Cummins Inc. (NYSE: CMI) — A Diesel Cyclical Re-Rated as a Data-Center Power Play

Date: 2026-06-13 · Price (ref): ~$656 · Market cap: ~$90.6B · EV: ~$96B Sector: Industrials — Engines, Power Systems, Components & Distribution · Fiscal year: December · CIK: 0000026172

⚡ Claude’s Take

This block is the author’s own independent, subjective opinion and general information only — not investment advice. The analysis that follows it takes no position and carries no price target; the single opinion in this article is contained in this block.

Verdict: HOLD / trim into strength. Accumulate only on a cyclical pullback toward a ~14–15x EV/EBITDA, roughly the high-$400s to mid-$500s ($480–540). Not a short. Conviction: medium.

Tag: “A truck-cycle company wearing a data-center halo.”

Cummins is a genuinely good, durable industrial franchise — the largest independent engine maker on earth, with a real aftermarket/distribution annuity and best-in-class emissions know-how — that the market has, in the space of nine months, repriced from a ~12x mid-cycle cyclical to a ~19x EV/EBITDA “AI power” stock. The re-rate rests on something real: the Power Systems segment (about 22% of sales) is riding a data-center backup-power boom, posted a 22.7% EBITDA margin in 2025, has orders booked into 2028, and at the May 2026 Analyst Day management raised its 2030 data-center revenue target to ~$9B (from ~$5B in 2026) and group EBITDA to >20%. UBS chased it to a $850 target on June 8. That is the bull case, and it is not fantasy.

But here is what the $656 price is quietly underwriting: that the ~55–60% of Cummins still tethered to the North American heavy-duty truck cycle — Engine, much of Components, and the whole-goods half of Distribution — will hold up while the data-center engine compounds. The truck side is the problem. NA heavy-duty volumes fell ~20% into early 2026 and are now inflecting up only because of a 2026 pre-buy ahead of the 2027 EPA NOx rule (which the EPA has confirmed it will keep). Pre-buys borrow demand from the future; they are followed by air-pockets. Cummins’ own CFO flagged “volatility between the second half of this year and the first half of next year.” So you are being asked to pay a peak multiple for a business with one segment near a structural high (data-center genset margins that the capital cycle says will attract CAT, GE Vernova, Caterpillar and Generac capacity) and another approaching a cyclical air-pocket. That is the wrong time in the cycle to pay the highest multiple in the company’s recorded history. The market is mispricing the durability and blend of the two stories, not their existence. What flips me bullish: evidence that data-center genset demand is structural rather than a one-cycle AI capex spike — orders extending materially beyond 2028 and Power Systems sustaining >24% margins through a truck downturn. What flips me bearish: a 2027 post-pre-buy truck air-pocket landing harder than guided while data-center order intake decelerates — the combination that collapses both the earnings and the multiple at once.

1. Executive Summary

Cummins Inc. is a ~$34B-revenue global power-systems company built around four cash-generative segments — Engine, Components, Distribution, and Power Systems — plus a small, loss-making zero-emissions unit (Accelera). It is the world’s largest independent manufacturer of diesel and natural-gas engines, with a vertically connected portfolio spanning aftertreatment/emissions systems, turbochargers, fuel systems, axles and brakes (via the 2022 Meritor acquisition), power generation, and a 600-plus distributor / 7,200-dealer service network across ~190 countries.

The investment debate is no longer about whether Cummins is a good business — it is — but about what cycle and what multiple you are buying. Three facts frame everything:

1. Revenue has been flat at ~$34B for three years (2023–2025), but the mix has shifted dramatically. Power Systems EBITDA roughly doubled (margin 14.7% → 22.7%) on data-center power-generation demand, and Distribution EBITDA rose ~50%, while Engine and Components declined on lower North American truck volumes. The growth engine has changed.
2. Reported earnings are noisy. 2023 net income was crushed to $735M by a $2.0B EPA/DOJ emissions settlement charge; 2024’s $3.95B was flattered by a ~$1.3B non-taxable gain on the Atmus filtration split-off; 2025’s $2.84B absorbed ~$896M of Accelera losses (including $458M of hydrogen-exit charges). Normalized earnings power is materially higher than 2025 GAAP suggests — which is why the stock trades at ~32x trailing but ~19.5x forward P/E.
3. The stock has re-rated to the top of, and beyond, its own decade-long range. At ~$656, CMI trades at ~19x EV/EBITDA and ~3.0x book — versus a 10-year average of ~12x EV/EBITDA and ~1.8x book. It is, however, the cheapest member of the data-center-adjacent industrial peer set (CAT ~31x, ETN ~27x, DE/PH ~22–23x).

Business quality: moderate-to-good moat — brand/reliability, an installed-base aftermarket annuity, distribution scale, emissions-technology depth, and engine scale — but with a structural vulnerability: its largest customers (PACCAR ~13% of sales) and rivals (Daimler, Volvo, Traton/International) increasingly make their own engines. ROIC of ~12–14% is solid but not elite, consistent with that “real but contestable” moat.

Capital allocation: competent and conservative. Meritor (2022) was a sensible bolt-on; Atmus was monetized cleanly; the loss-making hydrogen/electrolyzer venture is being wound down with discipline; the dividend has risen ~16 consecutive years (now $2.00/qtr); buybacks were paused 2023–2025 to fund the EPA settlement and de-lever, and are now restarting. The balance sheet is healthy (net debt/EBITDA <1x, A-rated, overfunded pension).

The tension: Cummins is a high-quality cyclical that has been awarded a secular-growth multiple at a point when ~60% of the business sits near a cyclical top facing a 2027 pre-buy air-pocket, and its star segment’s margins are likely near peak. The valuation embeds the bull case; the cyclical core embeds the risk. No recommendation or price target is offered in this body.

2. Business Overview

Cummins designs, manufactures, distributes and services power solutions. Founded in 1919 in Columbus, Indiana, it employs ~69,600 people and sells in ~190 countries. It reports five segments. Because Cummins sells extensively between its own segments (an engine carries Cummins components and is sold through Cummins distribution), segment “total sales” sum to ~$41.4B before ~$7.7B of intersegment eliminations, reconciling to ~$33.7B consolidated. The segment picture (FY2025, $M):

Engine — diesel and natural-gas engines (4 to 78 liters) for heavy- and medium-duty trucks, buses, RVs, construction, agriculture, marine and rail. This is the historic core and the most cyclical piece; heavy-duty truck engines (~$3.5B of segment external sales) swing with the North American Class 8 cycle.

