This Boring Pipeline Stock Just Signed a Deal to Power AI Data Centers

ONEOK agreed to supply gas to a 1-gigawatt power plant for data centers, a likely early example of additional AI-power-related infrastructure deals. Management said the $100M project should generate a “very nice return,” while also indicating it is in late-stage discussions for “a couple of other opportunities” to supply AI data centers. The company reported Q2 net income up 13% and raised its full-year net income forecast to $3.6B at the midpoint, with ONEOK’s dividend yield currently above 4.5%.

Analysis

This is more important as a business-development signal than as an earnings event. For OKE, the real value is not the first project’s dollars but the proof that AI load growth can create a repeatable pipeline-lateral pipeline, which is accretive because it tends to be fee-based, capital-light relative to legacy growth, and quicker to monetize than large-diameter greenfield builds. The second-order winner is broader midstream with local takeaway/storage optionality—WMB and KMI should be watched for similar announcements—while utilities that were counting on being the sole AI-power gatekeeper may see some load growth bypass the regulated grid entirely.

The near-term market reaction is likely to overstate the economic impact. A single deal won’t move OKE’s 2026-2029 cash flow profile much; the catalyst path is 1-3 months of follow-on announcements, permitting commentary, and any incremental capex guidance tied to data-center laterals. If those repeat, the story becomes a structural volume-growth lane for gas infrastructure; if they don’t, this fades back into a narrative trade. Key falsifiers are slower-than-expected AI buildouts, a shift back toward grid-tied utility solutions, or a collapse in project economics if gas prices/basis volatility make on-site generation less compelling.

Contrarian view: the market may be underpricing how selective this opportunity is. AI power demand is not a blanket bull case for every energy-adjacent name; it favors assets with short cycle times, existing rights-of-way, and contractual visibility. That argues for owning the midstream toll collectors rather than chasing the most crowded utility/nuclear proxies. But it also means the upside is probably steady compounding, not a rerating to growth-stock multiples—unless OKE proves the pipeline of deals is much larger than investors currently assume.

Key Decisions for Investors

• Long OKE on pullbacks as a 6-18 month compounding trade; thesis is incremental, high-return laterals adding to an already visible growth stack. Treat the first deal as a proof point, not the full story.
• Pair long OKE / short XLU for 1-3 months if the market keeps rotating toward utility-only AI beneficiaries; this expresses the view that gas infrastructure captures faster monetization than regulated load growth.
• Watch WMB and KMI for follow-on AI-power announcements; if they surface within the next quarter, add a basket long in midstream names with the cleanest fee-based growth and storage exposure.
• Do not chase OKE on the headline alone; require evidence of repeat deals or capex guidance creep before adding size. Falsifier: no additional AI-related project disclosures by next earnings and no change in 2026-2029 growth guidance.