Cummins

Phase A — Understand the business

Lens 1 · Company Overview

Cummins Inc. (NYSE: CMI), founded 1919 in Columbus, Indiana as one of the first diesel-engine makers, is "a global power leader" selling engines, integrated power-generation systems, critical drivetrain components, and zero-emissions powertrains into roughly 190 countries through ~640 wholly-owned/JV/independent distributor locations and >13,000 certified dealer locations. It sells to OEMs, distributors, dealers and end customers, competing on "performance, price, total cost of ownership, fuel economy, emissions compliance, speed of delivery, quality and customer support".

Five reportable segments:

• Engine — diesel/natural-gas engines for on- and off-highway (trucks, buses, light-duty pickups, construction, ag).
• Components — aftertreatment, turbochargers, fuel systems, controls, transmissions, axles, brakes (the Meritor acquisition, ~$3.7B in 2022, built the axle/brake scale).
• Distribution — the service-and-parts network plus power-generation product distribution; the recurring-revenue layer.
• Power Systems — high-horsepower engines and integrated power-generation systems for "data center, consumer, commercial, industrial, health care, prime rental fleet and defense applications". This is the data-center franchise.
• Accelera — battery and electric/fuel-cell powertrains, electrolyzers (the decarbonization bet — see Lens 4/10, now in retreat).

The business model is razor-and-blade across an installed base: sell the engine/genset (Engine, Power Systems), then monetize a multi-decade tail of parts, service and overhauls through Distribution. Contract structure is mostly transactional OEM/dealer sell-through rather than take-or-pay, but the data-center genset book now carries multi-quarter visibility: management cites "data center products extending out six to eight quarters" of order coverage, and large-genset backlog reportedly runs into late-2026/early-2027. Strategy banner is "Destination Zero" (decarbonize the installed base over time while keeping the diesel/gas franchise profitable).

Lens 2 · Supply Chain

Upstream → Cummins → end customer, with named stakeholders:

• Upstream inputs: steel, aluminium and castings; precision machined components; semiconductors/electronic control modules (engine controls, the SCR/aftertreatment electronics implicated in the emissions case); fuel-cell stacks and electrolyzer materials (Accelera); battery cells — sourced via the Amplify Cell Technologies JV with Daimler Truck, PACCAR and EVE Energy, building a 21 GWh plant in Marshall County, Mississippi (production targeted 2027). Cummins runs supply-chain financing programs (≤$574M capacity, $153M used at YE25) and a Wells Fargo A/R sales program (≤$500M, unused in 2025) — working-capital levers, not distress signals.
• The company: designs and manufactures its own engines and assembles large gensets in-house — it controls the two scarcest links (the high-horsepower engine and the integration). Capacity itself is the chokepoint: an announced $150M expansion of the Fridley, Minnesota plant (Feb 2026) to lift QSK95 output ~30%, explicitly to address ~18-month lead times on high-horsepower units.
• Downstream: OEMs (Stellantis/RAM for light-duty engines; truck OEMs incl. PACCAR, Daimler; off-highway OEMs); the ~640 distributor + >13,000 dealer network; and the new buyer cohort — hyperscale and colocation data-center operators (Amazon, Microsoft, Google and the colo builders behind them) buying standby/prime gensets.

Chokepoints / single-source dependencies: (1) engine + genset capacity is the binding constraint — hence the lead times and the Fridley expansion; (2) high-horsepower engine castings and large-bore machining are not easily second-sourced; (3) the distribution/service network is itself a hard-to-replicate asset (see Lens 3). This lens is concrete: the chain runs castings/electronics/cells → Cummins-built engine → in-house genset integration → distributor → hyperscaler.

