacks — “The impact of share buybacks will not be considered as an adjustment” — which is rare, correct, and removes the incentive to manufacture EPS with repurchases. Relative TSR is measured against a credible fifteen-company peer set (including Schneider, Siemens, Eaton, Emerson, Legrand, Rockwell). And executive shareholding requirements are demanding — 500% of base salary for the CEO, tightening from a net- to a gross-salary basis in 2026 — with Wierod holding ~1,200% of net salary, roughly 2.4x his requirement. His ~CHF 9m+ of ABB stock against CHF 6.1m of annual pay is meaningful personal wealth at risk.

Ownership and board. Investor AB (the Wallenberg vehicle) is the largest shareholder at 14.4% with one attributable board seat (Johan Forssell) — influence without control, and a patient, industrially literate anchor that is a credible check on short-termism. Cevian Capital, the activist behind the 2019–23 breakup agenda, cut its stake below 5% in February 2024 and subsequently below 3%, with its board representative standing down. The activist that drove the disciplined-seller era has left, precisely as the premium-buyer era begins — a correlation worth noting. Chair Peter Voser has served eleven years; board fee structure (fixed only, half in restricted shares, no options) is best practice, but a chair of that vintage presiding over this pivot is the situation in which fresh challenge is most valuable.

R&D is the quiet weakness. At 4.0% of revenue ABB sits at the bottom of its peer set — against Siemens at 8.3% and Schneider at ~5% — and management has conceded the point by targeting 4.5–5.0%. A target is not spending. Capex at ~3.0% of revenue is in line with peers.

Verdict: a strong historical record that has clearly changed direction, governed by an incentive system that does not measure the thing that matters. The 2019–24 era — divest, simplify, buy back stock at ~$30 — was excellent capital allocation, marred by one very large error in Power Grids. The 2024–26 era is different: the largest acquisition in company history at a full price and a below-WACC first-year return, funded by proceeds management had promised to spread across four uses, with the buyback slowing and the activist gone. Management is not reckless and the balance sheet is not stretched. But with ROCE at 10% of the bonus and zero in the long-term plan, and bonus outcomes that have not been below target in five years, there is no mechanism in the pay structure that would stop the next Rotork.

8. Changes and Headwinds — Last Two Years

Sourcing note: this timeline is built from ABB’s own SIX ad-hoc releases, the Q2 2026 interim report and slides, the FY2025 Financial Report, the Rotork offer announcement and SoftBank’s own release.

Litigation and compliance are quiet. Aggregate recognised liabilities for regulatory, compliance and legal contingencies were $25m at 31 December 2025 (down from $72m), with “none of the individual liabilities recognized was significant.” The 2017 South Korea forex/embezzlement fraud and older cartel and bribery matters do not appear as live contingencies, and no cybersecurity incident is disclosed. [OPEN QUESTION: $25m at a $33bn-revenue industrial with this history is either genuinely clean or thinly disclosed; the disclosure enumerates no individual matter. Nothing public contradicts it. Residual exposure sits in $1,944m of third-party guarantees running to 2032 and one legacy non-core EPC project of ~$100m whose exit timing ABB says is “difficult to assess due to legal proceedings.”]

Tariffs are a disclosed, absorbed cost rather than a hypothetical. FY2025 operating income growth “more than offset the negative impact of United States tariffs, rising raw material spend and higher labor costs.” Q2 2026 gross margin fell 50bp to 40.0% on “higher expenses related to commodities and tariffs,” offset by volume leverage and ~2% positive pricing. [OPEN QUESTION: ABB never quantifies an absolute tariff cost.]

Verdict: the changes strengthen the operating thesis and simultaneously raise the risk attached to it. Three things are unambiguously better — the demand environment, the margin structure (now inside a band ABB itself raised in November 2025), and the E-mobility wound, cauterised from a $445m annual bleed to ~$18m a quarter. Leadership has also proved willing to reverse itself in shareholders’ favour: abandoning a complex spin-off for $5.375bn of cash is the right kind of U-turn. But the thesis now rests on a single, unhedged capital-allocation swap executed at what is visibly a cyclical high — ROCE 28.4% against a >20% target, margins above the through-cycle band, orders +28%, and a backlog carried disproportionately by one theme. Buying a peak-multiple asset with peak-cycle currency is how good industrials destroy a decade of compounding. The company is better; the margin of safety is thinner.

9. Risk Analysis

How to read this matrix. The high-likelihood rows — FX, tariffs, withholding tax, buyer power — are chronic frictions of moderate impact: costs of owning this security rather than thesis-breakers. The rows that decide the outcome are the three medium-likelihood, high-impact ones: order normalisation, the Rotork price, and the margin’s distance above its own through-cycle band. Critically, they are correlated. The same cycle that produced the record orders and the 28.4% ROCE is the cycle that set the currency ABB is paying with, and the industry capacity that ends the shortage lands in the same 2027–28 window in which the data-centre book converts. If it turns, these three go wrong together — which is precisely why the position sizing implied by any one of them individually understates the risk.

