Georgia Power's contract with OpenAI approved; Latest approval part of portfolio of large-load contracts delivering approximately $950 million in annual savings beginning in 2029

Georgia Power’s contract to serve OpenAI’s new project in Effingham County has been approved by the Georgia PSC, adding 3,200 MW of new demand. The deal (with OpenAI paying infrastructure costs) is expected to contribute savings of about $950 million per year starting in 2029, lifting typical residential savings to $180/year (from $102/year committed in Dec. 2025; about $15/month). The framework also relies on up to 1,000 MW of flexible demand response to limit incremental generation buildout.

Analysis

The key market mechanism is not the branded customer win; it is regulatory validation that Georgia Power can convert incremental load into lower system costs while keeping recovery largely off its own balance sheet. That should modestly improve the perceived durability of SO’s earnings path and, more importantly, support the utility multiple by reducing the fear that AI-driven load growth is automatically margin-dilutive for incumbents. The second-order beneficiary is the broader regulated-utility complex: if Georgia remains a template, peers with large data-center pipelines can argue for similar cost-allocation frameworks.

The catch is timing mismatch. The economic benefit is back-end loaded while capex, transmission planning, and financing happen now, so the near-term upside to equity is limited unless investors start capitalizing 2029 load growth into the stock today. The real risk is that the demand is more flexible than firm: if the customer curtails usage more often than expected, Georgia Power may get less revenue than modeled while still carrying the infrastructure and execution burden. That makes the story more credit-positive than pure growth-positive.

Contrarian view: consensus may be overestimating how much of this is a direct earnings driver and underestimating how much is a policy precedent. If the PSC keeps approving customer-funded infrastructure and demand-response terms, SO’s downside from load growth is lower than the market assumes; if regulators tighten cost recovery or any large-load contract gets renegotiated, the thesis weakens quickly. Watch for the next rate-case language, financing plan, and any evidence that the 3,200 MW pipeline is slipping from a firm backlog into a marketing narrative.

Key Decisions for Investors

• Tactically accumulate SO on any post-event weakness; this is a 6-18 month multiple-support story, not a same-day EPS catalyst. Use a tight thesis stop if Georgia PSC commentary starts implying slower or partial cost recovery.
• Pair trade: long SO / short XLU for 3-6 months to isolate utility-specific data-center optionality versus the broad utility basket, which still prices in generic rate-case pressure. Falsify if SO underperforms XLU after the next PSC filing cycle.
• Prefer SO credit over equity at the margin: if spreads widen on capex concerns before 2029 benefits are visible, the bonds should better reflect the de-risked recovery structure than the stock. Exit if leverage metrics rise without confirmed approved recovery mechanisms.
• Do not chase into the print with calls; the market can easily overpay for a far-dated savings story. If a trade is needed, wait for a pullback and reassess only after the next regulatory filing or guidance update.
• Watch item: data-center cancellation/curtailment risk. If management starts qualifying the load as more flexible or less firm than implied, reduce exposure immediately because the upside case depends on load actually materializing.