Lens 3 · Competitive Advantages (moats)

1. Vertical integration in the scarce link. In data-center gensets, the engine is the bottleneck, and Cummins owns engine design, manufacture and genset integration — most rivals integrate someone else's engine. That is the moat that lets the QSK95/high-HP book run 6-8 quarters out.
2. Installed base + service network. ~640 distributor and >13,000 dealer locations across ~190 countries create a multi-decade parts/service annuity (the Distribution segment) and high switching costs — a data center or fleet buys the brand it can get serviced anywhere.
3. Brand & emissions know-how. "Cummins" is a category brand in heavy diesel; 106 years of combustion and aftertreatment IP is a real barrier — ironically the same control-system depth that produced the defeat-device liability (Lens 10) is the depth rivals can't shortcut.
4. Duopoly economics in hyperscale backup. Caterpillar #1, Cummins clear #2; together the two hold the majority of hyperscale backup capacity. In a capacity-constrained, reliability-critical purchase, a credible #2 with its own engines has pricing power — visible in Power Systems EBITDA margin rising to 29.5% in Q1 2026 from 23.6%.

Bargaining power: strong over distributors/dealers (they need the brand and parts); strengthening over data-center buyers while capacity is short (lead times = pricing power); weaker over truck OEMs in the cyclical trough, where Cummins is one of several engine suppliers and the buyer holds volume leverage.

Lens 4 · Segments

FY2025 vs FY2024 vs FY2023, all ``. (Segment "total sales" include intersegment; consolidated external net sales = $33,670M / $34,102M / $34,065M — essentially flat for three years on the surface, masking a violent internal mix-shift.)

The trend is the thesis: Power Systems EBITDA +44% and Distribution +31% — both driven by "increased power generation volumes in North America… especially data center and commercial markets" — fully offset Engine −16% and Components −12% from "lower demand in on-highway commercial truck markets." Accelera is a widening, deliberately-shrinking loss (−$896M; see Lens 10). The accelerating leg (power-gen, ~22-30% margins) is replacing the decelerating leg (trucks, ~10-14% margins), which is why consolidated revenue can be flat while the earnings power and multiple re-rate.

Geographically, FY2025 international strength (China, Q1 2026 intl +16%) is offsetting North American truck weakness (Q1 2026 NA −6%).

---

Phase B — Measure performance

Lens 5 · Earnings Result (latest print: Q1 2026, qtr ended 2026-03-31)

• Revenue $8,398M, +3% YoY (vs $8,174M); +$224M, NA −6% / international +16% (China).
• GAAP net income $654M, diluted EPS $4.71 (vs $824M / $5.96 PY). The decline is non-operating: a $199M charge on the low-pressure fuel-cell business sale cut EPS by $1.44.
• Adjusted EPS $6.15, beat consensus $5.63 by $0.52. This is the figure the tape trades on — a clean beat, not the optical GAAP miss.
• Margins: Q1 EBITDA $1.29B = 15.4% GAAP; 17.7% ex-charges vs 17.9% PY (essentially held).
• Segment drivers: Power Systems $2.0B +19%, EBITDA margin 29.5% (from 23.6%) — the standout; Distribution $3.1B +7% (14.2%); Engine $2.7B −4%, EBITDA margin 10.4% from 16.5% (truck trough); Components $2.5B −5% (13.3%).
• Guidance raised: FY2026 revenue to +8% to +11% (from +3-8%); EBITDA to 17.75%-18.50% (from 17.0-18.0%).
• Balance-sheet flags — clean: FY2025 operating cash flow $3,621M (vs $1,487M in 2024, which was depressed by the $1.9B settlement payment), capex $1,235M → FCF ≈ $2,386M. Net leverage 0.22x vs a 0.65x covenant; $4.0B of undrawn revolvers, only $353M commercial paper out. A/R rose ($612M working-capital use) — worth watching but not alarming given the volume ramp.
• Market reaction: stock near 52-week highs into and after the print ($704.86 on 2026-06-30 vs a $322.45 12-month low) — the market has clearly priced the mix-shift, so a beat-and-raise was largely expected rather than a surprise.

Unusual vs its own history: the Engine margin compression (10.4%) is a genuine cyclical trough, and the Power Systems margin (29.5%) is a record — the two together define a company in transition.

Lens 6 · Earnings Calls (sentiment trend)

No transcripts on disk (transcripts=0), so this is `` plus the filings' MD&A tone.