Catastrophic-loss risk is low. Net debt at 0.3x EBITDA, an overfunded pension, 96% cash conversion and a $30bn backlog make insolvency a remote scenario. The realistic bear case is a de-rating and an earnings reset, not a permanent capital loss — though the 2002 precedent (−96.4% peak-to-trough, see the price-action section above) is a reminder that “remote” is not “impossible” for this particular company.

10. Valuation Discussion

No price target and no recommendation. This section establishes what the current price requires the business to deliver.

10.1 Where the multiple sits

At SEK 945.20 (ADR $98.15), with 1,815m shares and net debt of $2.32bn, market capitalisation is ~$177.8bn and enterprise value ~$180.3bn.

Per share: TTM EPS $2.77 (SEK 26.8); book value $8.86 (SEK 86); tangible book $2.94 (SEK 28).

10.2 Own-history context — above the top of the twelve-year range

A twelve-year multiple series taken from the ADR line serves as the own-history reference:

The 2019–21 P/E readings are distorted by disposal gains and are not comparable run-rate figures; read P/S and P/B across those years instead.

On the two cleanest series, ABB is above the top of its own twelve-year range, not merely at the high end of it. Price/sales at ~5.4x compares with a 2014–25 range of 1.11–4.07x; price/book at ~7.9x against 1.80–5.94x. EV/EBIT at ~27.9x is below only the distorted 2020 reading. This is the highest valuation ABB has carried in over a decade, on a business whose margins its own management describes as within — not below — a through-cycle band.

10.3 Peer context — a uniformly re-rated cohort

Drawn from comparable-company analysis, June–July 2026.

Three observations. Eleven of twelve peers sit at or near their richest-ever own-history valuation — Ingersoll Rand the only exception — and every prior report concluded HOLD or worse; not one was a fresh-money buy. Returns are far weaker than the multiples imply: median cohort ROIC is ~13–15%, barely above a ~9% WACC, against 20–35x EBITDA. The sector re-rating has been a multiple event, not a returns event. And ABB’s ~17% ROIC is genuinely top-quartile — behind only Vertiv and Powell, level with Parker — but its ~24.6x EV/EBITDA is top-quartile too. Relative cheapness inside a uniformly re-rated cohort is not a margin of safety.

Note a real gap: Schneider Electric, ABB’s single most direct global comparable, is not covered in this comparative set.

10.4 Embedded expectations — what the price requires

Rather than construct a false-precision DCF, the more useful exercise is to ask what must be true. Taking EV of $180.3bn and a ~$7.19bn TTM Operational EBITA ($6.81bn ex-gain):

What the market is underwriting at ~26.5x ex-gain Operational EBITA:

1. That ~20% Operational EBITA margins are the new structural base, not a cycle peak. ABB’s own CMD band is 18–22%; the sustainable margin after recurring add-backs is ~17.5% (see section 6.3). The current multiple makes little sense on a 17.5% through-cycle margin and considerable sense on a sustained 21–22%.
2. That order growth normalises gently rather than reverting. A $30bn backlog is ~10 months of revenue. Sustaining anything near current revenue growth requires the data-centre order rate to persist for several more years — while ABB’s own CEO says “I cannot promise you that we will continue to make that kind of records.”
3. That the ~2% pricing environment improves, or that volume leverage substitutes for it indefinitely. Gross margin is currently falling at peak demand.
4. That Rotork’s synergies materialise. At the entry return of ~3.5%, roughly $85–90m of incremental EBITDA is required merely to reach a mid-teens multiple, and considerably more to earn the cost of capital.
5. That the 2027–28 capacity wave does not compress industry pricing as it lands into the same window in which the data-centre book converts.

What the market appears to be pricing correctly: the genuine quality of the transformation, the strength of the balance sheet, the cash conversion, the low SBC, and the reality of the electrification demand backdrop. None of these is in dispute.

Scenario frame (ASSUMPTIONS throughout; illustrative, not forecasts):

The asymmetry to note: in the base case the investor waits several years for earnings to justify today’s price; in the bear case both terms of the equation move against them simultaneously; the bull case requires ABB to do something — earn pricing power over hyperscalers, or materially raise Electrification service attach — that it has not yet demonstrated.

Verdict: the price embeds continuation of exceptional conditions in a business whose own management describes current margins as within, not below, its through-cycle band, and whose own CEO declines to extrapolate the order rate. ABB is a genuinely better business than its twelve-year multiple history reflects, so some re-rating is warranted and arguing for a return to a 2.5x price/sales multiple would be wrong. But at ~5.4x sales and ~7.9x book — above the top of the prior range — the stock is priced for the boom to be the base case, with the supply response already funded and landing in the same window as the demand risk.

