ABB Ltd

Refresh delta (since 2026-07-06)

What changed

1. Q2 2026 was a beat-and-raise, and the stock fell anyway. Orders were $12,042m (+30%, +28% comparable), revenue $9,475m (+12% comparable) and Operational EBITA $1,925m at a 20.2% margin (+90bp). Book-to-billBook-to-billNew orders divided by orders filled. Above 1 means the backlog is growing; below 1 means the company is working through it faster than it is replacing it. was 1.27 and backlog a record $30,007m (+28% comparable). The FY2026 guide was raised for the second time: 6–9% comparable growth in January, high-single-digit to low-double-digit in April, low-double-digit to low-teens in July. ABBN closed −5.9% on the day, from CHF 83.18 to CHF 78.26. Conflict: Investing.com's call-transcript headline says "shares slip 2.8%", which was probably an intraday figure. The close-to-close number is −5.9%.
2. ABB is buying Rotork for about $5.5bn cash (503p a share, recommended; announced 2026-07-16). That is about 19.5x Rotork's 2025 EV/EBITDA and about 5.3x EV/sales. Management says synergies take it to the "mid-teens", it will be EPS-accretive in year two, and it closes in H1 2027. Rotork shareholders approved it on 2026-09-03: 96.01% of votes cast, 48.10% of eligible shares voting. It is funded by the ~$4.8bn of net Robotics proceeds. The prior dossier said ~$5.3bn, which was the October 2025 headline. Five bolt-ons were also announced or closed: Netcontrol, IPEC, Premium Power, Specialtrasfo, Høglund, plus Advantics for silicon-carbide DC technology [interim-report-2026-q2.md; transcripts/2026-q2.md].
3. The Robotics sale to SoftBank has still not closed. The Q2 Financial Information says "expected to be completed in the second half of 2026, subject to regulatory approvals… and the completion of certain legal entity reorganizations". No completion announcement turned up in searches as of 2026-10-05. This makes it a live Q3/Q4 catalyst, bringing a ~$2.4bn pre-tax book gain.
4. New items for the forensic file. These are a US EPA Superfund potentially-responsible-party (PRP) claim, about $100m of legacy-claim provisions, a loss-making acquisition and a project carved out of guidance because of litigation:

• EPA Superfund: ABB was named a potentially responsible party. It booked a $31m provision against an estimated loss range of $31–204m, with a later remediation phase "not presently estimable".
• Legacy claims: about $100m of "provisions for unasserted legacy claims and remediations" hit Q2 income from operations [interim-report-2026-q2.md, "Earnings"].
• Gamesa Electric: the acquisition is loss-making and cut about 70bp from Motion's margin. Management expects it to stay dilutiveDilutionIssuing new shares, so each existing share owns a smaller slice of the same company. through 2026 [interim-report-2026-q2.md; transcripts/2026-q2.md].
• Excluded project: a ~$100m non-core project is excluded from the 2026 guidance framework because its exit timing "is difficult to assess due to legal proceedings etc." [interim-report-2026-q2.md, "Additional 2026 guidance", fn 1].

1. Corrections to the 2026-07-06 dossier (that run was web-only). Primary documents now contradict several of its numbers:

• FY2025 business-area (BA) margins were wrong. The real figures are Electrification 23.5%, Motion 19.4% and Automation 14.0%. The 22.6% / 18.3% / 13.9% it quoted are the Q4 2025 margins.
• The "19.0% all-in vs 17.6% continuing" reading was wrong. 19.0% is the FY2025 continuing-operations margin and 17.6% is the Q4 2025 margin [interim-report-2025-q4.md key figures].
• The "2026 guide conflict" was not a conflict. The 6–9% guide (January) and the high-single-digit to low-double-digit guide (April) were successive guides; July raised it again.
• The valuation row was wrong. The prior dossier showed ~23.5x forward P/E and ~$146bn market capMarket capitalisationThe share price multiplied by the number of shares. What the market says the equity is worth.. At CHF 81.70 and 1,815m shares, market cap is ~$178.5bn and the trailing P/E is ~36x on continuing EPS (see Lens 7).
• The Kusile penalty was overstated. The FY2023 20-F records ~$325m provisioned and $313m expensed in 2022 for regulatory penalties, plus $104m paid to Eskom in 2020. The prior "$460m to US authorities" is not supported by the filing.