• Recurring, intensifying phrase: "data center" / "strong demand for data center backup power" — it now leads the CEO's quote ("Our teams executed with discipline to meet continued strong demand for data center backup power" ) and recurs throughout the 10-K MD&A. Two years ago the call narrative was trucks-and-emissions; today it is power-gen and AI.
• What they stopped saying: the aggressive hydrogen/electrolyzer optimism. The 2025 language is a managed retreat — "rapidly deteriorating conditions in our electrolyzer markets… we intend to stop new commercial activity in the electrolyzer space". Tone on Accelera shifted from growth-story to cost-discipline.
• Net trajectory: more confident and more concrete on Power Systems (guidance raised twice into 2026), more sober and disciplined on the energy-transition bets. Management is leaning into the cash-cow that's working and pruning the option that isn't.

Lens 7 · Comps

All multiples ``; trailing/forward as labeled. EV/EBITDA for CMI specifically n/a from a clean feed; do not fabricate.

Read: CMI at ~23x forward is the cheapest of the data-center-genset cohort — below CAT (~37x) and Generac (~32x), roughly in line with the truck-cycle name PACCAR (~19-22x). The market is valuing CMI as a truck company with a power-gen kicker, not yet as a power-gen company with a truck drag. The bull case is that the multiple migrates toward CAT/Generac as Power Systems' share of EBITDA keeps rising; the bear case is that the truck-cyclicality cap is correct and the re-rating is largely done. 5-year average ROE: n/a.

Lens 8 · Stock-Price Catalysts (>5% moves, ~5 years)

Mostly ``; corroborated by filings where noted.

• 2021 — COVID-recovery cycle; FY guide raised to +20-24% revenue. Up-moves on reopening demand.
• Aug 2022 — Meritor acquisition (~$3.7B) closed; strategic, expanded Components.
• Dec 2023 — $2.0B emissions-settlement charge booked (EPS to $5.15); the stock took the hit but recovered as the market read it as one-time.
• Mar 2024 — Atmus split-off, $1.3B gain, −5.6M shares; Q2 2024 $1.9B settlement cash payment hammered operating cash flow to $1,487M.
• 2024-2026 — the dominant catalyst: the data-center power-gen re-rating. From a ~$322 12-month low to ~$705, the stock roughly 2.2x'd as Power Systems EBITDA compounded (+44% FY25, +19% rev / 29.5% margin Q1'26) and FY2026 guidance was raised on data-center demand.
• 2025 — Accelera/hydrogen reset (electrolyzer wind-down, $458M charges); a sentiment positive (removing a cash drain) more than a fundamental hit.

Pattern: historically CMI traded on the North American truck cycle and one-off legal/portfolio events. Since 2024 the marginal price-setter is data-center power demand — the stock now reacts to Power Systems orders, genset lead times and hyperscaler capex far more than to Class-8 truck builds. That regime change is the whole investment debate.

---

Phase C — Judge people & books

Lens 9 · Management

• CEO: Jennifer Rumsey — CEO since Aug 2022, Chair & CEO since Aug 2023; a 20-plus-year Cummins engineer who ran the Components segment and was COO (Mar 2021) before the top job; first woman to lead Cummins. Insider-operator archetype, not a parachuted financial CEO.
• Track record: her tenure spans (a) settling the decade-long emissions overhang ($2B, paid down), (b) the Atmus split-off (portfolio simplification), (c) the pivot to data-center power, and (d) the disciplined retreat from hydrogen/electrolyzers — i.e. she has both pushed the growth engine and cut a failing bet, which is the harder, more credible signal.
• Skin in the game / comp: total comp ~$19.9M, ~92% variable (stock/options/bonus) — heavily equity-weighted, reasonable for a $97B-cap industrial. Specific insider-ownership % n/a (no insider-transactions.csv on disk). She joined the 3M board in Jun 2026 — a mild attention-bandwidth flag, common for sitting CEOs.
• Capital allocation: 16th consecutive annual dividend raise territory — quarterly dividend +9.9% to $2.00 in Jul 2025 (FY annual $7.64/sh), after +8.3% (2024) and +7.0% (2023). No buybacks 2023-2025 ($2.2B authorization idle) — capital deliberately conserved for the settlement, Accelera funding and the capacity build; LT goal is to return ~50% of operating cash flow to shareholders, $519M returned in Q1 2026. This is conservative, shareholder-friendly stewardship — net leverage 0.22x is almost too under-levered for a business with this cash generation and a clear growth runway.
• Red flags: the defeat-device history predates her CEO tenure but is a cultural mark on the franchise (Lens 10). No related-party or promotional-behavior flags surfaced.