11. Variant Perception

The consensus view is that ABB is a structurally transformed, high-quality electrification compounder with a decade-long runway from grid investment, data centres and industrial electrification, deserving a premium multiple and a place among the best industrials globally. Every element of that is defensible and most of it is correct. Sell-side sentiment is broadly constructive — Morgan Stanley judged that Q2 “cleared the key hurdles,” Berenberg expected “further earnings upgrades,” and Nordea raised its target into the print.

The strongest bull case, stated at its most persuasive: the electrification demand cycle is a genuine multi-decade structural shift, not a capex spike — grid investment must rise 50% by 2030 with 2,500 GW stalled in queues, and AI power demand is additive to, not a substitute for, that. ABB has three top-tier positions, a decentralised model that has demonstrably worked, a fortress balance sheet at 0.3x leverage, best-in-class cash conversion, negligible dilution, and returns at 17–25% against an 8.5% WACC. Rotork is a 24.6%-margin, 38% ROCE asset with 24% service revenue — exactly the kind of business that should be bought when available, and available assets of that quality are rare at any price. The record orders are evidence of share gain in the most important end-market of the decade. On this reading, arguing about 19.5x versus 15x for Rotork is quibbling over the entry price of a permanent quality upgrade, and the current multiple is what quality costs.

The strongest bear case: this is a cyclical industrial at the top of the best cycle in its history, being valued as a secular compounder. Margins at 20.2% are above ABB’s own through-cycle band midpoint and the sustainable figure after recurring add-backs is ~17.5%. The H1 profit improvement is a property sale. Growth is one segment, one end-market, one geography, and roughly half of the US surge is lumpy large orders that will not repeat. Pricing power failed its cleanest test — ~2% price at a 1.28 book-to-bill with year-plus lead times, because hyperscalers hold the negotiating leverage — and gross margin is already falling at peak demand. The industry is committing >$2.5bn of capacity that lands in 2027–28, exactly when the order book converts, and ABB’s own capex/D&A has crossed 1.0x for the first time in seven years. Management has pivoted from disciplined seller to premium buyer at the cycle top, with incentives that reward EPS and ignore return on capital, and with the activist that enforced the prior discipline now gone. The stock is above the top of its twelve-year price/sales and price/book ranges in a peer cohort that is uniformly at record valuations. Tangible book is about to go to zero.

The 3–5 assumptions that actually matter:

1. Is ~20% Operational EBITA structural or cyclical? Everything else is secondary. ABB’s own CMD band (18–22%) and the ~17.5% post-add-back figure argue cyclical-to-fair; the mix shift toward Electrification argues structural.
2. Is the data-centre order rate an annuity or a spike? Unanswerable from outside because ABB will not disclose the exposure — the single most consequential non-disclosure in this report.
3. Can ABB ever price above ~2%? If buyer power permanently caps price, the through-cycle margin is capped with it regardless of demand.
4. Does the 2027–28 capacity wave compress industry returns? Marathon’s framework says yes by default; the burden is on the bull to name what prevents it.
5. Will Rotork earn its cost of capital, and is it the first of several? The deal itself is survivable at ~3% of revenue. A pattern of 19.5x acquisitions funded by disposal proceeds would be thesis-breaking.

What the tape and factor positioning add. The positioning read is unusual and cuts against the lazy conclusion. ABB is not a crowded momentum trade: beta is 0.47, alpha 0.26, idiosyncratic volatility 21.2% against 32.2% total, and only ~55% of variance is factor-explained. Most strikingly, its largest factor loading is to a basket labelled “Robotics & AI” (+0.52), which returned −10.8% over the twelve months in which ABB rose ~50% — so the return was earned despite its dominant factor exposure, not because of it. Quality carries a zero loading in all four models. Risk-adjusted, the last year (Sharpe 1.62, max drawdown −16.3%) is the smoothest stretch in the security’s recorded history, against a lifetime Sharpe of 0.40 and a lifetime drawdown of −71.9%.

The variant perception this report offers is therefore not “ABB is a bad business” — it plainly is not — nor “the electrification cycle is fake” — it plainly is not. It is narrower and, I think, more defensible: the market is treating a demonstrated margin transformation as though it were a demonstrated pricing-power transformation, and they are not the same thing. ABB has proved it can take cost out, prune a portfolio, decentralise accountability and convert profit to cash. It has not proved it can price above its competitors or above its customers’ tolerance — the 2% print at maximum scarcity, the falling gross margin at peak demand, and the 7.3% Electrification service attach all say the opposite. Consensus is extrapolating the first achievement into the second. If the market is offsides anywhere, it is there — and the correlated arrival of the supply response and the demand risk in 2027–28 is the mechanism most likely to expose it.

