Two Fossil Fuel Companies Are Betting Big on Data Centers

Oil and gas demand is being bolstered by the AI-led data center boom, with Williams and Chevron pitching natural gas/pipeline power as a multi-year growth driver. BloombergNEF estimates US natural gas production must rise 36% by the mid-2030s (partly due to data centers), while permit applications suggest gas-fired power plants connected to data centers could emit up to 21 million tons of GHGs/year (with Williams’ permits potentially 2/3 lower than modeled). Chevron’s multi-gigawatt Texas project with Microsoft includes a 20-year power purchase agreement, and Williams is building six “behind-the-meter” gas plants for data centers after announcing $5B+ investments in the segment.

Analysis

The real market implication is not higher oil beta; it is a rerouting of AI-related capex into assets that look more like contracted infrastructure than cyclic commodity exposure. That favors midstream/power-builders with rights-of-way, permitting, and financing muscle, while making regulated utilities and grid-tied renewables less relevant to the incremental load story because the load is being met off-grid. For CVX, the value is not the megawatt itself but the optionality to package energy reliability into long-dated, fee-like cash flows that can support a higher quality multiple.

Near term, the catalyst path is regulatory rather than operational: permit approvals, local tax fights, methane scrutiny, and whether hyperscalers keep signing multi-year PPAs. The reversal risk is in the 1-3 month window if gas prices spike enough to dent project economics or if state/federal agencies slow behind-the-meter approvals; over 6-18 months, cheaper batteries, SMRs, or faster interconnection reform could reduce the need for private gas islands. If that happens, the current enthusiasm for gas-backed AI power could prove a bridge, not a secular step-up.

The contrarian miss is that investors may be treating this as an ESG-negative headline when it is also a capital-allocation story: the highest-value customer in the market is effectively pre-committing demand to whoever can deliver reliable electrons fastest. The underappreciated loser is not just renewables; it is any utility valuation that relies on data-center load as a growth leg, because behind-the-meter builds steal that demand before it hits the grid. What would falsify this thesis is a meaningful shift by Microsoft/Meta toward grid-connected clean PPAs, a wave of permit denials, or a material collapse in gas-fired project returns versus contracted utility yields.

Key Decisions for Investors

• Long WMB / short XLU for 3-6 months: express the view that behind-the-meter gas and pipeline monetization win while utility load-growth assumptions get bypassed. Stop if state regulators accelerate interconnection reform or if WMB’s data-center backlog stops expanding.
• Long XLE / short ICLN for 6-12 months: gas-linked infrastructure should capture the first derivative of AI power demand, while pure-play clean-power names face a weaker load-growth narrative. Falsify if hyperscalers pivot back to grid-tied renewables at scale or carbon policy tightens unexpectedly.
• Buy CVX 12-18 month call spreads on pullbacks: the direct earnings impact is modest, but the market may re-rate the stock if it starts valuing the infrastructure option embedded in its long-duration power contracts. Risk/reward works best only if the market gives back post-print enthusiasm.
• Watch KKR rather than chase it: it benefits from financing private power assets, but the edge is dealflow-dependent. Only initiate if announced commitments to AI power projects re-accelerate above the current pace.