Components — aftertreatment/emissions systems, turbochargers (Holset), fuel systems, electronics/controls, automated transmissions, and — since the 2022 Meritor deal — axles and brakes. Components revenue fell from $13.4B (2023, total) to $10.1B (2025) partly because the Atmus filtration business (~$1.6B of revenue) was deconsolidated in 2024.

Distribution — now the largest segment by external revenue ($12.4B). Sells and services the full product line through a captive plus independent network: parts, whole goods (engines, gensets), and field service. This is the razor-and-blades annuity — high-recurring, less cyclical, and a beneficiary of the data-center genset wave (North American distribution revenue +13% in 2025).

Power Systems — high-horsepower engines, power generation (gensets), and industrial applications. This is the growth and re-rating engine. Power-generation revenue rose ~19% to $4.7B (total-sales basis) in 2025, EBITDA margin expanded to 22.7%, and management attributes ~$3.5B of 2025 group revenue (Power Systems plus Distribution) to data-center end-markets, targeting ~$9B by 2030.

Accelera — the zero-emissions unit (battery-electric, fuel cells, electrolyzers, e-powertrains). It is a structural loss-maker (-$896M segment EBITDA in 2025, including $458M of charges as Cummins exited green-hydrogen/electrolyzers and sold its low-pressure fuel-cell business to Alstom). Management is shrinking it toward battery-electric powertrains and guiding 2026 losses down to ~$270–300M.

Segment economics through time ($M segment EBITDA / margin on total sales):

The table is the whole story in five rows. Power Systems EBITDA doubled in two years (836 → 1,694) and its margin expanded ~800 bps — that is the entire bull case and the source of the re-rating. Distribution EBITDA rose 50% (1,209 → 1,808) on data-center genset pull-through and parts. Meanwhile Engine EBITDA fell as NA truck volumes rolled over and R&D ramped for the 2027 launches, and Components EBITDA declined as the Atmus deconsolidation removed a high-quality filtration stream. Accelera’s loss widened every year (it is being staunched only now). So the headline “flat $34B revenue, flat ~$5B segment EBITDA” hides a portfolio in which the cyclical legs are weakening while the power/aftermarket legs surge — a favorable mix shift that the market has aggressively extrapolated.

How it makes money: ~30–40% of value sits in the recurring aftermarket/parts/service annuity (Distribution parts + Components service parts), which smooths the cyclicality of new-engine sales — Distribution grew EBITDA right through the 2024–2025 truck downturn, proof of the annuity’s counter-cyclical ballast. The original-equipment side (engines into trucks, gensets into projects) is cyclical and OEM-customer-concentrated. Roughly two-thirds of revenue is North America-weighted; China (~$1.0B long-lived assets) and India (~$0.6B) are the largest international footprints. Verdict: a diversified, vertically-integrated power-systems platform with a healthy recurring-revenue core, currently enjoying a once-in-a-cycle mix tailwind from data-center power that has lifted blended margins even as the truck core softened.

3. Industry Dynamics

Cummins straddles three distinct industries with very different structures.

(1) On-highway commercial-vehicle powertrains (Engine + much of Components). This is a deeply cyclical, mature, regulation-driven market. North American Class 8 truck production swings 30–50% peak-to-trough. Demand is being actively distorted right now by the 2027 EPA low-NOx rule (35 mg/bhp-hr standard). The EPA confirmed in late 2025 it will keep the 2027 timeline but plans a spring-2026 proposal to reduce compliance cost (warranty/useful-life relief). The new emissions hardware adds an estimated $8,000–$15,000 of content per heavy-duty truck, much of it Cummins powertrain — a genuine content tailwind, but one that triggers a classic pre-buy: fleets accelerate 2026 purchases to beat the price hike, then under-buy in 2027–2028. This is the single most important cyclical fact in the thesis. Structurally, the industry is an oligopoly (Cummins, Daimler/Detroit, Volvo, PACCAR, Traton/International) — but a worrying share of it is captive: truck OEMs increasingly build their own engines, structurally shrinking the independent-engine pool that is Cummins’ heartland.

(2) Power generation / data-center backup power (Power Systems + Distribution). This is the bright spot and the reason for the re-rating. AI data-center build-out is driving unprecedented demand for high-horsepower standby gensets. Order books extend into 2028; Cummins doubled its 95-liter genset capacity in 2025 and committed a further $450M to add ~20 GW of capacity (to 55 GW nameplate) by 2030. Marathon’s capital-cycle lens flashes a caution here: segment margins of 22.7% and “record order intake” are exactly the conditions that attract capital. CAT, GE Vernova, Caterpillar, Generac, Rolls-Royce Power Systems and others are all expanding genset capacity. High returns invite supply; supply eventually compresses margins. The demand is real, but a 22.7% Power Systems margin should be treated as near-cyclical-peak, not a new permanent baseline.

(3) Zero-emissions / hydrogen (Accelera). Structurally unattractive today — green-hydrogen economics “dried up” (management’s words), demand is policy-dependent, and the whole sector is loss-making. Cummins’ retreat here is rational.

Profit-pool & value-chain note. In the on-highway chain, the profit pool has migrated toward (a) emissions/aftertreatment content (regulation-driven, where Cummins is strong) and (b) the aftermarket (parts/service over a 10–20-year engine life). The new-engine OE margin itself is thin and cyclical; the durable economics are in the regulated content and the service tail. In power-gen, the pool currently sits with whoever has available high-horsepower capacity — demand so outstrips supply that lead times, not price, are the constraint, which is why margins spiked. That is precisely the condition the capital cycle predicts will not last: extraordinary returns on scarce capacity are the signal that pulls in new capacity. Cummins itself is adding 20 GW; CAT, Generac, GE Vernova and Rolls-Royce Power Systems are all expanding. By 2027–2028 the binding constraint may shift from capacity to demand, and 22.7% margins would face the classic mean-reversion that follows every capacity boom.