1. The Q3 2026 print is Tuesday 20 October 2026, not Thursday 15 October (release time not stated; ~07:00 CEST ). Source: ABB's financial calendar. ABB also held a webcast on DC technology for high-power data centres on 2026-09-24.

What held. The structural thesis still stands. Electrification is the engine (Q2 margin 24.9%, a record), demand is broad (orders excluding data centres still grew double digits in Electrification), management sees no pre-buying and lead times are unchanged [transcripts/2026-q2.md]. Lens 2 (supply chain) and Lens 3 (moat) are unchanged in substance.

---

Phase A — Understand the business

Company Overview — carried (see the previous dossier), with corrections

The business description stands. These corrected FY2025 continuing-operations figures replace the prior table:

FY2025 group results: orders $36,765m (+15% comparable), income from operations $6,047m, basic EPS $2.59 (continuing $2.50), free cash flowFree cash flowCash left after paying to run and maintain the business. Unlike profit, it is hard to flatter with accounting choices. (FCF) $4,566m, return on capital employed (ROCE) 25.3%, dividend CHF 0.94 [interim-report-2025-q4.md]. At the November 2025 Capital Markets Day, the target bands became an Operational EBITA margin of 18–22% with per-BA targets and ROCE above 20% [interim-report-2025-q4.md, "Significant events"; our figures].

Supply Chain — carried (the previous dossier)

Two additions:

• ABB is investing about $200m over three years in medium-voltage (MV) manufacturing capacity in Europe [our figures; interim-report-2026-q2.md].
• Management says capacity coming online is what lets it take more orders. In the CFO's words, it is "more that we're getting more capacity online" [transcripts/2026-q2.md, Q&A answer to Citi].

Moats — carried, with one new datapoint

ABB launched the HiPerGuard 34.5kV MV UPS, which connects data centres directly to the grid without a voltage-conversion step [interim-report-2026-q2.md CEO summary]. It reinforces the MV moat. Management also said it has "zero order in the backlog for 800 volt DC data centers. Nobody has". It does not expect that architecture to earn revenue before late 2027 or 2028 [transcripts/2026-q2.md, CEO]. In the meantime ABB is building the portfolio, including the Advantics silicon-carbide acquisition. So DC data-centre architecture is an option, not current revenue.

Segments — re-run (segment mix moved)

All figures below are. Operating income is the segment's Operational EBITA.

• Electrification now carries the group. It was 55% of Q2 group revenue (5.200/9.475) and about 63% of segment Operational EBITA ($1,301m of $2,052m segment total). Its book-to-bill was 1.39, the sixth straight quarter above 1 [transcripts/2026-q2.md, CFO]. The weak spot is the gross margin, which fell 140bp. About half of that is unrealised FX/commodity derivatives; the rest is a price/cost gap that management expects to be "at least neutral" for the full year.
• Motion's margin is going backwards. The 130bp decline breaks down as Gamesa (−70bp), High Power inefficiencies (to be resolved in H2) and Traction production timing [interim-report-2026-q2.md].
• Automation's order drop is a comparison effect. Orders fell −14% against a $600m single booking in Q2 2025. Its 15.4% margin includes about 70bp from a one-off provision release [transcripts/2026-q2.md, CFO]. Adjusted, it is about 14.7%, which is still an improvement on 14.2%. Rotork would add about 12% to Automation's revenue and would have lifted its 2025 Operational EBITDA margin to 15.2% pro forma [transcripts/2026-q2.md].
• By geography, the US is driving growth. Q2 orders: Americas +52% comparable (US +62%; base orders about +30% excluding large bookings), Europe +12% (Germany down), AMEA +12% (China +10% comparable) [interim-report-2026-q2.md; transcripts/2026-q2.md].

