S06.4 · Energy, Power & Climate
Refining crude into finished fuels and distributing them to end markets, with thin refining margins offset by steadier retail marketing economics.
Downstream refining, marketing and fuels converts crude into finished fuels and moves them to end markets. Global refining capacity runs roughly 102-105M bbl/d in 2025 (EIA/OPEC), with downstream revenue estimated at $2.5-3T (2025E, industry aggregates, sensitive to fuel prices). Capacity is roughly flat, growing about 1% a year; demand growth concentrates in Asia while OECD demand plateaus and declines with EV penetration. The earnings-quality point is the one to hold onto: refining margins are thin and volatile, and the durable earnings sit in retail marketing, behind the pump. AI enters on the cost side — refinery digital twins compressing the process-engineering line — and changes nothing about the demand trajectory.
New capacity concentrates in Asia-Pacific, particularly China and India; US Gulf Coast refiners run large, complex facilities; European refiners are rationalizing capacity. Ownership spans national oil companies (Sinopec, Aramco), international majors (ExxonMobil, Shell) and independents (Valero, Marathon), with the top 10 players holding roughly 25-30% of global capacity — genuinely fragmented.
Downstream buys crude from upstream and midstream and sells finished fuels into retail, wholesale and petrochemical channels. Crack-spread margins are thin and volatile, the segment is capital-intensive and cyclical, and retail marketing carries a materially higher and steadier margin than refining itself — which is why forecourt economics matter as much as refining throughput to overall segment profitability. Blended earnings quality is a function of mix: how much of the business sits behind the pump versus at the crack spread. An integrated operator with a large retail network and a pure-play refiner exposed entirely to the spread between crude and finished-product prices carry fundamentally different risk, whatever the capacity numbers say.
Petrochemical integration hedges long-run fuel-demand decline. Renewable diesel, sustainable aviation fuel, retail convenience and EV charging run the other way — they monetize forecourt real estate as fuel volumes plateau, adding revenue rather than defending it.
Agentic AI automates refinery process optimization and predictive turnaround scheduling, compressing the process-engineer cost line, and the operating moat migrates from operator tacit knowledge to plant digital twins, with widespread digital-twin adoption expected within 2-5 years. The money accrues to the operator as avoided downtime and lower maintenance spend, not incremental fuel revenue. With capacity and demand both roughly flat, this is a margin story, not a volume story.