Consensus is most likely to be right if Electrification service attach genuinely inflects from 7.3%, because that would convert equipment economics into annuity economics and would justify the re-rating on its own.

12. Fact vs. Interpretation

13. Open Questions

1. What share of Electrification revenue and orders is data centres? ABB declines to disclose it; the CFO refused the question directly on 16 July, offering only that ex-data-centre Electrification orders still grew “double digits.” This is the single most consequential non-disclosure for the thesis — the concentration risk cannot be sized without it.
2. What is the firm-versus-cancellable split of the $30.0bn backlog? Undisclosed. Also undisclosed at group level is the large-order-versus-base-order split (only US base-order growth of ~30% and Q4 2025 dollar amounts are given).
3. Is ~20% Operational EBITA structural or cyclical? Resolvable only with time and a downturn. ABB’s own 18–22% band and the ~17.5% post-add-back figure are the best available anchors.
4. Will Electrification service attach actually inflect from 7.3%? Four years of data show it essentially flat (6.5–7.3%) despite service orders growing +16%. This is the highest-value thing to monitor: it is the one development that would convert the bull case from assertion to evidence.
5. What are the Rotork synergies, quantified? Management declined twice. Without a number, the ~3.5% entry return is the only concrete fact about the deal’s economics.
6. What is the antitrust status and break-fee structure of the SoftBank Robotics sale? No jurisdiction-level clearance status or break terms are publicly disclosed, yet Rotork is explicitly funded from those proceeds.
7. What is the undescribed ~$100m provision for “unasserted legacy claims and remediations” taken in Q2 2026? Given ABB’s asbestos and compliance history, this warrants follow-up in the annual report.
8. Does Chinese competition reach ABB’s third markets? Currently unsupported by any primary source — ABB, FANUC, Yaskawa and Nidec all report the price war inside China only. The falsification test is whether Chinese vendors close the safety/NRTL certification gap on export-bound equipment.
9. Is the $25m contingency liability disclosure complete? It enumerates no individual matter at a $33bn-revenue industrial with a significant legal history.
10. Schneider Electric is absent from the comparative peer set — a genuine gap, given it is ABB’s closest global comparable.
11. No analyst Q&A is on the record for Q4 2025 or Q1 2026 (transcript-corpus gap plus paywalls) — so there is no direct management questioning on the quarter in which the $377m gain was booked.

14. What Must Be True

For the bull case

For the bear case

The single most informative datapoint over the next twelve months is the gross-margin line at a book-to-bill above 1.1. If ABB can expand gross margin from here, the pricing-power question resolves in the bulls’ favour and the multiple is defensible. If gross margin continues to erode while demand is this strong, the bear case is confirmed by the company’s own income statement, and no amount of order-book growth will offset it.

15. Source Appendix

See the accompanying source appendix for the full citation list. Principal primary sources: ABB Q2 2026 press release and results presentation (16 Jul 2026); ABB Q4 2025 press release and Financial Information booklet (29 Jan 2026); ABB Form 20-F FY2023 (the last US filing before SEC deregistration); ABB Compensation Report 2025; ABB Capital Markets Day materials (18 Nov 2025); the ABB/Rotork Rule 2.7 announcement (16 Jul 2026); SoftBank’s Robotics acquisition release (8 Oct 2025); ABB earnings-call transcripts Q1 2024 – Q2 2026; the IEA’s Electricity 2026 and Energy and AI; peer filings from Schneider Electric, Eaton and Hubbell; OSHA 29 CFR 1910.7/.303/.399 and the NRTL scope schedules; 2 CFR Part 184 and FAR 25.101; and FANUC, Yaskawa, Nidec and Mitsubishi Electric investor-relations disclosures.

APPENDIX A — Standard Diligence Questionnaire

ABB Ltd (Nasdaq Stockholm: ABB.ST) — 18 July 2026

Supplemental to the main analysis. Answers are labelled Fact / Interpretation / Assumption where the distinction matters.

General

What thoughtful questions have other investors asked about this company?

The most penetrating questions came from the Q2 2026 call and they cluster on three points. On concentration: James Moore (Rothschild) asked directly for the data-centre share of the $7.2bn Electrification order book; the CFO declined to answer, offering only that ex-data-centre Electrification orders still grew “double digits.” On the Rotork return: Alessandro (Octavian) computed a 5–10% entry ROIC on the $5.5bn and said ABB would need to “double EBIT” to justify it — and CFO Nilsson did not dispute the starting return, contesting only the required uplift. On the moat itself: Karri Rinta (SB1 Markets) asked why price was such a modest driver at record demand, drawing the most consequential answer of the call (see Business Quality below). Analysts also pressed on whether record Electrification orders reflected genuine demand or lengthening delivery windows into 2028+, and on 2027 guidance, which Wierod deferred to January.