---

Phase B — Measure performance

Earnings Result — Q2 2026 (reported 2026-07-16)

All figures are.

Headline numbers

• Revenue $9,475m: +14% reported, +12% comparable. The Q2 guide (given 2026-04-22) was high-single-digit to low-double-digit, so 12% landed at the top of the company's own range; the CFO called it delivering "on our guidance" [our figures; transcripts/2026-q2.md].
• Gross margin 40.0%, down 50bp, mainly unrealised FX/commodity derivatives.
• Income from operations $1,585m, a 16.7% margin, down 100bp. Hits included ~$60m derivative mark-to-market, ~$100m legacy claims and remediation, and ~$30m equity fair value.
• Operational EBITA $1,925m, a 20.2% margin (+90bp). Pricing contributed "close to 2%".

The GAAP/non-GAAP gap is widening. In Q2 2026, Operational EBITA ran $340m ahead of income from operations, against $132m in Q2 2025.

Earnings, cash and balance sheet

• EPS: basic $0.68 (+8%), of which continuing $0.66 vs $0.59. H1 basic EPS was $1.41.
• Cash: operating cash flow from continuing operations $1,303m (+34%). FCF $881m (+4%). H1 FCF $2,131m includes about $425m from the Q1 real-estate sale [transcripts/2026-q2.md, CFO].
• Balance-sheet flags:
• Inventories rose $725m in H1 and receivables $538m. They were funded by payables (+$709m), so trade net working capital stayed at 11.9% of revenue, below 13.8% a year earlier.
• Short-term debt rose to $1,610m from $475m at December 2025 as maturities came current.
• Net debt was $2,320m, or 0.3x net debt/EBITDA.
• Treasury stock fell from 26m to 8m shares after the cancellation of 20.7m shares.
• Discontinued operations: Robotics revenue in Q2 was $576m and income from operations fell to $47m from $102m. Expenses rose to $143m from $105m, which looks like separation cost [financial-information-2026-q2.md, Note 3].

Q1 2026 for context [our figures 2026-Q1]: revenue $8,734m (+11% comparable), Operational EBITA 23.5% including a $377m real-estate gain, FCF $1,250m. Excluding the gain, Q1 Operational EBITA margin was about 19.1%.

Market reaction: −5.9% on the day (see delta item 1), followed by −4.3% on 2026-07-07, before the print. Most likely the market read four things together: record orders that were already priced in, falling gross margins, Motion's slide, and paying ~19.5x for Rotork.

Earnings Calls — re-run (five transcripts now on the shelf: 2025-Q2 → 2026-Q2)

Mention counts per call, Q2-25 → Q2-26:

What the counts show

• Tariffs and uncertainty dropped out of the vocabulary. In Q3-25, management said tariff impacts were "limited to the tens of millions" thanks to its local-for-local footprint [transcripts/2025-q3.md].
• Pricing became a recurring topic as input costs rose. From Q2-26: "still have a bit of a gap in price versus input cost to cover" [transcripts/2026-q2.md].
• M&A replaced buybacks in the capital-allocation conversation. Buybacks were mentioned 0 times in the Q2-26 call, against 4 in each of Q4-25 and Q1-26.
• The tone is confident and increasingly acquisitive. Management repeated "No pre-buys" and "lead times… the same" [transcripts/2026-q2.md], and in Q1-26 said "we are growing probably a bit more than the market" [transcripts/2026-q1.md].

What changed: the story moved from margin self-help to growth plus M&A. That is a classic later-cycle shift, and it warrants watching capital discipline (Rotork at 19.5x EBITDA).

