US data centers to use more natural gas than most countries - report

BloombergNEF projects US data-center electricity generation will add 15 Bcf/d of natural-gas demand by 2035, more than double its prior 6.9 Bcf/d forecast. Gas is expected to provide 69% of power for new grid-connected facilities, while total power-sector demand rises 18 Bcf/d to 54 Bcf/d and LNG-export demand increases 21 Bcf/d. US producers are projected to add 35 Bcf/d of output through 2035, but would need another 11 Bcf/d to satisfy forecast demand, supporting a constructive long-term outlook for gas infrastructure and producers despite substantial forecast uncertainty.

Analysis

The investable implication is not simply higher Henry Hub: incremental load is concentrated near transmission-constrained data-center corridors, creating basis and capacity scarcity before a national gas shortage appears. Appalachia producers EQT, AR and CTRA have the lowest-cost molecules but remain constrained by takeaway; midstream owners WMB, KMI and TRP should capture more durable returns through contracted pipeline, storage and lateral-build demand. Merchant gas-fired generation is a more selective beneficiary: VST and NRG can monetize tight reserve margins where data-center load arrives ahead of new generation, while regulated utilities may face capex pressure and political resistance to rate recovery.

Over the next 1-3 months, the market will likely discount the forecast because project announcements are not equivalent to energized load. The relevant catalysts are signed utility service agreements, interconnection milestones, pipeline expansion FIDs and power-price forwards in ERCOT, PJM and the Southeast; these can re-rate infrastructure names before physical gas demand appears. Over 6-18 months, a sustained tightening thesis requires gas-directed drilling to remain capital-disciplined despite stronger strip pricing—otherwise associated gas from oil basins can cap Henry Hub while regional spreads widen.

Consensus may be over-allocating to AI semiconductors and underpricing the physical bottleneck: reliable power, firm gas transport and gas storage. The contrarian risk is that utilities force data centers to procure behind-the-meter generation, renewables plus storage, or demand-response contracts, shifting value from upstream gas to equipment providers such as CEG, GEV and ETN. The thesis is falsified if announced hyperscale capacity fails to convert into load, regional power forwards soften, or producer capex lifts supply growth faster than firm transport and LNG demand absorb it.

Key Decisions for Investors

• Initiate a 6-12 month long WMB / short XLU pair: WMB has fee-based exposure to incremental gas transport and storage, while utility-sector returns face funding and regulatory-lag risk from accelerated grid investment. Target 10-15% relative return; exit if announced pipeline/lateral projects fail to reach FID or utility rate-case outcomes improve materially.
• Accumulate EQT and AR on Henry Hub-driven weakness rather than chase spot-price spikes; use a 12-18 month horizon and size as a supply-discipline trade. Prefer EQT for scale and marketing optionality, AR for liquids diversification; reduce if 2027 gas strip falls below full-cycle reinvestment economics or management budgets materially accelerate.
• Maintain a tactical long VST versus short NEE over 3-6 months in ERCOT/PJM load-tightness scenarios. VST has more direct merchant-power upside from firm-load scarcity, whereas NEE's valuation is more sensitive to financing costs and regulated/renewable execution; invalidate on falling regional forward power prices or evidence of substantial data-center load deferrals.
• Set an alert—not a position—for KMI and TRP when new data-center service agreements identify gas-backed generation or firm pipeline capacity. The missing diligence item is contract tenor and whether capacity is take-or-pay; without it, projected demand should not be capitalized as durable midstream EBITDA.