Cyclicality & Earnings Nature

Are earnings at a cyclical high or low? [Interpretation, high confidence: a cyclical high.] Operational EBITA margin of 20.2% sits in the upper half of the 18–22% band ABB itself only raised in November 2025; ROCE of 28.4% compares with the company’s own “>20%” target. Management’s own framework therefore describes current results as above-normal. Orders grew 28% comparable to an all-time high with a 1.28 book-to-bill. Every one of those is a cycle-peak marker.

Driven by the external environment or internal actions? [Both, and the split matters.] The margin structure is substantially internal and durable — the decentralised operating model, portfolio pruning, and cost discipline took gross margin from 33.0% (FY2022) to 41.1% (FY2025). The current level of orders and volume leverage is overwhelmingly external: a US data-centre capital-expenditure boom. The honest split is that ABB earned the right to participate in this cycle through internal action, but is not creating the cycle.

How stable are revenues? Moderately. The $30.0bn backlog covers roughly ten months of revenue — real visibility, but far thinner than long-cycle peers. A book-to-bill below 1.0 would begin eroding revenue within three to four quarters. A meaningful short-cycle component sells through distribution and destocks quickly; Rockwell’s shares fell ~8% peak-to-trough on a mere destocking cycle.

Outlook for products/services? Structurally favourable and independently corroborated. Grid investment is rising ~20% to ~$550bn in 2026 and needs +50% by 2030, with >2,500 GW stalled in interconnection queues; hyperscaler capex is heading from >$400bn toward ~$725bn. The caution is that ABB’s growth is currently coming from data centres, not from the grid/utility pillar that anchors the structural story — utility orders were merely “stable.”

How big will this market be — growing, shrinking, domestic or international? Growing, and genuinely global, though ABB’s growth is presently concentrated in the Americas (+52%, USA +62%) against Europe +12%, Germany −2% and Brazil −26%. China is ~14% of revenue and is a share-defence market rather than a growth market.

Business Quality & Competitive Moat

Is the industry getting more or less competitive? [Interpretation: more, on a two-to-three-year view.] More than $2.5bn of announced Western capacity expansion — Eaton, Schneider, Siemens Energy and ABB itself — lands in 2027–28. ABB’s own capex/D&A ratio crossed above 1.0x in 2025 for the first time in seven years, the classic capital-cycle warning. Lead times (44–144 weeks) have not yet shortened, so the tightening has not yet arrived; the capacity that ends it is already funded.

How profitable is the business (ROIC, ROE)? Genuinely good, but state the range rather than the flattering point. On FY2025 NOPAT of $4,523m: ~24.7% against equity plus net debt; ~17.4% against equity plus gross debt; ABB discloses 25.3% ROCE (28.4% TTM) on its own definition. All three clear an ~8.5% WACC by a wide margin, and all three are dramatically better than FY2019’s 5.5–6.6%. The conclusion is robust to definition; the magnitude is not.

How profitable is the industry — how many competitors, what barriers to entry? A loose oligopoly. The top five hold >40% of switchgear but only ~20–25% of low voltage — too fragmented for industry-wide barriers in Greenwald’s sense. Barriers are nonetheless real and documented: OSHA NRTL certification (29 CFR 1910.303/.399/.7) with no IEC reciprocity for assemblies — IEC 61439 and UL 61439 are both absent from OSHA’s recognised list while UL 67/489/891/845/1558 are present, so international certification confers literally zero US credit; Build America Buy America domestic content at 55%, with the separate FAR schedule escalating to 75% by 2029; utility and EPC qualification; and code specification. These barriers exclude entrants but are shared equally by all incumbents — which is the best explanation for why no Western entrant of scale has succeeded in a generation.

Can the business be easily understood? Yes at the segment level, no at the consolidated level. Three businesses with a 950bp margin spread, different customers, channels and cycles, reported as one company. Valuing ABB as a single franchise obscures that ~65% of segment profit comes from Electrification.

Can it be undermined by foreign low-cost labour? Partly, and the evidence is genuinely contested. Chinese vendors are reported to undercut by 25–30% in low-voltage drives, and Japanese peers describe a severe price war — FANUC states the CNC “price war is getting fiercer every year”; Nidec’s statutory filing alleges pricing that “may spoil the market fairness.” But all of that erosion is reported inside China; FANUC, Yaskawa and Nidec describe India, SE Asia and Europe as margin sanctuaries, and Yaskawa states plainly “in India, we don’t see price competition.” ABB’s own management denies the read-across to electrification, and says China price declines have “reversed.” [Open Question, not an asserted headwind. The falsification test is whether Chinese vendors close the safety/NRTL certification gap on export-bound equipment.]