A note on regulation as a moat-and-risk. Emissions regulation is double-edged for Cummins. It raises barriers (only scale players can certify across the power range and fund the multi-fuel R&D) and adds content ($8–15k per truck in 2027). But it also creates the pre-buy distortion that whipsaws the truck cycle, and it carries tail risk — the 2023 $2.0B EPA settlement is the cautionary example of what happens when certification goes wrong. The EPA’s planned 2026 cost-relief proposal could, paradoxically, soften the 2027 content uplift (less expensive hardware) while also softening the air-pocket (smaller pre-buy incentive) — a wash that the market has not clearly handicapped.

Verdict: A mixed industry portfolio. The trucking core is a cyclical, content-rich oligopoly with a structural captivity headwind; power generation is a genuinely attractive up-cycle that the capital cycle warns is sowing the seeds of future margin pressure; the hydrogen adjacency is structurally poor and being de-emphasized. Net structural attractiveness: moderate, cycle-dependent — better than a pure truck supplier, worse than a pure data-center electrical name.

4. Competitive Position

Cummins’ moat is real but moderate — and best understood through Greenwald’s taxonomy.

Economies of scale + intangibles (the strongest leg). Cummins is the largest independent engine manufacturer globally. Scale funds ~$1.4B of annual R&D (~4% of sales) — the price of staying ahead on each emissions-regulation step-change (Tier 4, EPA 2027, the “HELM”/fuel-agnostic platforms). Emissions/aftertreatment certification and the multi-year, multi-fuel R&D cadence are a genuine barrier: a sub-scale entrant cannot economically clear EPA/CARB/Euro VII hurdles across the full power range. Holset turbochargers and the aftertreatment franchise are technology-share leaders.

Customer captivity / switching costs (the recurring-revenue leg). Once a Cummins engine is in the field, it pulls a 10–20-year tail of Cummins parts and service through the captive distribution network. Fleet operators standardize on a powertrain to simplify maintenance, training and uptime. This installed-base annuity is the highest-quality, least-cyclical part of the company and the reason Distribution EBITDA grew through the truck downturn.

Brand / reliability. “Cummins inside” carries real pull-through with fleets; reliability and resale value support pricing.

Where the moat leaks — and it matters. Cummins’ core engine business sells to OEMs who are also competitors. PACCAR alone is ~13% of consolidated sales (~$4.4B in 2025, ~16% in prior years) — and PACCAR, Daimler (Detroit Diesel), Volvo, and Traton/International all make their own engines for at least part of their lineups. Every truck maker that vertically integrates its powertrain shrinks Cummins’ addressable independent-engine market. This is the structural reason Cummins’ ROIC sits at ~12–14% rather than the 20%+ of a true wide-moat compounder: customer captivity runs outward (Cummins’ aftermarket holds end-users) but not upward (OEMs can and do defect). The Power Systems / data-center franchise is currently the strongest competitive position Cummins holds — but it is also the most contested over a five-year horizon .

The OEM-integration math. Consider PACCAR: it buys ~$4.4B of Cummins product a year (~13% of CMI sales) while offering its own PACCAR MX engines as an alternative in Kenworth/Peterbilt trucks. Daimler Truck (Detroit Diesel), Volvo Group (Volvo/Mack engines) and Traton/International (International engines) are likewise dual-tracked. Cummins wins where its emissions/multi-fuel technology lead, breadth, or fleet preference outweighs the OEM’s integration incentive — which is why it retains a large independent share — but the secular vector points toward more insourcing, especially as electrification (whoever builds it) threatens to disintermediate the engine entirely over a 10–20-year horizon. This is the structural ceiling on the moat and the reason a “Cummins as wide-moat compounder” framing is wrong: the company must re-earn its position every product cycle against customers who are also competitors.

Greenwald tests applied. (1) Market-share stability — Cummins has held leadership in independent heavy-duty engines and aftertreatment for decades: stable share, supportive of a moat. (2) ROIC test — sustained ~12–14% ROIC (mid-teens in good years) confirms a real but moderate advantage; a no-moat commodity supplier would show ROIC at/below WACC, while a wide-moat franchise would show 20%+. Cummins sits in between — the financial fingerprint of a genuine-but-contestable moat. (3) The size of the pond (independent engines) is structurally eroding as OEMs integrate, which caps the value of the share-stability. Verdict: a durable, moderate moat anchored in scale, emissions IP, and an installed-base aftermarket annuity — strong enough to defend mid-teens ROIC, not strong enough to justify a wide-moat compounder multiple. The data-center genset position is a real near-term edge but a contestable one, and it does not change the structural verdict on the core engine business.

5. Growth History and Forward Opportunities

History (revenue, $M): 2018 $23.8B → 2019 $23.6B → 2020 $19.8B (COVID trough) → 2021 $24.0B → 2022 $28.1B (Meritor added ~$3B) → 2023 $34.1B → 2024 $34.1B → 2025 $33.7B. The 2021–2023 surge was part organic (truck up-cycle, pricing) and part acquired (Meritor, Aug 2022). The 2023–2025 plateau at ~$34B masks a major internal rotation: Power Systems and Distribution grew strongly while Engine and Components shrank (lower NA truck volumes plus the Atmus deconsolidation removing ~$1.6B). EPS has been whipsawed by one-offs; on a cleaner basis, underlying earnings power has drifted up despite flat revenue, thanks to Power Systems margin expansion.

Forward opportunities:

• Data-center power generation — the headline. Management’s raised 2030 target: ~$9B of data-center revenue (vs ~$5B in 2026) and group revenue of $45–50B with >20% EBITDA. This is the single biggest swing factor and the basis of the re-rating. It is a high-quality, high-margin, capacity-constrained opportunity — but demand visibility beyond ~2028 is the open question (orders are “well into 2028,” not 2030).
• EPA 2027 content + pre-buy. Near-term: a 2026 pre-buy lifts truck volumes (management raised its 2026 HD forecast to 230–250k units). Medium-term: ~$8–15k of new content per HD truck is a durable revenue/mix uplift once the air-pocket passes. The X15/X10 launch in 2027; the B-series (7-liter) launch was pushed to January 2028 pending the EPA’s revised rule.
• Natural-gas engines. The X15N (15-liter NG) launched 2024; a new 4 MW NG engine for data centers was announced at the 2026 Analyst Day. NG is a credible “bridge fuel” in trucking and a growth vector in power-gen.
• Aftermarket / Distribution compounding. The installed-base annuity grows mechanically with each year of new-engine shipments and is the most reliable growth leg.
• Accelera optionality — heavily de-risked/de-funded; effectively a call option, not a growth driver.

