S06.2 · Energy, Power & Climate
Gathering, processing, transport and storage between the wellhead and end markets; a fee-based segment now being pulled into direct datacenter gas supply.
Midstream infrastructure — gathering, processing, transport and storage between the wellhead and end markets — runs on fees, and that single fact separates its economics from everything upstream of it. Global revenue is $900B-1.1T in 2025, a range that reflects inconsistent scope definitions across Mordor Intelligence and Precedence Research rather than genuine uncertainty about the business. Growth has run 3-5% historically and should hold in the mid-single digits on US LNG export buildout and Permian gas takeaway. The live question sits on the demand side: AI-driven datacenter gas load is pulling midstream operators into direct behind-the-pipe supply arrangements that blur the boundary with power generation — a new commercial category, not an automation story.
Geography follows the LNG and gas-export buildout: the US Gulf Coast, the Permian Basin, Canadian oil-sands pipeline corridors and Middle East gas-gathering systems. Tariffs are FERC-regulated in the US and state-controlled across much of Asia and the Middle East. North American pipeline capacity concentrates in five hands — Enterprise Products, Energy Transfer, Kinder Morgan, Williams and Enbridge hold the majority between them — with a long tail of smaller regional gatherers beneath.
The segment sits between the wellhead and refining, LNG export or the utility burner-tip. Revenue is fee-based, with 80-90% under take-or-pay contracts, pairing low direct commodity exposure with high capital intensity. That contract structure is the analysis: within the same value chain, revenue accrues to whoever holds the long-term shipping or processing contract, not to whoever wears the spot price of the commodity moving through the pipe. The binding constraint has shifted as well. Multi-year permitting, as pipeline approvals grow increasingly contested, now determines which projects get built more than capital availability does.
The adjacencies all point outward: LNG export terminals, carbon transport and storage, and renewable natural gas gathering each turn existing right-of-way and compression assets into new fee streams rather than defending against a threat.
Agentic AI automates pipeline integrity monitoring, leak detection and pigging schedules — collapsing the inspection labor cost line — but the durable moat is right-of-way and permitting, and AI replicates neither. The demand side matters more. AI-driven datacenter gas demand is pulling midstream operators into direct behind-the-pipe supply deals with data-center developers, a commercial category that did not exist at scale before. This is new-build demand, not a reallocation of existing gas volumes: it comes from incremental datacenter load rather than displacement of existing burner-tip customers, and it accrues to whichever operator already holds compression and right-of-way capacity near a given datacenter cluster. Predictive maintenance applications are already in use; fully autonomous pipeline operations remain 5-10 years out.