Chevron eyes more deals to power US data centers

Chevron is expanding into AI-driven data center power with Project Kilby, a 2.67-gigawatt natural gas facility dedicated to Microsoft’s Pecos, Texas campus, with first power expected in 2028. The company is also exploring additional deals across the Midwest, Gulf Coast, Rockies, and Utah, potentially creating a new revenue stream less exposed to commodity price swings. Chevron plans a final investment decision by year-end, though analysts say it is still too early to gauge the scale of the business.

Analysis

This is less an isolated Chevron story than an early signal that data-center power is becoming a quasi-infrastructure asset class, with long-duration contracted cash flows that sit somewhere between midstream and IPP economics. The second-order winner is not just CVX, but the whole “picks-and-shovels” stack around large-load interconnects: turbine OEMs, EPC/engineering, gas gathering, transmission gear, and grid services. If this template scales, the bottleneck shifts from compute demand to permitted power delivery, which tends to re-rate scarce assets in the Southwest and Gulf Coast first.

For Chevron, the strategic value is option creation: it monetizes gas optionality without taking full exposure to merchant power volatility, but the real margin pool likely accrues to whoever controls the site, transmission, and interconnection queue position. That means the competitive moat is more local than national; utilities and gas processors with existing infrastructure near AI clusters could quietly outperform the headline beneficiaries. Microsoft gains supply certainty, but it also implicitly accepts long-dated execution risk, so the market may eventually reward hyperscalers that lock power early and punish those still reliant on spot grid capacity.

The main risk is timing mismatch: the revenue is years out, while capex, permitting, turbine lead times, and grid studies are immediate. If AI capex cools or power demand estimates normalize, the market could mark down the “AI power” theme before first electron. A more interesting contrarian angle is that the trade is likely under-discounting gas supply constraints and over-discounting turbine scarcity; if this becomes a broad trend, the constraint is equipment availability, not demand, which is bullish for pricing power at OEMs but could delay economics for developers.

Key Decisions for Investors

• Go long GEV on any pullback over the next 1-3 weeks; the setup improves if more AI-power announcements follow because turbine backlog and pricing power can compound before revenue recognition. Risk: project delays, but reward is multiple expansion on scarcity economics.
• Pair trade: long GEV / short MSFT into strength over the next 1-2 months if the market starts extrapolating power capex pressure onto hyperscalers. This isolates the infrastructure beneficiary versus the customer funding the buildout.
• Initiate a small long CVX position with a 6-12 month horizon, but size it as an option on incremental deals rather than core energy beta. Upside comes from de-risked, fee-like revenue; downside is that this remains a low-probability, long-dated earnings stream.
• Consider a basket long in industrial power chain names versus broad software if follow-on deals emerge: long GEV/CAT on dips, short a basket of crowded AI software leaders over 1-3 months. The thesis is that power scarcity, not compute enthusiasm, becomes the scarce monetizable theme.
• If you have access to listed options, buy longer-dated upside in GEV rather than chasing CVX. The market may overreact to the strategic narrative faster than fundamentals, giving better convexity in the equipment supplier than in the sponsor.