Lens 10 · Forensic Red Flags

Income statement / balance sheet / cash flow, all labeled.

• Earnings quality — clean. FY2025 consolidated net income $2,957M vs operating cash flow $3,621M — cash flow exceeds earnings, the healthy direction. The FY2024 OCF collapse to $1,487M is fully explained by the $1.9B settlement payment, not deteriorating quality.
• Non-GAAP vs GAAP gap is real and recurring. FY2025 GAAP EPS $20.50 vs an adjusted base materially higher because of Accelera impairments/exit charges of $458M in 2025 ($210M goodwill impairment + $119M inventory write-downs + PP&E/intangible impairments + $47M severance/contract-termination), on top of $312M of 2024 Accelera charges, plus the $199M Q1 2026 fuel-cell-sale charge. The adjustments are legitimate (genuine non-cash impairments of a strategy being wound down), but the pattern — recurring "one-time" charges in the green-energy unit — means the adjusted number flatters the optics; track whether Accelera charges truly stop in 2026.
• Working capital: Q1 2026 A/R was a $612M use of cash on a volume ramp — monitor that receivables don't outrun the data-center revenue.
• Goodwill/intangibles: the electrolyzer goodwill was fully impaired in 2025 (no overhang left there); broader goodwill from Meritor sits on the balance sheet — no impairment flagged.
• SBC / dividends / leverage: modest dilution (138.7M diluted shares, down from 142.7M in 2023 via the Atmus exchange); dividend well-covered by FCF; leverage conservative. No leases or revenue-recognition anomalies surfaced (deferred revenue $1,606M current / $1,054M long-term, growing with the genset book).

Regulatory findings (required sub-section).

• SEC: No Litigation Releases and no AAERs naming Cummins in the 2021-06-30 → 2026-06-30 window, verified via SEC EDGAR EFTS.
• Non-SEC enforcement — material and large. The EPA/DOJ/CARB emissions ("defeat device") settlement: Cummins agreed to a $1.675 billion civil penalty — the largest in Clean Air Act history and the second-largest environmental penalty ever — with total cost (penalty + mitigation + recalls) ~$2 billion; $1.9B paid in Q2 2024. It covered ~630,000 MY2013-19 RAM trucks with software defeat devices plus ~330,000 MY2019-23 vehicles with undisclosed auxiliary emission-control devices (~1M engines), and requires repairing ≥85% of affected trucks within three years (with Stellantis) or face additional penalties. This is fully accrued and largely paid — a closed financial chapter, but a permanent reputational mark and a reminder of regulatory tail-risk in the diesel franchise.
• 10-K Item 3 / Note 14 (Commitments & Contingencies): the company's own disclosure of the Settlement Agreements (final and effective April 2024, $1.9B paid) is in the FY2025 10-K; no new material proceedings beyond ordinary-course litigation were flagged.
• Net: one very large, now-resolved environmental enforcement action; no SEC accounting enforcement; ongoing recall-execution obligation to monitor.

---

Phase D — Project & stress-test

Lens 11 · Forward Projection (FY2026 / FY2027 / FY2028)

Built bottom-up from FY2025 actuals + the raised FY2026 guide. Base/bull/bear; output ``, inputs labeled. (No forecast.ts create — unattended --watchlist rule.)

Anchors: FY2025 clean GAAP EPS $20.50; FY2026 guide rev +8-11%, EBITDA 17.75-18.5%; consensus FY2026 EPS ~$26.44 (range 24.56-30.46), FY2027 ~$31.53 (range 26.45-37.66); Q2'25 already printed adj EPS $6.43 and Q1'26 adj $6.15.