Do brands matter? Less than assumed. Published institutional specifications name three or four acceptable manufacturers — and it is nearly always the same three or four — so brand matters enormously for exclusion and barely at all for differentiation among incumbents. At the installer level, practitioner evidence indicates brand choice is driven by distributor stocking depth and same-day parts availability, not design lock-in.

What is the nature of competition? Greenwald’s shared advantages among a few — a repeated pricing prisoner’s dilemma. Incumbents collectively enjoy real barriers against outsiders while competing vigorously with each other, including for each other’s aftermarket.

Customers’ switching costs? Highly variable by segment, and the single most important finding in this report. In Automation, switching costs are genuine and documented: System 800xA ships orderable migration products for 1980s-generation controllers (MOD 300, Advant Master, INFI 90, Symphony, Melody), ABB guarantees ten years’ minimum support after a product leaves active sale, and the installed base is 35,000 DCS systems and ~100 million I/O points. In Electrification, service attach is only 7.3% and has been stuck in a 6.5–7.3% band for four years — there is very little aftermarket annuity anchoring the customer. And the aftermarket generally is contested rather than captive: Schneider publishes that “over half of our work is on competitive brands,” with a printed cross-reference table for retrofitting Eaton, Siemens, GE and Square D gear.

The decisive test: at a 1.28 book-to-bill, with industry lead times of 44–144 weeks and data-centre orders growing at triple-digit rates, ABB extracted ~2% price — and gross margin fell 50bp (140bp in Electrification). Asked why, Wierod answered: “many of the Hyperscalers are very large customers, and they have also leverage in a price negotiation.” A genuine bottleneck prices far above 2%. Buyer power offsets ABB’s supply position.

Financial Condition & Balance Sheet

Assets not fully recognised on the balance sheet? The installed base and the specification positions in institutional standards are real economic assets carried at nothing. Against that, the DCS installed base’s value is partly capitalised in goodwill already.

Off-balance-sheet liabilities? $1,944m of third-party guarantees (performance guarantees $1,926m), maturities to 2032. One legacy non-core EPC project of ~$100m whose exit timing ABB says is “difficult to assess due to legal proceedings.” Recognised regulatory/compliance/legal contingencies are just $25m (down from $72m), “none individually significant” — [Open Question: that is a small number for a $33bn-revenue industrial with ABB’s legal history, and the disclosure enumerates no individual matter.] Pension is a non-issue: defined-benefit plans are overfunded by $519m.

How conservative is the accounting? Mixed, and worth separating. Conservative/creditable: the Operational EBITA add-back wedge narrowed from $837m (2.74% of revenue) to $267m (0.80%); management excluded a $210m unrealised derivative gain from Operational EBITA, which very few managements do; cash conversion is 1.16x net income; SBC is 0.29% of revenue. Aggressive/needs adjustment: the $377m real-estate gain is booked inside Operational EBITA, the metric on which management is bonused — H1 margin is 21.8% reported against ~19.8% ex-gain, and H1 EPS grew ~2.4% rather than the headline ~14%. Recurring add-backs of ~$498m/year (~1.5% of revenue) mean the sustainable operating margin is ~17.5%, not 19.0%. And in Q2 2026 the adjusted and statutory margins moved in opposite directions (+90bp vs −100bp), which always argues for weighting the statutory measure more heavily.

How CapEx-hungry is the business? Not very, though it is rising. Capex was $1,001m in FY2025 — 3.0% of revenue and 1.23x depreciation — in line with Eaton, Schneider and Rockwell at ~3%. The 25% increase in 2025 and the crossing of capex/D&A above 1.0x is a capital-cycle signal rather than a burden.

Capital Allocation & Management

How much FCF does the business generate, and what is the philosophy? FY2025 free cash flow of $4,566m — 96% of net income; TTM $5,200m. Philosophy through 2024 was disciplined: divest, simplify, return cash. Since August 2024 it has shifted decisively toward acquisition.

Significant acquisitions recently? Yes — the largest in company history. Rotork at ~$5.5bn EV, announced 16 July 2026: 503p/share cash, ~60% premium to the three-month average, ~5.3x EV/sales and ~19.5x EV/EBITDA, closing H1 2027. Rotork is a high-quality asset (24.6% adjusted operating margin, 38.4% ROCE, 24% service). But the arithmetic is unforgiving: ~$256m of operating profit taxed at ~22% is ~$192m NOPAT on $5.5bn, a ~3.5% first-year return against an ~8.5% WACC; even fully synergised it is ~4.7%. Management declined twice to quantify synergies. Also acquired: Gamesa Electric (currently operating at a loss, costing Motion ~70bp), Siemens Wiring Accessories, Specialtrasfo, Sensorfact, Brightloop.

