OpenAI deal adds marquee load but doesn't resolve pipeline uncertainty

What happened

Georgia Power, a subsidiary of Southern Company, announced it will serve OpenAI's new project in Effingham County, Georgia, with OpenAI committing up to 1,000 MW of flexible demand response to help manage grid reliability. The news bolsters Southern's narrative as a key beneficiary of AI-driven load growth, adding to the roughly 9 GW of contracted large-load capacity since 2023. However, the company's latest filings reveal significant pipeline volatility, with 14.3 GW of data center projects withdrawn and a net 6 GW decline in the interconnection queue, underscoring the gap between signed contracts and actual energization. The DeepValue report assigns a WAIT rating at $91.76, citing high leverage (4.5x net debt/EBITDA), negative free cash flow, and the need for visible queue stabilization before the growth story can be de-risked. This OpenAI commitment is a positive signal but does not alter the risk that further withdrawals or regulatory cost-allocation challenges could compress near-term returns.

Implication

Long-term investors should recognize the OpenAI agreement as a vote of confidence from a major AI player, but the core thesis remains dependent on sustained regulatory support in Georgia and successful conversion of the pipeline into rate base. Given pipeline churn and high capex needs, waiting for a more attractive entry near $85 or evidence of sequential improvement in energized load offers a better risk-reward profile.

Thesis delta

The OpenAI announcement adds a prominent customer to Southern's large-load roster, incrementally supporting the demand growth narrative, but it does not address the key uncertainties: 14.3 GW of withdrawn projects, a declining net queue, and the risk of resource delays if load materialization falls short. The fundamental challenge of converting signed contracts to energized load within the 2027-2028 window remains unchanged, and the stock's premium valuation (23.6x P/E) leaves little room for execution stumbles. Accordingly, the WAIT stance is maintained, with the bull case requiring tangible proof of queue stabilization and energization milestones over the next two quarters.