What happened
Business Insider reports that a wave of speculative AI data‑center projects—dubbed “Bragawatts” and “ghost projects”—are inflating the perceived scale of the AI infrastructure boom. Bernstein Research analyzed thousands of announced U.S. data‑center plans and concluded that roughly 35% of the announced capacity is likely to be built. Many developers have filed multiple electricity‑connection requests, especially in Texas, where the total demand from these requests exceeds the state’s existing power supply by more than five times. The article cites Amazon’s documented struggles to secure enough electricity for its AI workloads as a contrast to less‑experienced developers announcing massive facilities without secured power, permits, or customers.
Bernstein Research examined a database of announced U.S. AI data‑center projects, cross‑referencing them with electricity‑connection applications, construction permits, and known customer contracts. The study found that only about one‑third of the announced capacity is likely to be realized.
The report identifies two informal industry terms: “Bragawatts,” referring to large‑scale projects announced by firms with limited data‑center experience, and “ghost projects,” where developers submit multiple power‑connection requests hoping at least one will be approved. In Texas, the cumulative power requests from these ghost projects exceed the state’s current electricity supply by a factor of five.
Business Insider contrasts these speculative projects with Amazon’s documented challenges in securing sufficient power for its AI workloads, underscoring that experienced operators still face real grid constraints.
Source details: businessinsider.com ↗
Why it matters
The overstatement of AI data‑center capacity can mislead investors, policymakers, and local communities about the true scale of electricity demand and environmental impact. If utilities plan for a larger load than will actually materialize, they may over‑invest in generation or transmission, leading to stranded assets and higher costs for ratepayers. Conversely, under‑estimating the real projects—like Amazon’s—could leave the grid vulnerable to shortages as AI workloads grow. The analysis also highlights a risk of “ghost” projects crowding permitting processes, potentially delaying legitimate builds and creating regulatory fatigue. Understanding the gap between announced and actual capacity is crucial for accurate forecasting of AI‑related energy consumption, for shaping responsible AI‑infrastructure policy, and for investors assessing the viability of data‑center developers.
Inflated capacity forecasts can distort utility planning, leading to unnecessary generation investments or grid upgrades that may never be needed, raising costs for consumers and taxpayers.
Environmental assessments based on overstated data‑center builds could misallocate resources intended to mitigate AI’s carbon footprint, while under‑estimating real projects could leave the grid under‑prepared for actual demand spikes.
Regulators may need to tighten reporting requirements for data‑center developers to ensure that announced projects are backed by concrete power contracts and permits, reducing the prevalence of ghost projects.
Interactive Mechanism: How It Actually Works
Explore the underlying technology behind this development interactively.
What should a useful AI forecast state?
What to watch next
Watch for utility filings and state regulator responses to the surge of electricity‑connection requests, especially in Texas and other high‑growth regions. Monitor whether developers of speculative projects secure permits, power contracts, or anchor customers, which would move them out of the “ghost” category. Follow any policy proposals aimed at tightening disclosure of data‑center capacity plans to prevent over‑promising and to protect grid reliability.
Utility commission filings in Texas and other high‑growth states for new electricity‑connection requests tied to AI data‑centers.
Announcements from developers confirming power purchase agreements, construction permits, or customer contracts that would move a project out of the speculative category.
Legislative or regulatory proposals aimed at increasing transparency of data‑center capacity announcements, such as mandatory disclosure of secured power and permitting status.