Buying back shares? Yes, and historically very well. ABB repurchased 218,686,689 shares for ~$6.6bn at ~$30 average in 2020–22 into a stock that has since roughly tripled — genuinely good timing. Subsequent programmes ran $1,258m (2023), $1,247m (2024) and $1,499m (2025); share count is down from ~2,133m (2019) to 1,815m, −14.8%. But the pace decelerated ~35% quarter-on-quarter in Q2 2026 to $147m as cash was retained for Rotork.

Issuing large amounts of new shares to insiders? No. SBC is $97m — 0.29% of revenue, 2.1% of FCF — with a diluted-versus-basic gap of only 4m shares. Exceptionally low.

Compensation policy of directors/management? [The most important governance finding.] The annual bonus weights Operational EBITA margin 30%, revenues 30%, FCF conversion 20%, ROCE 10%, sustainability 10%. The long-term plan weights average EPS 50% and relative TSR 30% (sustainability 20%), moving to EPS 60% / TSR 40% from 2026 — with no ROCE, no return-on-capital and no cash-flow measure at all. So return on capital carries 10% of the bonus and zero of long-term pay. A cash-funded acquisition of a high-margin asset is near-certain to raise EPS and group margin percentage — both heavily weighted — while diluting return on capital, which is barely weighted. The structure does not merely fail to discourage Rotork; it rewards it. Targets are not demanding: CEO bonus outcomes over five years were 145.0%, 120.0%, 149.0%, 120.7% and 121.8% — never below target; LTIP vesting averaged 149.8%, with the EPS threshold and maximum set only ±11.4% around target. Directors are paid fixed fees only, half in restricted shares, no options — best practice.

Motivations of management? Wierod is a 28-year ABB insider with meaningful skin in the game — roughly 1,200% of net salary in stock, ~2.4x his requirement, tightening to a gross-salary basis in 2026. This is not a promotional management team. But the entire Executive Committee and Board each own <1%; these are professional managers, well-aligned individually and modestly aligned in aggregate. Note the governance context: Cevian Capital, the activist that drove the 2019–23 disciplined-seller era, cut its stake below 3% and left the board in 2024 — precisely as the premium-buyer era began. Investor AB holds 14.4% with one board seat: influence without control, and a patient anchor, but a dispersed register imposes little discipline on a large, richly-priced acquisition.

Valuation & Market Data

Is the stock an ADR, MLP, or K-1 issuer? ABB.ST is the ordinary share on Nasdaq Stockholm (primary listing SIX Swiss, ABBN). A 1:1 ADR trades OTC as ABBNY. No MLP, no K-1. Note the three-currency structure: USD reporting, SEK quotation, CHF dividend and cost base.

Dividend policy? A single annual dividend in CHF: 0.80 (2019) rising to CHF 0.94 (2025) — five consecutive increases but only a ~2.7% CAGR, far below EPS growth, with the cash payout ratio drifting down from ~47% to ~41%. Deliberate design: a low-volatility dividend floor with the buyback as the swing instrument. Swiss withholding tax is 35% at source; the Sweden–Switzerland treaty reduces this toward 15% but only via a reclaim filing, so 20 points of the gross dividend is withheld pending an administrative process, and non-treaty or tax-exempt holders can suffer permanent leakage.

How profitable is the business? See above — ~17–25% return on capital depending on definition, against ~8.5% WACC.

Is net income diverging from cash from operations? No — this is a clean area. OCF/net income ran 1.15x, 1.19x and 1.16x in 2023–25. The single poor year, 2022 (0.52x), was a $1,602m inventory build since unwound. (One correction to a common assumption: ABB does not run negative working capital — trade NWC was $4,059m at end-2025, 13.0% of revenue.)

Valuation summary. At SEK 945.20 / ADR $98.15, market cap ~$177.8bn and EV ~$180.3bn: EV/Sales 5.04x; EV/Operational EBITA 25.1x (26.5x ex-gain); EV/EBITDA ~24.6x; EV/IFRS EBIT 27.9x; P/E 35.3x (37.4x ex-gain); FCF yield 2.93%. Against its own twelve-year history, price/sales at ~5.4x (range 1.11–4.07x) and price/book at ~7.9x (range 1.80–5.94x) are above the top of the prior range, not merely at it. P/B and P/tangible-book are near-meaningless here — tangible book is $2.94/share and Rotork takes it to approximately zero.