Comps — ABB row re-derived from primary data; peer rows carried as stale

ABB as of 2026-10-05

• Price and market cap: CHF 81.70. At USD/CHF 0.8306 that is about $98.36, which matches the ABBNY ADR at $98.58. Market cap is about $178.5bn.
• EV: about $181.4bn unadjusted. That EVEnterprise valueWhat it would cost to buy the whole company: its market value plus its debt, minus the cash you would get with it. Often very different from market cap. still contains held-for-sale Robotics while EBITDA below is continuing-only; continuing-basis EV ≈ $176.6bn.
• Trailing multiples:
• P/E about 35.9x on LTM continuing basic EPS of $2.74 (Q3-25 0.64 derived [FY 2.50 − 0.57 − 0.59 − 0.70] + Q4-25 0.70 + Q1-26 0.74 + Q2-26 0.66; financial-information-2025-q4/2026-q1/2026-q2.md income statements). On total basic EPS incl. discontinued Robotics ($2.77 = 0.66 + 0.70 + 0.73 + 0.68) it is ~35.5x. The continuing basis matches the forward P/Es in Lens 11.
• EV/EBITDA about 24.1x on continuing-basis EV ($176.6bn) and LTM continuing EBITDA of $7,327m; about 25.4x excluding the $377m real-estate gain. (Unadjusted EV: 24.8x / 26.1x.)
• EV/Sales about 5.1x on LTM revenue of $35,752m.
• EV/Operational EBITA about 25.2x, or about 26.6x excluding the gain.
• Forward P/E about 34x FY26E / 30x FY27E on our estimates (Lens 11). The prior dossier's "~23.5x forward P/E" does not reconcile with ABB's own EPS and is withdrawn.
• Dividend yield about 1.2%.
• Consensus and price target: MarketBeat shows a Hold consensus (13 analysts: 2 Strong Buy, 1 Buy, 10 Hold) and no consensus PTConsensus price targetThe average of what published analysts think the share price should be. An opinion poll, not a forecast.. A search snippet attributed a ~$101.70 average PT on 26 analysts to MarketScreener, but it was not verified, so PT is n/a.

Peers: the Schneider, Siemens, Eaton, Emerson and Rockwell multiples in the previous dossier were aggregator snapshots from mid-2026. They were not re-sourced this run, so treat them as stale. The model.xlsx Comps sheet is scaffolded for sourced peer multiples. The 5-year average ROE per peer is still n/a.

Stock-Price Catalysts — re-run

Selected large moves since March 2025 (weekly closes over 7% plus notable daily closes, incl. one −4.1% day; earlier >7% weeks — 2022-01 −7.3%, 2022-06 −9.5%, 2023-10 −10.3%, 2024-07 −7.3%, 2024-09 −7.1%, 2025-01 −7.5%, 2025-02 −7.6% — are omitted here):

Pattern: the market reacts to earnings, guidance and M&A, and to macro or geopolitical shocks. The new feature is that a record-orders print sold off. Expectations, not fundamentals, are now the swing factor.

THE NEXT CATALYST — Q3 2026 results, Tuesday 20 October 2026 (time not stated; ~07:00 CEST ). Date confirmed in ABB's own calendar; not Thursday 15 October. What to watch, in order of information value:

1. Comparable revenue growth against the "low- to mid-teens" guide, roughly 11–15% [our figures 2026-07-16]. The Q3 2025 base is $8,491m [our figures 2025-Q3, derived]. Below 11% would be ABB's first miss of its own quarterly guide this year.
2. Operational EBITA margin above Q2's 20.2%. Management guided a sequential improvement. Q3 2025 was about 19.2%.

• Electrification: management guided growth "at least similar" to Q2 (19% comparable) and a margin above 24.9%.
• Motion: margin "similar to Q2" (18.5%).
• Automation: margin up year on year [transcripts/2026-q2.md, CFO].
• Watch for any widening of the GAAP gap beyond Operational EBITA (Q2 gap: $340m).

1. Electrification orders and book-to-bill after +58%. Whether data-centre orders keep a triple-digit pace, and whether management still says "no pre-buy". A book-to-bill below 1 in Electrification would break a six-quarter run.
2. Price/cost and gross margin. Has the gap closed? Q2 group gross margin was 40.0%.
3. Robotics close. Whether the SoftBank deal has closed or slipped into 2027; the size of the ~$2.4bn pre-tax gain; and the $4.8bn redeployment into Rotork.
4. Whether FY2026 guidance is raised a third time (currently low-double-digit to low-teens), plus any first comment on 2027.
5. Legacy provisions. Any growth in the EPA Superfund matter (range up to $204m) or in the legacy-claims line.