Verdict: medium-quality growth with a high-quality near-term spike. The data-center and aftermarket legs are high-quality; the truck-content leg is real but cyclically front-loaded by the pre-buy; Accelera has been (rightly) demoted. The risk is that 2026’s reported growth borrows from 2027–2028.

6. Financial Quality

Eight-year financial summary ($M except per-share / %):

Read the EPS line with care: the 2023 collapse to $5.15 is the EPA charge; the 2024 spike to $28.37 is the Atmus gain; the 2025 $20.50 absorbs Accelera charges. The cleaner signal is the operating-margin trend ex-2023 drifting from ~9–10% to 11.5%, and the steady share-count reduction (162.8M → 138.7M, ~15% over eight years) compounding per-share value.

Revenue & margins. Consolidated gross margin is structurally ~24–25% and ticked up to 25.3% in 2025; operating margin reached ~11.5% in 2025 (best of the period) despite flat revenue, driven by Power Systems/Distribution mix. Group EBITDA margin was ~14.8% in 2025 (ROIC basis), and management’s 2026 guide is 17.75–18.5% (their “Segment EBITDA”-style basis runs higher than the all-in ROIC EBITDA figure). Either way, the trajectory is up on mix — the question is how much of the Power Systems margin gain is durable.

Earnings quality — read past the GAAP line. Three consecutive years of large, distinct distortions:

• 2023: a $2.0B EPA/DOJ emissions settlement charge (RAM 2500/3500 engine certification; $1.7B of it non-deductible) cut operating income to $1.28B and net income to $735M. $1.9B was paid in 2024.
• 2024: a ~$1.3B gain on the Atmus split-off (tax-free exchange of 67M Atmus shares for 5.6M Cummins shares) inflated net income to $3.95B and pushed the effective tax rate down to 17%.
• 2025: ~$896M of Accelera losses including $458M of hydrogen/electrolyzer exit charges held net income to $2.84B (EPS $20.62) and the tax rate up to 25.4%.

The cleanest way to think about normalized earnings: 2025’s ~$20.6 GAAP EPS understates run-rate by roughly the after-tax Accelera charge, while 2026 consensus (~$32–34) reflects volume recovery, the pre-buy, Power Systems strength, and a smaller Accelera drag. This is why trailing P/E (~32x) and forward P/E (~19.5x) diverge so widely — the trailing figure is on depressed, charge-laden earnings.

Cash flow & conversion. Operating cash flow was $3.62B in 2025; capex was ~$1.24B, so free cash flow was ~$2.4B (~85% of net income on a clean basis). FCF conversion is respectable but not exceptional — this is a working-capital-intensive, ~3.7% capex/sales business. 2024 OCF was unusually low ($1.49B) because $1.9B of the EPA settlement was paid that year. Capex is rising (734M → 916M → 1,213M → 1,208M → 1,235M, 2021–2025) and guided to $1.35–1.45B in 2026 to fund genset capacity.

Returns on capital. ROIC ~14.0% (2025), ~15.2% (2024), with a ~12–14% normalized range; ROE 12.8% (2025), distorted to ~20% in 2024 by the Atmus gain. Against an estimated ~9–10% WACC, Cummins earns a modest but positive economic spread — value-creating, but not a high-return compounder. The spread widens in up-cycles and compresses (briefly negative on a GAAP basis in 2023) in shock years.

Balance sheet. Strong. Total debt ~$7.6B against ~$3.6B cash & securities → net debt ~$4.7B, net debt/EBITDA <1x, ~36% of total capital, A-rated, with a well-laddered maturity profile (no near-term wall) and an overfunded pension (+$489M US, +$70M non-US) — i.e., no hidden liability drag. Goodwill/intangibles (~$6.6B combined) are largely Meritor-related; book value per share is ~$168 and tangible book ~$65, so the stock trades at ~3.0x book and ~10x tangible book — rich versus history.

Verdict: good financial quality with genuinely improving economics-with-scale on the Power Systems mix shift — but returns are mid-teens, not elite, and reported earnings require active normalization. Cash generation and the balance sheet are sound; the quality story is the margin mix, not the headline EPS.

7. Capital Allocation

Management (CEO Jennifer Rumsey, CFO Mark Smith) has run a competent, conservative, shareholder-aware program, with a few notable judgment calls.

M&A. The Meritor acquisition (Aug 2022, ~$2.86B purchase price / ~$3.7B EV incl. assumed debt) added axles, brakes and drivetrain (into Components) and electric-powertrain capability (into Accelera). It was a sensible, in-footprint bolt-on that broadened content-per-vehicle; it raised leverage temporarily, which the company then prioritized paying down. Earlier deals (Hydrogenics, electrolyzer assets) made into the hydrogen thesis that has since been written down — a real, if modest, capital-allocation error now being corrected.

Portfolio surgery. Two clean moves: (i) the Atmus filtration separation (IPO 2023, full tax-free split-off March 2024) crystallized value and simplified the portfolio, netting a ~$1.3B gain and retiring 5.6M Cummins shares via the exchange; (ii) the disciplined wind-down of Accelera’s hydrogen/electrolyzer businesses (sale of low-pressure fuel cells to Alstom; $458M of 2025 charges). Cutting a loss-making “strategic” venture is harder than starting one — management deserves credit for the retreat.