The base case is essentially "consensus is right" — a high-$20s 2026 building to low-$30s 2027 as power-gen leverage compounds and trucks stop falling. The spread between bull and bear is driven almost entirely by two independent variables: (1) how durable AI/data-center genset demand is, and (2) when the North American truck cycle troughs. Brier forecast to log when conviction-grade (deferred here): "CMI FY2026 adjusted EPS ≥ $26" — directionally likely given the raised guide and the Q1 beat.

Lens 12 · Bull vs Bear

Bull case. Cummins is a mispriced AI-infrastructure beneficiary wearing a truck-engine costume. Data centers cannot run without reliable backup power, gensets are capacity-constrained (18-month high-HP lead times), and Cummins is the vertically-integrated #2 in a duopoly with CAT — owning the scarce link (its own engines) and the integration. Power Systems already earns ~22% segment EBITDA (29.5% in Q1'26) and is the fastest-growing, least cyclical leg; Distribution layers a recurring service annuity on top. Management raised FY2026 guidance, set a $45-50B / 20%-EBITDA 2030 framework, runs 0.22x net leverage with $2.2B of idle buyback authorization, and has been cutting (not feeding) the loss-making hydrogen bet. As power-gen's share of EBITDA rises, the multiple should migrate from a truck-cycle ~23x toward the genset-peer ~30-37x — a re-rating and an earnings tailwind. The contrarian surprise: investors still bucket CMI with PACCAR, not with the AI-power complex.

Bear case (permanent-impairment risks).

1. The truck franchise is structurally, not just cyclically, challenged. Engine EBITDA margin fell to 10.4%; if the energy transition (or a prolonged freight recession) permanently shrinks diesel-truck volumes faster than power-gen grows, the largest revenue base erodes. The HELM X15 diesel launch was already delayed to late 2026 on regulatory uncertainty — a tell that the core product roadmap is hostage to emissions politics.
2. Data-center genset demand could be a capex bubble. A meaningful slice of the re-rating assumes hyperscaler backup-power buying sustains. If AI capex digests, or if grid-scale / fuel-cell / on-site-gas alternatives erode the diesel-genset TAM, the highest-margin growth leg de-rates hard — and a lot of that is already in the ~$705 price.
3. Capacity-driven pricing power is self-limiting. Today's margins reflect short supply (18-month lead times). CAT (2 GW order in hand), Rolls-Royce mtu, Generac and others are all expanding; once the industry catches up, genset pricing — and Power Systems' 29.5% margin — normalizes.

Pre-mortem (18 months out, thesis broke). Most likely failure: AI data-center order growth decelerated just as genset capacity additions (Cummins' own Fridley +30%, plus CAT/mtu) hit the market, so Power Systems revenue growth halved and margins compressed toward the low-20s — while the North American truck cycle, instead of recovering in 2027, stayed in a freight recession. Consensus FY2027 EPS got cut from ~$31.5 toward the mid-$20s, and a stock priced for a power-company re-rating got re-bucketed back to a truck-cycle ~18x. The two legs disappointed together.

Multiples too high? At ~23x forward CMI is cheaper than its genset peers but richer than its own truck-cycle history; the multiple is fair-to-full only if the power-gen mix-shift is durable. It is not an obvious bargain at a 52-week high.

Contrarian view (what the market refuses to see): the bears under-weight the Distribution annuity — the recurring parts/service layer (+31% EBITDA in 2025) that compounds off both the truck and genset installed base and is far stickier than either OEM sale. The genuine debate is durability and timing of two cycles, and the service razor-and-blade is the under-discussed shock absorber.

