開發商為了在人工智慧建設中搶佔先機，經常將同一個專案向多家公用事業公司提出申請，以尋求最優惠的條件和最快的審批速度。這種被業界稱為「幽靈」申請的行為，不僅使審批時間大幅延長，也讓電網營運商難以準確預測真實的電力需求，甚至可能削弱美國在全球人工智慧競賽中的競爭力。
為了應對激增的申請量，許多公用事業公司開始提高門檻，要求支付高昂的預付款、提供鉅額擔保以及具備近乎完美的信用評級。例如在德州，州長已下令暫停審批並對相關專案進行審查，而一些公用事業公司也改採「準備就緒者優先」的模式，以淘汰缺乏實際執行能力的開發商。
According to projections from Wood Mackenzie, over two-thirds of the electricity sought for the US artificial intelligence boom is unlikely to materialize due to "phantom" projects and unrealistic pitches. Grid operators and utilities are expected to commit to only about 28 percent of the requested power capacity for data center projects.
To gain a competitive edge in the AI build-out, developers often pitch the same project to multiple utilities in search of the best deal and fastest approval. This practice, known in the industry as "phantom" applications, significantly stretches approval timelines, makes it difficult for grid operators to accurately forecast true power demand, and could potentially undermine US competitiveness in the global AI race.
In response to the surge of requests, many utilities are raising barriers by demanding steep upfront application costs, significant collateral, and near-perfect credit ratings. In Texas, for instance, the governor has ordered a pause and audit of such projects, while some utilities are shifting to a "first-ready, first-served" model to weed out developers lacking the actual capability to deliver.