Buybacks & dividends. Open-market buybacks were effectively paused 2023–2025 (after ~$1.4B in 2021 and $0.4B in 2022) to fund the $1.9B EPA payment and de-lever — a defensible prioritization. They are now restarting ($243M in Q1 2026 at an average ~$537/share). The dividend has risen for ~16 consecutive years, raised ~10% to $2.00/quarter in mid-2025 (~$8.00/yr, ~1.2% yield, ~35–40% payout). Management frames a return target of ~50% of operating cash flow to shareholders. Note the buyback-restart price (~$537) is well below today’s ~$656 — a mild signal that management itself found the stock more attractive a quarter ago than the market does now.

R&D / capex intensity. R&D ~$1.4B/yr (~4% of sales) and rising capex are appropriate for a regulation-and-capacity-driven business; the incremental capex is demand-pulled (genset capacity with order backing), not speculative.

Incentive alignment & insiders. Compensation is performance-weighted (ROIC and EBITDA feature in the program). Insider behavior is a mild negative tell: a two-year review of Form 4 filings shows no discretionary open-market purchases (code P) by any named officer or director — activity is entirely routine option-exercise-and-sell, grants, tax-withholding and estate transfers. No conviction buying at any price, including the 2024–2025 lows.

Verdict: above-average, disciplined capital allocation — good portfolio surgery, sensible M&A, conservative balance-sheet management, a long dividend record, and an honest retreat from a failing venture. The blemishes (the hydrogen write-down; the absence of insider buying) are modest. Management has earned trust; it has not signaled that it is a buyer at $656.

8. Changes and Headwinds — Last Two Years

• The data-center re-rating (2025–2026): Power Systems EBITDA margin 14.7% → 22.7%; ~$3.5B of 2025 data-center revenue; record order intake into 2028. The May 21, 2026 Analyst Day raised 2030 targets (revenue $45–50B, EBITDA >20%, data-center revenue ~$9B) and announced +$450M / +20 GW genset capacity and a new 4 MW NG data-center engine. This is the dominant positive change.
• EPA 2027 rule confirmed, B-series delayed: EPA will keep the 2027 NOx timeline but plans a 2026 cost-relief proposal; Cummins pushed its B-series (7-liter) launch to January 2028 (X15/X10 still 2027). Sets up a 2026 pre-buy and a 2027–2028 air-pocket.
• EPA/DOJ settlement closed (2023–2024): $2.0B charge, $1.9B paid in Q2 2024 — now behind the company but a reminder of regulatory/legal tail risk in emissions certification.
• Atmus fully separated (March 2024).
• Accelera retrenchment (2025–2026): hydrogen/electrolyzer exit, fuel-cell sale to Alstom, losses guided down.
• Customer mix: PACCAR concentration ~13% (down from ~16%); China power-gen +84% in Q1 2026 on data-center demand (a new, fast-growing but geopolitically exposed channel).
• Tariffs: management characterizes the net 2026 tariff EBITDA impact as immaterial (~20–30 bps drag), recovered largely dollar-for-dollar.
• Leadership/board: routine — CAO retirement (2025); ex-GM executive Matthew Tsien added to the board (2025).

Verdict: on balance these changes have strengthened the narrative and the multiple more than the underlying through-cycle earnings power — the genuine improvement is the Power Systems franchise; the genuine risk added is the pre-buy-driven cyclical setup into 2027.

9. Risk Analysis (Risk Matrix)

Catastrophic-loss risk is low — strong balance sheet, no liquidity/solvency concern, diversified end-markets. The dominant risks are earnings-cycle and valuation-multiple risks (#1–#3), which are correlated and could compound.

10. Valuation Discussion (Embedded Expectations)

Where it trades (ref ~$656). Market cap ~$90.6B (138.1M shares); net debt ~$4.7B; NCI ~$1.06B → EV ~$96B. On ~$4.97B trailing EBITDA → ~19.4x EV/EBITDA. Forward P/E ~19.5x (on ~$32–34 2026E EPS); trailing P/E ~32x (on charge-depressed 2025 EPS); P/B ~3.0x; P/sales ~2.7x; dividend yield ~1.2%.

(Note: EV is built from the 138.1M filing share count and FY2025 net debt; aggregator EV figures that rely on a stale share count are not used, and the result cross-checks to a market-data EV/EBITDA of ~19.4x.)

Versus its own history — the key fact. Cummins’ 10-year EV/EBITDA range is roughly 8–16x with a ~12x average; the prior decade high was ~15.9x (2025). At ~19x, the stock is above its entire recorded range. P/B (~3.0x) is likewise at a decade high versus a ~1.8x average. By its own yardstick, Cummins is more expensive than it has ever been.

Versus peers — the bull’s anchor.

Cummins is the cheapest of the data-center-adjacent industrial cohort on both EV/EBITDA and (ex-PCAR) forward P/E, and trades at less than half CAT’s and ETN’s P/sales. The bull argument (UBS $850, ~30% above spot) is explicitly a convergence trade: if Cummins’ data-center franchise deserves to be valued like ETN’s or CAT’s, the multiple has further to run. The bear retort is that the discount is deserved — CMI’s revenue growth (~3%) trails ETN/CAT/PH, its margins are lower (~15% EBITDA vs ETN/CAT/PH at ~20%+), its ROIC is mid-teens vs those names’ higher returns, and its closest structural comp is actually PCAR (the truck OEM, ~21x) and the truck cycle — not the electrical/data-center pure-plays. CMI sits between the two cohorts, which is exactly where a half-cyclical-half-secular business should sit.

Embedded-expectations / scenario analysis. What must the ~$96B EV justify?

• Bear (~$430–490, ~12–14x mid-cycle EBITDA on ~$4.0–4.3B normalized through-cycle EBITDA): the 2027 air-pocket hits, data-center order growth cools, Power Systems margin reverts toward high-teens, and the multiple normalizes to history. This is roughly a 25–35% drawdown — the “it’s a cyclical and the cycle turned” outcome.
• Base (~$560–680, ~15–17x on ~$5.5–6.0B forward EBITDA): 2026 pre-buy + Power Systems strength lift EBITDA toward $5.5–6B; the market grants a structurally higher (but not peer-equal) multiple in recognition of the improved mix; the 2027 air-pocket is shallow/short. Today’s price sits near the top of this zone.
• Bull (~$800–900, UBS $850 territory): data-center demand proves structural through 2030, Power Systems sustains ~25%+ margins, group EBITDA pushes toward $7B+ on the way to the $45–50B/>20% 2030 target, and CMI re-rates toward ETN/CAT multiples (~25x+).

