Data center companies delay IPOs amid growing public opposition

Several AI data-center-linked companies have postponed IPOs amid local opposition to power-intensive facilities, elevated interest rates and uncertain development demand. SoftBank subsidiary SB Energy delayed a planned IPO that sought a valuation of at least $50 billion after bankers reportedly struggled to find buyers at targeted price levels. Holtec paused its planned Nasdaq listing indefinitely, which had targeted up to $900 million in proceeds and a valuation of as much as $10 billion, while Aggreko also slowed its offering process.

Analysis

The relevant signal is not a change in AI compute demand but a higher cost of converting announced demand into operating capacity. If private capital cannot clear at valuations built on rapid data-center deployment, the bottleneck shifts from GPUs to interconnection, local permitting and firm-power availability. That favors listed owners of scarce, already-permitted generation and transmission assets—Constellation Energy (CEG), Vistra (VST), GE Vernova (GEV), Quanta Services (PWR), and Eaton (ETN)—over pre-revenue power developers whose terminal values assume uninterrupted construction pipelines.

For META, the near-term read-through is modestly negative to capex efficiency rather than to advertising earnings: delayed campuses can defer depreciation and cash outlays, but they also constrain the pace at which inference capacity can be deployed. The market has rewarded hyperscalers for aggressive AI capex; a 1-3 month rise in project delays would create dispersion between firms with contracted power/owned infrastructure and those reliant on greenfield buildouts. Over 6-18 months, local opposition raises the probability that capacity is redirected toward politically receptive regions, nuclear-adjacent sites, and behind-the-meter gas generation—potentially improving pricing power for CEG/VST while increasing stranded-development risk for merchant developers.

The contrarian case is that IPO weakness reflects financing-market selectivity, not a broad demand impairment. A lower-rate move or a high-profile hyperscaler power-purchase agreement could quickly reopen the financing window and reverse a short in speculative data-center infrastructure. The thesis is falsified if hyperscaler capex guidance remains intact while announced projects obtain permits/interconnection approvals on schedule, or if CEG/VST forward power-price curves fail to firm despite incremental data-center contracting.

Key Decisions for Investors

• Add a 3-6 month long CEG / short META relative-value position, sized modestly: CEG captures scarcity rents from firm clean power while META faces incremental deployment friction. Target 10-15% relative outperformance; exit if META guides higher AI capacity availability without a corresponding increase in contracted power costs.
• Prefer long PWR and ETN over private-market data-center-development exposure for the next 6-12 months. Their revenue is tied to grid hardening, interconnection and electrical equipment regardless of which proposed campuses ultimately proceed; reassess if backlog growth or book-to-bill falls below 1.0x.
• Maintain VST as a watch-to-buy on any 8-12% pullback rather than chase strength. The key confirmation is disclosed long-duration data-center power contracting and tightening ERCOT/PJM forward curves; a sustained decline in those curves would invalidate the scarcity-premium thesis.
• Do not infer a directional NDAQ trade from delayed listings. Monitor the 1-3 month IPO calendar and pricing discounts: multiple withdrawn infrastructure offerings would be a broader risk-appetite signal, but isolated postponements are insufficient to impair exchange earnings.