Energy Transfer Is Quietly Becoming One of the Biggest Natural Gas Suppliers to AI Data Centers

Energy Transfer signed multiple gas-supply deals tied to AI data-center power demand, including ~900,000 Mcf/d for Oracle (three U.S. sites) plus 150,000 Mcf/d deals for Nexus and 900 MW support for Crusoe. It also secured long-duration utility supply, including a 20-year Entergy agreement for at least 250,000 MMBtu/d starting Dec 2028, and 300,000 Mcf/d to four Oklahoma gas plants through 2028. The article cites EBITDA growth of at least 17.5% in 2026 and guidance for its >6% distribution to rise 3%-5% annually, tempered by a permitting delay that pushes one Oracle-linked project ~6 months.

Analysis

ET is best thought of as a toll-road on a structural power bottleneck, not just a dividend name. The market usually underprices this kind of demand because the first dollar of AI spend gets booked to GPUs and cloud capex, while the enabling infrastructure earns slower, but more durable, contracted cash flows with high incremental ROIC. That argues for a modest multiple re-rate if management can keep converting announcements into commissioned laterals and mainline expansions.

Second-order beneficiaries are the gas basin and midstream ecosystem, but the public-market payoff will likely lag the headlines. Contract wins do not equal EBITDA until permits, interconnects, and downstream power plants are actually online; that gap is where consensus is too optimistic. The more important read-through is regional basis strength in Texas/Arizona/Louisiana and improved utilization for adjacent pipes/processors, while merchant power developers and grid-constrained data-center builds face the highest execution friction.

The key risk is substitution: if gas prices, basis, or local opposition push delivered power costs too high, operators will slow campuses, redesign load plans, or wait for grid upgrades. Over the next 1-3 months, watch for slippage in in-service dates and whether ET’s growth guide is reiterated without caveats; over 6-18 months, the thesis breaks if AI load growth proves more deferred than incremental, or if permitting delays become the norm rather than exceptions.

Key Decisions for Investors

• Long ET on pullbacks over the next 1-3 months; thesis is that contracted data-center gas volumes and low-capex laterals justify a growth multiple above a traditional MLP. Risk/reward is attractive if the market starts pricing ET as AI-infrastructure leverage rather than yield-only.
• Buy ET 6-12 month call spreads instead of outright stock if you want convexity to contract conversions and project FIDs without full balance-sheet exposure. Use this only if near-term price action stays choppy and implied vol remains reasonable.
• Pair trade: long ET / short BE over 3-6 months. ET owns the bottleneck economics with contracted transport and gathering exposure; BE has more site-specific execution risk and is more vulnerable if customers choose cheaper gas-turbine or grid-backed solutions.
• Watchlist alert for ORCL and META next earnings: any commentary on power-driven capex delays or campus timing slippage would be a bearish read-through for AI infrastructure names, while a cleaner commissioning schedule would validate ET’s backlog conversion.
• Do not chase utilities broadly here; wait for evidence that regulated returns are translating into earnings, not just higher capex. If ET’s 2026 EBITDA guide is raised without new delay disclosures, add to longs; if guidance is reaffirmed but project timelines slip, take profits.