Explicit embedded-expectations arithmetic. Reverse the multiple. At ~$96B EV, a buyer at a “fair” through-cycle ~13x EV/EBITDA would need ~$7.4B of sustainable EBITDA — roughly 50% above the 2025 level of ~$5.0B and above even management’s 2026 guide. Put differently, today’s ~19x on ~$5.0B trailing EBITDA is only ~13x if you believe EBITDA reaches ~$7.4B and stays there. Management’s 2030 framework (revenue $45–50B, >20% EBITDA) implies ~$9–10B EBITDA — which at a normalized ~13–14x would support an EV well above today’s, if delivered and if the market grants a normal multiple on it. So the price is internally consistent with management’s 2030 plan landing and the multiple normalizing — it is not internally consistent with the through-cycle reality of the current (2025) earnings base. The gap between those two is the risk: you are paying today for 2030 delivery, with a 2027 air-pocket in between.

The crux: at ~$656 the market is paying a base-to-bull price. You are underwriting (a) Power Systems margin durability near 22–25%, (b) data-center demand extending well beyond 2028, and © only a shallow 2027 truck air-pocket — simultaneously. Each is plausible; all three together, at a record multiple, leaves little margin of safety. The variant question is not whether Cummins is good (it is) but whether a moderate-moat, mid-teens-ROIC cyclical should hold its highest-ever multiple at a cyclical inflection. No price target is set.

11. Variant Perception

Consensus (post-Analyst-Day, mid-2026): constructive-to-bullish. The sell-side has rotated CMI from “cyclical engine maker” to “low-risk AI/data-center power play.” UBS upgraded to Buy with a $850 target (June 8, 2026); Morgan Stanley turned positive; the stock rose from ~$510 (2025 close) to ~$656. Consensus accepts management’s raised 2030 framework (revenue $45–50B, EBITDA >20%, $9B data-center).

Strongest bull case: Cummins is the cheapest credible way to own the AI-power-infrastructure theme. Power Systems has visible, capacity-constrained, high-margin demand into 2028 and a clear path to $9B data-center revenue; the recurring aftermarket/Distribution annuity de-risks the cyclicality; the balance sheet is fortress-grade; buybacks are restarting; and the multiple still sits below electrical/data-center peers. EPA 2027 content adds a durable mix uplift on the truck side. Re-rating + earnings growth = the $850 case.

Strongest bear case: This is a ~12x-history cyclical wearing a ~19x data-center costume. ~55–60% of the business is tied to a NA truck cycle that is only up because of a 2026 pre-buy that mechanically creates a 2027–2028 air-pocket; the prized Power Systems margin (22.7%) is near a capital-cycle peak that will attract competing capacity from CAT/GE Vernova/Generac; ROIC is mid-teens, not elite; insiders aren’t buying; and the stock is at its highest multiple ever just as the cyclical core peaks. Mean-reversion of either the earnings or the multiple — let alone both — produces a 25%+ drawdown.

The 3–5 assumptions that decide it:

1. Data-center genset demand durability beyond 2028 (structural vs. one-cycle AI-capex spike).
2. Power Systems margin sustainability (can ~22–25% hold as capacity floods in?).
3. Depth/timing of the 2027 truck air-pocket (shallow and short, or deep?).
4. The multiple (does ~19x EV/EBITDA hold/expand, or revert toward ~13x?).
5. Accelera ending as a manageable drag (already largely de-risked).

Falsification evidence: Bull breaks if data-center order intake decelerates or Power Systems margins roll over while truck volumes drop post-pre-buy. Bear breaks if order backlog visibly extends into 2029–2030 and Power Systems holds >24% margins through a truck downturn — proving the franchise has structurally de-cyclicalized.

12. Fact vs. Interpretation Table

13. Open Questions

1. How far does the data-center order book actually extend — is there visibility into 2029–2030, or does it stop at 2028? (Determines whether Power Systems is structurally de-cyclicalized.)
2. What is the true depth of the 2027 truck air-pocket given the EPA’s planned cost-relief proposal — could relief blunt both the pre-buy and the air-pocket?
3. What normalized Power Systems margin should we underwrite once new genset capacity (CMI’s +20 GW plus competitors’) comes online — 25% or back toward high-teens?
4. What is clean, through-cycle EPS stripping pre-buy benefit, Accelera drag, and one-offs? (Our ~$32–34 2026E is a peak-ish, not mid-cycle, figure.)
5. Will buybacks scale materially now that de-levering is complete, or will capex for genset capacity absorb the incremental cash?
6. How exposed is the China power-gen ramp (+84% in Q1’26) to trade/geopolitical disruption?

14. What Must Be True (Bull and Bear, with Falsification Tests)

Bull thesis — what must be true:

• Data-center backup-power demand is structural, not a single AI-capex cycle, sustaining double-digit Power Systems growth and ~25% margins through 2030.
• Power Systems + Distribution data-center revenue reaches ~$9B by 2030 as guided, lifting group EBITDA toward $7B+ and >20% margin.
• The 2027 truck air-pocket is shallow and brief, cushioned by EPA cost-relief and the aftermarket annuity.
• CMI re-rates toward its data-center peer group (~25x+ EV/EBITDA).
• Falsification test: if data-center quarterly order intake decelerates for two-plus consecutive quarters or Power Systems EBITDA margin falls below ~20% while truck volumes drop, the bull thesis is broken.

Bear thesis — what must be true:

• ~55–60% of Cummins remains a NA-truck cyclical whose 2026 strength is borrowed from 2027–2028.
• Power Systems’ 22.7% margin is a capital-cycle peak that competing capacity (CAT/GE Vernova/Generac) compresses.
• The ~19x EV/EBITDA multiple — above the entire 10-year range — reverts toward mid-cycle (~13–14x) as the truck cycle turns.
• Falsification test: if the order backlog visibly extends into 2029–2030 and Power Systems sustains >24% EBITDA margin through a truck-volume downturn, the bear (cyclical-at-peak-multiple) thesis is broken — the franchise has genuinely de-cyclicalized.