Other dated catalysts

• Rotork scheme: court sanction and regulatory approvals, effective H1 2027.
• Q4 2026 results: 28 Jan 2027.
• Q1 2027 results: 20 Apr 2027 [interim-report-2026-q2.md calendar].

---

Phase C — Judge people & books

Management — carried (the previous dossier)

Wierod and Voser are unchanged. The capital-allocation posture has shifted: a $5.5bn acquisition at 19.5x EBITDA is the largest deal of Wierod's tenure, and the buyback ran at only ~$147m in Q2 against a $2bn programme [interim-report-2026-q2.md]. Weekly buyback notices continue through September. Capital is being redirected from buybacks to M&A. Whether that discipline holds is the open question for Lens 14.

Forensic Red Flags — re-run (new filings and provisions on the shelf)

Regulatory findings

• SEC LR/AAER (automated): regulatory/regulatory-findings.md again reports "no CIK", and the script skipped because the file was under 7 days old. This is a false negative: ABB's CIK is 0001091587. A direct EDGAR full-text search for LR/AAER on "ABB Ltd" since 2021 returned 0 hits; a search for "ABB Kusile" returned 14 hits, all ABB's own 6-K/20-F disclosures.
• Kusile (the material historical matter), quoted from the FY2023 20-F Item 8: "In December 2022, ABB settled with the SEC and DoJ as well as the authorities in South Africa and Switzerland. The matter is still pending with the authorities in Germany…".
• Cost: a provision of about $325m in Q3 2022, about $315m paid in 2022, and $104m paid to Eskom in December 2020.
• Ongoing obligation: a Deferred Prosecution Agreement with the DOJ, which was a CEO incentive goal in 2023 [same filing].
• The prior dossier's "~$460m to US authorities" is not supported by the primary filing and should be read as about $315–325m in penalties plus the $104m Eskom settlement.
• Current US GAAP disclosure (Q2 2026): only boilerplate under "Regulatory, Compliance and Legal". There is no named matter, and the company "does not expect… a material adverse effect" [financial-information-2026-q2.md, Note 11]. Since ABB left the SEC there is no 20-F Item 3. The Item 8 equivalent now lives in the annual report, which is not yet on the shelf.
• Non-SEC (new): the EPA Superfund PRP on a site tied to a business acquired in 2012. Loss range $31–204m; $31m booked in Q2 2026 [financial-information-2026-q2.md, Note 11]. The upper end is about 6.6x the amount booked.

Accounting-quality risks

1. The gap between Operational EBITA and GAAP is widening. Q2 2026's $340m gap included ~$100m of "legacy claims and remediations", ~$60m FX/commodity mark-to-market and ~$30m equity fair value. "Certain other non-operational items" is management's discretionary bucket [definition: filings/20-f-2023-q4.md]. Watch whether it recurs.
2. 2026 profit is flattered by one-offs. In H1, the $377m real-estate gain sat inside Operational EBITA, and FCF got about $425m from the sale. In H2 the ~$2.4bn Robotics gain is due, booked below the line in discontinued operations. Reported FY2026 EPS will overstate the run-rate twice over.
3. Inside segments: a one-off provision release (~$15m, about 70bp) supported Automation's Q2 margin. Gamesa losses depress Motion.
4. Off-balance-sheet guarantees of $1,611m (performance $1,594m), down from $1,944m. These include $657m tied to the divested cables business [financial-information-2026-q2.md, Guarantees].
5. A non-core project (~$100m) is excluded from guidance because its exit timing is hard to assess "due to legal proceedings etc."

Cash vs earnings: cash conversion is still healthy. H1 operating cash flow from continuing operations was $2,315m against $2,569m of continuing net income. Working capital stayed disciplined (trade net working capital 11.9% of revenue). There are no cash-earnings divergence flags beyond the real-estate timing.