Risks & Downside

What factors would cause the stock to decline? In order of importance: (1) order normalisation as the data-centre capex cycle rolls over — Automation orders already fell 14% with a flat backlog; (2) margin mean-reversion from 20.2% toward the ~17.5% sustainable level, which the market is not underwriting; (3) evidence that Rotork will not earn its cost of capital, or a further premium acquisition establishing a pattern; (4) the 2027–28 capacity wave compressing industry pricing; (5) multiple compression alone, given the stock sits above the top of its own twelve-year range in a peer cohort that is uniformly at records.

Critically, these are correlated. The same cycle that produced the record orders and 28.4% ROCE is the cycle that set the currency ABB is spending on Rotork, and the industry capacity that ends the shortage lands in the same 2027–28 window in which the data-centre book converts. They go wrong together.

Risk of a catastrophic loss? Low. Net debt of 0.3x EBITDA, an overfunded pension, 96% cash conversion, a $30bn backlog and no covenant pressure make insolvency remote. Goodwill impairment is possible — intangibles are 65% of equity and Rotork takes tangible book to ~zero — but that is non-cash.

Chance of a total loss? Remote on any reasonable horizon. It is worth recording, however, that this specific company has been to the brink once: on the adjusted price series ABB fell −96.4% from its May 2000 peak to its October 2002 trough during the Combustion Engineering asbestos and liquidity crisis. “Remote” is not “impossible,” and the current rating embeds no memory of it.

Recent News & Events

Note: built from ABB’s own SIX ad-hoc releases, quarterly reports, the Rotork offer document and SoftBank’s release.

Has the business environment changed recently? Yes, materially and favourably on demand. Q2 2026 (16 July) delivered record orders of $12,042m (+28% comparable), revenue +12%, Operational EBITA margin 20.2%, backlog $30,007m (+27%), and raised FY2026 guidance to “low double-digit to low-teens” revenue growth. Note the raise was revenue-only — the margin sentence is word-for-word identical to April’s. Electrification orders passed $7bn in a quarter for the first time. US tariffs de-escalated sharply in November 2025, with the ceiling on Swiss goods cut from 39% to 15%.

Significant acquisitions? Rotork (~$5.5bn, July 2026) — see above.

Change in accounting policies? No policy change, but two significant presentational changes: Robotics & Discrete Automation moved to discontinued operations from Q4 2025 (with only FY2024 restated, breaking multi-year comparability), and the third business area was renamed “Automation” with Machine Automation folded in.

Recent changes — new markets, facilities, management? CEO transition to Morten Wierod (August 2024) and CFO transition to Christian Nilsson (February 2026) — note the CMD targets anchoring the equity story were set by the departing CFO, and the largest acquisition in company history is being executed by a CFO five months into the job. Capital Markets Day (18 November 2025) raised targets: Operational EBITA margin 18–22% (from 16–19%), ROCE >20% (from >18%), FCF conversion >95%. Facilities: ~$230m of US capacity announced in 2025 (Selmer TN, Senatobia MS, Richmond VA, Arecibo PR, Pinetops NC) on top of ~$500m in 2022–24, plus ~$200m into European medium-voltage capacity (May 2026). Portfolio: Robotics sold to SoftBank for $5.375bn (October 2025, reversing an April 2025 spin-off plan, closing H2 2026); E-mobility pruned further with 60% of ChargeDot sold (December 2025), narrowing losses from $445m (FY2024) to $205m (FY2025).

APPENDIX B — Source Appendix

ABB Ltd (Nasdaq Stockholm: ABB.ST) — 18 July 2026

All sources accessed 18 July 2026 unless otherwise noted. Primary sources are listed first within each category.

1. ABB company filings, reports and releases (primary)

2. Transcripts (primary)

3. Rotork and SoftBank transaction documents

4. Peer and competitor filings (primary)

Note on citation form: FANUC and Yaskawa publish English Q&A summaries in which individual answers are not attributed to a named speaker. Citations are to the document, not to an individual.

5. Regulatory and standards sources (primary)

6. Industry data and market context

7. Market, price and factor data

8. Sourcing limitations disclosed

1. The SEC corpus is unavailable — ABB deregistered from SEC reporting in 2024; only filings through the FY2023 Form 20-F exist on EDGAR. FY2024–FY2025 figures come from ABB’s own Financial Reports and quarterly Financial Information booklets.
2. Items explicitly not found and therefore not asserted anywhere in this article: ABB’s data-centre revenue or order share; the firm-versus-cancellable backlog split; division-level revenue within the Automation business area; quantified Rotork synergies; any ABB group service-share target; and any quantified specification-position-to-order conversion rate.
3. No analyst Q&A is publicly available for the Q4 2025 and Q1 2026 calls, so there is no direct management questioning on the record for the quarter in which the $377m real-estate gain was booked.
4. Third-party aggregated financial data was reconciled against ABB’s own reporting throughout; where the two disagreed, the company’s figures were used and the discrepancy noted (see section 6.1).