15. Source Appendix

The full dated source list appears in the Source Appendix below. Primary sources: Cummins FY2021–FY2025 Forms 10-K and FY2026 Q1 10-Q (SEC EDGAR, CIK 0000026172); Q1 2026 (2026-05-05) and Q4 2025 (2026-02-05) earnings-call transcripts; the May 21, 2026 Analyst Day investor materials (investor.cummins.com); aggregated fundamentals and valuation history (reconciled to filings); peer comps (2026-06-13); and trade-press coverage of the EPA 2027 NOx rule (FleetOwner, CCJ Digital, Dec 2025).

This article discusses valuation only as embedded expectations and scenarios. It contains no investment recommendation and no price target. The single subjective view is the clearly-labeled “Claude’s Take” block at the top. It is general information, not investment advice.

APPENDIX A — Standard Diligence Questionnaire

Cummins Inc. (NYSE: CMI) — supplemental to the research memo (2026-06-13)

Labels: (F) Fact · (I) Interpretation · (A) Assumption. Figures from FY2025 10-K, aggregated fundamentals data, transcripts, and Analyst Day unless noted.

General

What thoughtful questions have other investors asked about this company? The central debates: (1) Is the data-center power-generation demand structural or a one-cycle AI-capex spike? (2) How deep is the 2027 post-pre-buy truck air-pocket? (3) Should a mid-teens-ROIC, OEM-customer-concentrated engine maker hold its highest-ever multiple? (4) Can Power Systems’ ~22.7% EBITDA margin survive incoming competitive capacity? (5) Is Accelera’s cash drain finally ending? (6) Why are insiders not buying after the de-rating lows? Bulls (UBS $850, Morgan Stanley) frame CMI as the cheapest AI-power play; bears frame it as a cyclical at a peak multiple.

Cyclicality & Earnings Nature

Are earnings at a cyclical high or low? (I) Mixed and arguably near a blended high. NA truck volumes are below mid-cycle (down ~20% into early 2026) but inflecting up on a 2026 pre-buy; Power Systems is at a cyclical/structural high (record orders, 22.7% margin). The multiple is at an all-time high. So earnings are not at a clean trough — the pre-buy and Power Systems peak offset the truck weakness. Driven by external environment or internal actions? (I) Both — external (AI data-center demand, EPA 2027 regulation, truck cycle) plus internal (Power Systems capacity build, Accelera wind-down, Meritor/Atmus portfolio moves). How stable are revenues? (F/I) Flat at ~$34B for three years, masking a large internal rotation; the aftermarket/Distribution annuity (~30–40% of value) is stable, the OE engine/genset side is cyclical. Outlook for products/services? (F) 2026 guide: revenue +8–11%, EBITDA margin 17.75–18.5%; Power-gen +15–25%. 2030 targets: revenue $45–50B, EBITDA >20%, data-center ~$9B. How big will this market be? (I) Trucking: mature/cyclical, content-growing via emissions regulation. Data-center power: rapidly growing (CMI targets $9B by 2030 vs ~$5B 2026). Global, with US the largest market; China power-gen growing fast (+84% Q1’26).

Business Quality & Competitive Moat

Is the industry getting more or less competitive? (I) Trucking: more competitive as OEMs vertically integrate engines. Data-center gensets: intensely contested over a 5-yr horizon (CAT, GE Vernova, Generac, Rolls-Royce adding capacity). How profitable is the business (ROIC, ROE)? (F) ROIC ~14.0% (2025), ~12–14% normalized; ROE 12.8% (2025), ~20% (2024, Atmus-inflated). Modest positive spread over ~9–10% WACC. How profitable is the industry / barriers to entry? (I) Engine/aftertreatment has high barriers (emissions certification, multi-fuel R&D ~$1.4B/yr, scale); independent-engine oligopoly. Power-gen barriers moderate (capacity + technology). Hydrogen: structurally unprofitable today. Can the business be easily understood? (F/I) Yes at the segment level, but GAAP earnings are noisy (3 straight years of large one-offs) — requires normalization. Undermined by foreign low-cost labor? (I) Limited — heavy, regulated, capital- and IP-intensive products; CMI predominantly makes/sources in the US for the US market. Do brands matter? (F/I) Yes — “Cummins inside” pull-through with fleets; reliability/resale support pricing. Nature of competition? Both OEM-captive engines and independent industrial peers; data-center gensets vs power-equipment majors. Customers’ switching costs? (I) High for end-users (installed-base parts/service annuity, training, uptime standardization); low for OEM customers, who can insource.

Financial Condition & Balance Sheet

Assets not fully on the balance sheet? (I) The installed-base aftermarket annuity and distribution network are under-represented intangibles; brand value. Off-balance-sheet liabilities? (F) None material flagged; pension is overfunded (+$489M US / +$70M non-US); small OPEB underfunding (-$108M). EPA settlement fully accrued and ~$1.9B paid in 2024. How conservative is the accounting? (I) Reasonably conservative; one-offs are clearly disclosed; “Segment EBITDA” is the management metric (runs higher than all-in EBITDA — reconcile). How CapEx-hungry? (F) Moderate — capex ~3.5–4% of sales ($1.24B 2025; $1.35–1.45B 2026E), rising to fund genset capacity (demand-backed).

Capital Allocation & Management

FCF generation & use / philosophy? (F) OCF $3.62B, FCF ~$2.4B (2025); target ~50% of OCF returned to shareholders. Dividend (16-yr growth, $2.00/qtr) + restarting buybacks + capex + de-levering. Significant acquisitions? (F) Meritor (Aug 2022, ~$2.86B purchase). Hydrogen/electrolyzer deals (Hydrogenics) now written down/exited. Buying back shares? (F) Paused 2023–2025 (EPA payment, de-lever); restarted Q1’26 ($243M @ ~$537). Issuing shares to insiders? (F) Routine equity comp; share count down (~138.1M, from ~165M a decade ago) — net reducer over time. Compensation policy? (F/I) Performance-weighted (ROIC, EBITDA); reasonable. No insider open-market buying in 2-yr Form 4 review — mild negative tell. Motivations of management? (I) Long-tenured, operationally credible (Rumsey/Smith); disciplined portfolio managers; not signaling personal conviction at $656.

