Energy Transfer vs. Williams Companies: Which Natural Gas Giant Wins the AI Power Buildout?

Williams is positioned as the more direct natural-gas beneficiary of AI and data-center power demand: it transports roughly one-third of U.S. natural-gas output and derives 100% of adjusted EBITDA from gas operations, versus about 40% for Energy Transfer. Williams trades at 14x forward adjusted EBITDA compared with Energy Transfer's 7x, reflecting its AI-power exposure and C-corp structure; Energy Transfer offers a higher 6.5% forward yield versus Williams' 2.9%. Both operators are insulated from commodity-price volatility through fee-based pipeline revenue, while natural gas supplies more than 40% of U.S. data-center grid electricity, according to the IEA.

Analysis

This is not a fresh fundamental catalyst; it is a retail-friendly reframing of an existing power-demand theme. The relevant question is not gas throughput today, but whether incremental data-center load converts into contracted, FERC-permitted expansion projects with returns above each company's cost of capital. WMB has cleaner upside sensitivity to incremental Northeast/Mid-Atlantic gas demand, but that also makes its valuation more exposed to any delay in utility interconnects, data-center construction, or pipeline permitting.

The valuation dispersion creates a more interesting contrarian setup than an outright AI-infrastructure purchase. WMB's premium embeds sustained rate-base-like growth and low regulatory friction, while ET's discounted multiple reflects both non-gas exposure and the MLP investor-base constraint; however, ET can still monetize the same upstream gas-supply response through gathering, processing, NGL and export-linked systems. If AI-driven electricity demand lifts gas production rather than merely regional basis spreads, ET's broader footprint may capture more of the second-order capital cycle than the market assigns.

Over the next 1-3 months, monitor announced Transco expansion commitments, binding shipper contracts, and the capex/EBITDA multiple implied by new projects rather than headline data-center announcements. Over 6-18 months, power generation additions, utility load forecasts, and regional gas basis behavior will determine whether throughput demand is real. A sustained decline in forward power prices, project cancellations by hyperscalers, adverse FERC outcomes, or WMB guidance that fails to translate demand into contracted backlog would falsify the premium-purity thesis.

Key Decisions for Investors

• No outright chase of WMB on this article alone; retain/add only after a disclosed contracted expansion or guidance increase. A failure to show backlog conversion by the next two earnings updates is a reason to avoid paying further premium multiple expansion.
• Consider a 6-12 month relative-value position: long ET / short WMB in equal dollar amounts if the EV/EBITDA valuation gap widens further without a corresponding increase in WMB contracted-growth guidance. The payoff is mean reversion plus ET distribution carry; exit if WMB secures material long-duration Transco commitments or ET cuts distribution/raises leverage guidance.
• Use KMI as a liquid additional watch-list comparator for U.S. gas-power demand. If WMB and KMI both announce contracted capacity expansions while ET's gas-linked EBITDA outlook remains unchanged, abandon the ET catch-up thesis rather than assuming broad gas demand automatically benefits all midstream assets equally.
• Set an event-driven alert around FERC approvals, utility integrated-resource-plan revisions, and hyperscaler power procurement announcements in PJM/SE markets. These are more actionable for WMB than generalized AI capital-expenditure headlines and should precede any durable earnings revision.