Lens 13 · Devil's Advocate (short-seller)

Dismantling the bull case:

• Revenue concentration & cyclicality. The Engine + Components legs (~$21B of segment sales) ride the North American Class-8 truck cycle — a notoriously volatile, OEM-buyer-power market where Cummins is one of several suppliers. Q1 2026 NA sales −6%; Engine margin halved. The "diversified power company" framing can't fully escape that the largest base is deeply cyclical and currently masking how much of the EPS story leans on one hot end-market.
• The moat may be thinner than bulls think. The data-center genset "duopoly" pricing power is a function of a temporary shortage, not a structural barrier. Diesel-genset technology is mature; CAT is bigger; mtu, Kohler/Rehlko, Generac and Mitsubishi all build credible product and are all expanding capacity. When 18-month lead times compress to 6, the 29.5% Power Systems margin is the first thing to go.
• Most dangerous competitor bulls underestimate: not CAT (priced in) but on-site natural-gas and fuel-cell prime power plus utility/grid solutions that could shrink the diesel-backup TAM entirely — and Cummins' own X15-gas/HELM and the failed electrolyzer bet show how hard it is to win the post-diesel architecture.
• Capital-allocation / accounting nits: the recurring "one-time" charges in Accelera (2024 $312M, 2025 $458M, Q1'26 $199M fuel-cell) mean the adjusted EPS the bulls quote is systematically above GAAP; the defeat-device history ($2B, largest CAA penalty ever) is evidence the franchise will cut corners under emissions pressure. No buybacks for three years also signals management isn't convinced the stock is cheap.
• Assumptions that must hold for ~$705: sustained AI data-center genset demand and a 2027 truck recovery and margins holding near record and the multiple staying above truck-cycle norms. That is a stack of independent bets.
• −20-30% growth shock: if FY2027 EPS comes in ~$25 instead of ~$31.5 and the multiple compresses to ~18x (truck-cycle), the stock is ~$450 — roughly −35% from here. The downside is real precisely because both the earnings and the multiple are leaning on the same optimistic narrative.
• Single permanent-impairment scenario: a structural collapse in diesel-truck demand (regulatory ban acceleration + freight depression) that outpaces power-gen growth, turning the cash-cow into a melting ice cube while the growth leg de-rates. Plausibility: moderate over 3-5 years, low over 12 months.

Lens 14 · Management Questions (ordered by information value)

1. What share of Power Systems' record margin is price (shortage-driven) vs structural mix/cost — and where does the segment margin settle once industry genset lead times normalize from 18 months to 6?
2. How much of the data-center backlog ("6-8 quarters") is firm, non-cancellable orders with deposits vs soft demand signals, and what are the cancellation terms?
3. What is your base-case North American Class-8 truck volume assumption for 2026 and 2027, and what freight/rate signposts would tell you the trough is in?
4. With net leverage at 0.22x, $2.2B of idle buyback authorization, and ~50%-of-OCF return target — why no repurchases for three years, and what changes that?
5. When do Accelera restructuring charges genuinely stop, and what is the steady-state annual cash burn of the remaining (post-electrolyzer) Accelera portfolio?
6. Is the X15 diesel launch (late 2026) still on track, and how much 2026-27 Engine revenue is at risk from further regulatory-driven delays?
7. What is the realistic 2030 Power Systems revenue and margin contribution inside the $45-50B / 20%-EBITDA framework — i.e. how much of the framework depends on data centers?
8. How exposed is the genset franchise to a substitution from diesel backup toward on-site natural-gas prime power, fuel cells, or grid/storage solutions over the next 5 years?
9. What is the return profile and ramp risk of the Amplify Cell (21 GWh Mississippi) battery JV, and what happens to it if US EV-commercial-vehicle incentives are further cut?
10. How do you think about capacity-addition discipline industry-wide — what stops Cummins, CAT and mtu from collectively over-building genset capacity into a demand peak?
11. What is the recall-completion status on the RAM defeat-device program (the ≥85%-in-3-years obligation), and is there residual financial exposure?
12. What is the China power-gen exposure (a Q1'26 growth driver), and how do you frame the geopolitical/tariff risk to that revenue?
13. How much of Distribution's +31% EBITDA is durable recurring service vs power-gen product pass-through that moves with the genset cycle?
14. What is the insider/management ownership trajectory, and how is the variable-comp scorecard weighted between power-gen growth and total-company returns?
15. What capital-allocation priority wins if data-center cash flow surges — buybacks, M&A, capacity, or accelerating the decarbonization roadmap?

---