Valuation & Market Data

ADR / MLP / K-1? (F) No — ordinary US common stock, NYSE: CMI. Standard 1099 dividend. Dividend policy? (F) $2.00/qtr (~$8.00/yr), ~1.2% yield, ~35–40% payout, ~16 consecutive annual increases. How profitable? (F) Net margin ~8.4% (2025, charge-affected); EBITDA margin ~14.8%; operating margin ~11.5%. Net income vs cash from operations diverging? (F/I) Yes in distorted years (2023 EPA charge; 2024 EPA cash payment depressed OCF to $1.49B vs $3.95B NI). 2025 OCF $3.62B > NI $2.84B (charges non-cash) — favorable. Normalize across years.

Risks & Downside

What would cause the stock to decline? (I) A 2027 truck air-pocket; data-center order deceleration / Power Systems margin reversion; multiple compression from a record ~19x; OEM insourcing; China/tariff shock; renewed emissions litigation. Catastrophic loss risk? (I) Low — strong balance sheet (net debt/EBITDA <1x, A-rated), diversified, no solvency concern. Chance of total loss? (I) Negligible.

Recent News & Events

Has the business environment changed recently? (F) Yes — the data-center power-gen boom re-rated the stock (~$510→~$656); May 2026 Analyst Day raised 2030 targets; EPA confirmed the 2027 NOx timeline (pre-buy/air-pocket setup); UBS upgrade to Buy ($850) June 8, 2026. Significant acquisitions / divestitures? (F) Atmus split-off completed March 2024; First Mode assets acquired Feb 2025; Komatsu/Wabtec hybrid-powertrain MOU Sept 2025; low-pressure fuel-cell business sold to Alstom (2026); electrolyzer business wound down (2025). Change in accounting policies? (F) None material noted; segment “New Power” renamed Accelera. Recent operational changes? (F) +$450M / +20 GW genset capacity expansion (to 55 GW by 2030); new 4 MW NG data-center engine; third shift at Rocky Mount; B-series launch delayed to Jan 2028.

APPENDIX B — Source Appendix

Cummins Inc. (NYSE: CMI) — research as of 2026-06-13

All non-obvious facts in the memo trace to the sources below. Primary sources (SEC filings, company releases, transcripts) take precedence; third-party data aggregators were reconciled to filings. Management commentary is treated as hypothesis, not evidence.

Primary — SEC filings (EDGAR, CIK 0000026172)

• Form 10-K, FY2025 (period 2025-12-31; filed ~2026-02-10) — segment data (Engine/Components/Distribution/Power Systems/Accelera), revenue by geography/end-market, PACCAR customer concentration (~13%), EPA settlement disclosure, Accelera charges ($458M), capex/R&D, debt maturity ladder, pension funded status, dividend history. https://www.sec.gov/cgi-bin/browse-edgar?action=getcompany&CIK=0000026172
• Forms 10-K, FY2021–FY2024 — historical segment & financial detail; Meritor purchase accounting (FY2022 10-K); Atmus separation (FY2023/FY2024); EPA settlement charge (FY2023).
• Form 10-Q, Q1 2026 (period 2026-03-31; filed ~2026-05) — Q1 segment trends, Power Systems +19% / 29.5% EBITDA, Accelera fuel-cell sale charge, buyback ($243M @ ~$537), 2026 guidance.
• Form 4 corpus (2024–2026) — insider transactions; no discretionary open-market purchases (code P) by named officers/directors; routine M/S/A/F/G activity (CEO Rumsey, CFO Smith).
• 8-K corpus (2024–2026) — quarterly earnings, dividend increase to $2.00/qtr (mid-2025), board/exec changes, EPA settlement announcement (2023-12).
• DEF 14A (2025 proxy) — compensation structure (ROIC/EBITDA metrics), board.

Primary — company materials

• Cummins Analyst Day, 2026-05-21 — raised 2030 targets (revenue $45–50B, EBITDA >20%, data-center revenue ~$9B vs ~$5B 2026); +$450M / +20 GW genset capacity (to 55 GW); new 4 MW natural-gas data-center engine. https://investor.cummins.com/news/detail/696/
• Q1 2026 earnings call transcript (2026-05-05) — NA truck outlook (HD 230–250k units), 2027 EPA pre-buy, Power Systems margin/orders, Accelera fuel-cell sale, capital allocation, 2026 guidance, tariffs.
• Q4 2025 earnings call transcript (2026-02-05) — record Power Systems order intake into 2028, data-center revenue ~$3.5B (2025), electrolyzer wind-down, capacity doubling, “low-risk play on the AI boom.”

Secondary — quantitative aggregators (reconciled to filings)

• Aggregated fundamentals data — multi-year income statement, balance sheet, cash flow, profitability ratios (ROIC/ROE/margins), per-share data, and own-history valuation multiples (FY2015–FY2025): EV/EBITDA range ~8–16x, avg ~12x; P/B avg ~1.8x. (Reconciled to filings; enterprise value rebuilt from the 138.1M filing share count + FY2025 net debt.)
• Market data (accessed 2026-06-13) — current price (~$656–660), stats (EV/EBITDA ~19.4x, forward P/E ~19.5x, trailing ~34x, dividend yield ~1.2%, ROE ~22%), and peer comps: CAT ~31.5x, ETN ~27.3x, DE ~22.7x, PH ~22.5x, PCAR ~21.4x EV/EBITDA.

Secondary — news & industry

• Benzinga (2026-06-08) — UBS upgrade to Buy, price target $565→$850 (analyst Steven Fisher).
• FleetOwner / CCJ Digital / Rush Truck Centers (Dec 2025) — EPA confirms it will keep the 2027 NOx (35 mg) timeline; plans a spring-2026 cost-relief proposal (warranty/useful-life); 2026 pre-buy dynamics; $8–15k content add per HD truck.
• TIKR / Sahm Capital / Simply Wall St / Yahoo Finance (May–June 2026) — Analyst Day coverage and Morgan Stanley positive stance (corroborating company release).

Accessed 2026-06-13. Prices and multiples are intraday/aggregator snapshots and will drift.