Cummins (CMI)

Diesel + natural-gas reciprocating engine duopolist (with caterpillar) — Power Systems is the company's fastest-growing segment on AI data-center backup and bridge power.

1. Core Product / Service

Cummins (NYSE: CMI) is a 105-year-old power-and-engine OEM headquartered in Columbus, Indiana. Five operating segments:

• Engine — heavy-duty diesel engines for commercial truck and off-highway
• Distribution — global parts, service, and aftermarket
• Components — fuel systems, filtration, emissions (Atmus, Cummins Filtration)
• Accelera (formerly New Power) — electrolyzers, fuel cells, electrified powertrains
• Power Systems — the AI data-center segment: high-horsepower diesel and natural-gas reciprocating engines, generator sets (gensets), and industrial power products

The AI tailwind sits almost entirely inside Power Systems. Cummins makes the diesel gensets that sit behind every hyperscale data center as backup, and — increasingly — the lean-burn natural-gas reciprocating engines (315 kW to 2 MW) used as bridge or prime power when utility interconnects are 3-5+ years out [1]. The product lineup includes the QSK95 (3.5 MW class), QSK60, and the smaller QSK series, plus natural-gas engine families.

2. Target Users & Pain Points

• Hyperscalers and colos — AWS, Microsoft, Google, Meta, Oracle plus equinix, digital-realty, qts, iron-mountain
• Neoclouds — coreweave, crusoe-energy, nebius, lambda-labs for off-grid / bridge sites
• Industrial — mining, marine, rail, oil & gas (legacy heritage)

Pain solved: immediately-available, multi-fuel backup or bridge power at hyperscale, plus the global service network needed when a campus is running engines 24/7 for months while waiting for utility interconnect. Compared to caterpillar, Cummins is stronger in the natural-gas reciprocating engine category — increasingly the preferred bridge-power technology because pipeline gas is cheaper and quieter than diesel.

3. Competitive Landscape

Cummins' edge: dual-fuel (diesel + gas) flexibility, deep service network, and a product gap CAT can't fully cover — the 1-3 MW lean-burn gas reciprocating sweet spot ideal for behind-the-meter prime power.

4. Unique Observations

• Section 4 focus — 1 MW build-cost share: backup gensets at 1.2-1.5× IT load capacity, reciprocating engines at ~$300-500/kW installed = **$0.4-0.7M/MW IT load**, or roughly 3-5% of a $11-20M/MW AI DC build cost. Bridge-power configurations (continuous run, emissions controls, exhaust treatment, fuel handling) push this materially higher.
• Q1 2026 Power Systems is the story: Power Systems revenue $2.0B (+19% YoY) with 29.5% EBITDA margin — record [1][3]. North America power generation +23%, international +16%, China power generation +84% — all driven by data-center demand [1][3]. Total company revenue $8.4B (+3%); the rest of the portfolio is flat-to-down, masking how dominant Power Systems' contribution to incremental earnings has become.
• AI in total revenue: Power Systems is ~24% of the company by revenue but rising — and almost all of Cummins' incremental 2026 growth is data-center driven [3]. Management raised full-year 2026 revenue guidance to +8% to +11% explicitly because of data center power demand [1][3]. AI exposure inside Cummins is concentrated but no longer dilute.
• Backlog and lead times — the bottleneck: backlog for large-scale gensets extends into late 2026 / early 2027, with some segments stretching into 2028 [5][6]. Industry-wide lead times for power-system components run 12-36 months [5] — the same supply-chain squeeze that Wood Mackenzie has documented for gas turbines (+195% price, 18-24 month lead) is also reshaping reciprocating-engine economics. This converts engine OEMs from "build to order" into "auction the slot."
• Multi-fuel as moat: Cummins is one of the only OEMs with both a competitive diesel lineup AND a credible 1-3 MW natural-gas reciprocating engine portfolio. As bridge power moves the use case from "rare run during outages" to "primary site power for 12-36 months," gas-engine economics matter (fuel cost, emissions, noise). This is the slice that's growing fastest.
• Token cost chain — the "shovel seller": Cummins sits in the L1 A.b backup-power layer. Engines are consumed by behind-the-meter power as bridge generation for AI campuses, then transition to standby. Every MW of new GPU compute requires ~1.2-1.5 MW of reciprocating engine capacity behind it — ahead of any token actually being generated. The current 12-36 month lead time means Cummins' Q1 2026 backlog already encodes the AI capacity coming online in 2027-2028.
• Customer concentration: extremely diversified globally; no single customer >5%. Less direct hyperscaler concentration than vertiv.

5. Financials / Funding

• Listed: NYSE: CMI; market cap ~$55-60B range mid-2026
• Q1 2026 revenue: $8.4B (+3% YoY) [1]
• Q1 2026 net income: $654M [1]
• Q1 2026 Power Systems revenue: $2.0B (+19% YoY); 29.5% EBITDA margin — record [1][3]
• Q1 2026 NA power generation growth: +23% [1]
• Q1 2026 China power generation growth: +84% [1]
• 2026 revenue guidance: +8% to +11% (raised) [1][3]
• 2026 EBITDA margin guidance: 17.75% – 18.50% [1]
• Q1 2026 capital returned to shareholders: $519M [1]
• Backlog: extends into late 2026 / early 2027, some into 2028 [5][6]
• Industry lead time: 12-36 months on large gensets [5]

6. People & Relationships

• CEO: Jennifer Rumsey (since 2022)
• CFO: Mark Smith
• HQ: Columbus, Indiana (1919 founding)
• Brand portfolio: Cummins (engines, gensets), Onan (commercial gensets), Atmus (filtration, spun off 2024), Accelera (zero-emissions)
• Top customers: extremely diversified — major hyperscalers + colos + AI cloud customers including crusoe-energy for behind-the-meter generation; truck OEMs (PACCAR, Daimler, Stellantis); industrial accounts
• Strategic position: viewed by 2026 markets — alongside caterpillar — as one of two genset-OEM AI infrastructure plays. Cummins' relative advantage is the 1-3 MW gas-engine sweet spot for bridge power