A recent report by energy intelligence firm Currence forecasts that between 30% and 50% of large-scale data center capacity anticipated for 2026 will likely face delays. The report cites rising electricity demand, construction bottlenecks, and unreliable project timelines as major contributors to this slowdown in growth.
Currence’s findings, based on 777 announced AI factories and large data centers with capacities exceeding 50 megawatts, reveal an industry where project announcements outpace actual construction. Since 2024, developers have declared plans for around 190 gigawatts of future capacity, with about 16 gigawatts set to commence operations in 2026. However, only approximately 5 gigawatts of this capacity is currently under construction, leaving 11 gigawatts in a planning phase without any visible progress, despite typical lead times of 12 to 18 months.
The report emphasizes that while there appears to be significant growth potential, the reality is different, with 30% to 50% of the planned capacity unlikely to be operational by year’s end. Challenges such as local opposition, limited availability of skilled labor, and increasingly unreliable schedules have hindered progress in data center construction.
The electricity supply crisis particularly affects the situation, as local communities have pushed back against data centers that have led to soaring electric bills. In response, some operators are now required to generate their own power rather than rely on local power grids. Interestingly, projects utilizing on-site or hybrid power systems represent less than 10% of announced facilities but account for nearly half of the total capacity.
To improve their public image, large tech companies, or hyperscalers, are adapting. For instance, Google is enhancing its renewable energy initiatives, while Amazon is increasing its investments in solar energy and battery storage solutions.
The distinction between announced and actionable projects is becoming crucial for utilities, suppliers, and investors trying to predict future demands. Although estimates of data center-related debt are difficult to quantify, Bloomberg suggests total outstanding debt may exceed $500 billion for AI data centers. Additionally, recent reports indicate that companies like Meta, Google, Amazon, Microsoft, and Oracle have accumulated approximately $1.65 trillion in “hidden debt,” significantly higher than their recorded liabilities.
Amidst these financial strains, Meta’s substantial off-balance sheet debt is notable, estimated to be about $420 billion, which is three times its reported debt. Despite having excess capacity, Meta is keen on expanding further, with BlackRock raising $12 billion for a new data center on its behalf.
In summary, while there is a predicted surge in data center projects, significant hurdles—ranging from local opposition to energy supply issues—threaten the timely realization of these expansions.