---

Phase D — Project & stress-test

Forward Projection — re-run

Model status (read before citing): model.xlsx was rebuilt today and has no computed values; it has never been opened. Its seeds are not usable as-is:

• The revenue-growth seeds are an artifact. FY+1 base is seeded at 100% (bear 47%, bull 118%) because the shelf only has 6 quarters, so the trailing-year comparison is half-empty.
• Tax rate is seeded at 15%; ABB's actual effective rate is 25.5% [interim-report-2026-q2.md].
• Pre-tax cost of debt is seeded at 1%.
• Share price is blank.
• The opening balance sheet is sourced from our figures (2026-Q2): Opening balance sheet sourced: yes. Held-for-sale assets of $3,931m and liabilities of $1,238m are excluded from the current-asset/liability lines, so the BS will not tie to total assets without a plug.

Connor should set growth to about 10–12%, tax to about 25% and the price before trusting any output. The projection below is prose.

Inputs:

• H1 2026 continuing basic EPS: $1.39.
• FY2026 guide: low-double-digit to low-teens comparable growth; margin up year on year excluding the real-estate gain.
• Q3 guide: low- to mid-teens growth; margin up sequentially.
• Effective tax ~25%; ~1,815m shares; buyback running slowly.

FY2026E

• Q3E: revenue $8,491m × ~1.14 (13% comparable + FX) ≈ $9.7bn. Operational EBITA margin ~20.7% ≈ $2.0bn. Less ~$170m of acquisition amortisation, restructuring and other items gives income from operations ≈ $1.83bn. Add ~$30m net finance income, tax at 25%, less NCI ≈ $1.38bn net income → ~$0.76.
• Q4E: $9,052m × 1.12 ≈ $10.1bn. Margin ~18.8% (Q4 is seasonally weakest; Q4-25 was 17.6%) ≈ $1.9bn. Income from operations ≈ $1.70bn; net income ≈ $1.28bn → ~$0.71.
• FY2026E continuing basic EPS ≈ $2.86. That is 1.39 + 0.76 + 0.71, and it includes about $0.16 of real-estate gain. Excluding the gain ≈ $2.70.
• Reported FY2026 EPS will be roughly $3.9–4.2 if Robotics closes in 2026, after a ~$2.4bn pre-tax gain at an unknown tax rate (≈ $1.0–1.3 a share). Exclude it.

FY2027E and FY2028E

• FY2027E ≈ $3.25 (base). Revenue +10%, supported by a +28% backlog, with Rotork adding ~3% for a part-year from H1 2027. Margin +60bp. Rotork is roughly EPS-neutral in year one, since its deal costs offset the lost interest on the $4.8bn. Share count −0.5%.
• FY2028E ≈ $3.60 (base). Revenue +8%, margin +50bp, Rotork accretive.
• Bear FY2027E ≈ $2.70: data-centre digestion, Electrification book-to-bill below 1, price/cost gap persists.
• Bull FY2027E ≈ $3.60: guide beats continue, Electrification margin reaches 26%, Rotork synergies arrive early.

At $98.36: ~36x FY26E (ex-gain) / ~30x FY27E base / ~27x FY28E.

Brier forecast: not logged (unattended). The candidate is: "ABB Q3 2026 comparable revenue growth ≥ 11% (bottom of low-to-mid-teens guide)", p ≈ 0.75, resolves 2026-10-20.

Consensus EPS by year: n/a. MarketBeat shows no estimates.

Bull vs Bear — re-run

Bull.

• ABB keeps beating its own guide. It raised FY2026 twice; Q2 comparable growth of 12% landed at the top of the high-single-digit to low-double-digit guide.
• Electrification is a 25%-margin franchise with a $13.7bn backlog (+59%) and double-digit growth even excluding data centres. There is a pricing tailwind ("sequential acceleration in pricing").
• The balance sheet is nearly unlevered (0.3x net debt/EBITDA) even before ~$4.8bn of Robotics cash arrives.
• Rotork completes the Sense–Control–Act loop and lifts Automation's margin to ~15.2% pro forma.
• 800V DC and MV-UPS are the next growth legs from 2027–2028.
• In this view, the July sell-off was a de-risking: the stock is 8% below its June peak (CHF 88.56) while estimates have risen.

Bear.

1. The multiple is the risk, and it is higher than we thought. ABB trades at ~30x FY27E and ~24x trailing EV/EBITDA (continuing basis), about 1.5x its long-run EV/EBITDA median of ~17x [prior dossier, gurufocus — not re-sourced]. With ~63% of segment profit from Electrification, a data-centre digestion would hit both orders and the multiple at once.
2. Margins under the surface are not improving as fast as the headline.

• Gross margin is down year on year for two straight quarters, after rising in Q4-25: Q1-26 −290bp, Q2-26 −50bp [our figures].
• The GAAP-to-Operational gap is widening ($340m in Q2).
• Motion's margin is falling.
• The Q1 headline relied on a $377m property gain.

1. Capital allocation has turned to M&A at a peak. ABB is paying ~19.5x EBITDA for Rotork while buybacks have slowed to ~$147m a quarter.

Pre-mortem (it is April 2028 and the thesis broke). Hyperscalers paused in H2 2027 and Electrification book-to-bill fell to 0.9. The price/cost gap reopened as copper spiked, and Rotork's process-industry exposure (oil & gas, chemicals) softened with the cycle. FY2027 EPS came in at about $2.80 against $3.25 hoped for. The multiple compressed from ~30x to ~22x, and the stock fell about 35%, roughly back to CHF 55, even though revenue never declined.

Contrarian view. The market sold the best quarter in ABB's history. That says positioning is full. The contrarian long case therefore needs evidence that margins, not orders, are the next leg. The contrarian short case is the gross-margin trend that headline Operational EBITA hides.

Devil's Advocate (short-seller) — re-run

• Concentration in disguise. On revenue alone, data centres look like a minor share. But Electrification is ~63% of segment profit, data-centre orders are growing triple digits, and US orders are up 62%. The marginal dollar of growth and margin is data-centre and US. If that cycle turns, the profit effect is far larger than data centres' revenue share suggests.
• Earnings quality is slipping. In H1 2026: a $377m property gain inside Operational EBITA; $100m of legacy claims kept outside it; an EPA exposure of up to $204m against $31m booked; a provision release supporting Automation; and a ~$100m project excluded from guidance because its exit timing is clouded by legal proceedings. None of it is large on its own. Together it is a pattern of the adjusted metric doing more work every quarter.
• The Rotork price is a bet that synergies arrive. At 19.5x EBITDA, about 5.3x sales, and "majority on the commercial and growth side" synergies, the deal relies on revenue synergies, the hardest kind to deliver. An analyst on the call pushed on returns, and management deflected [transcripts/2026-q2.md].
• What has to hold at today's price. Comparable growth of at least 10% through 2027; Electrification margin at 25% or better; the price/cost gap closed; Robotics closing on time; Rotork at least neutral. If growth disappoints by 20–30% (to about 7–8%) and the multiple de-rates to a 22–24x P/E, the downside is about 25–35%.
• The single permanent-impairment scenario is still a re-rating, not a business failure. The business is durable; the price is not.

Management Questions — carried, with three additions at the top

New

1. Excluding the $377m property gain and the H1 one-offs, what is the run-rate Operational EBITA margin, and why does the gap to GAAP income from operations keep widening ($340m in Q2)?
2. Rotork at ~19.5x EBITDA: what share of the synergies that get you to "mid-teens" is cost versus revenue, by when, and what ROIC hurdle did the board approve?
3. What is the realistic upper bound on the EPA Superfund exposure ($31–204m), and is there a second remediation phase?

Carried: the 15 questions in the previous dossier. Q6 on the use of Robotics proceeds is now answered (redeployed into Rotork). Q1–Q2 on data-centre concentration remain first in information value.

---