As the artificial intelligence (AI) investment boom continues to drive up the scale of data center construction, how many of these huge expansion blueprints can actually be implemented? Wall Street investment bank Bernstein released a research report stating that the North American data center capacity it tracks to be built has reached 586GW (gigawatts). However, after a comprehensive assessment of "construction progress, customer credit, developer strength and regulatory risks", only about 199GW has a high degree of credibility for implementation, accounting for about 34%. This means that nearly two-thirds of the planned capacity is still uncertain.
However, Bernstein emphasized that this does not mean that all other projects will be cancelled, but that many projects are still in the early planning stage, and there are great uncertainties in construction time, financing capabilities and final delivery scale.
The article points out that there is another figure in this report that deserves more attention, that is, even after risk adjustment, the credible construction capacity still increased from 136GW in July this year to 199GW in October, a significant increase of about 46% in just 3 months. In other words, although a large number of planned projects may not be implemented, the data center capacity in North America with high construction certainty is still expanding rapidly.
According to data center capacity tracked by Bernstein, the capacity to be built increased from 411GW in July this year to 488GW in August, and further climbed to 586GW in October. During the same period, the risk-adjusted credible construction capacity also increased from 136GW to 170GW and 199GW, accounting for 33%, 35% and 34% respectively. Although the planning scale continues to expand, the proportion of capacity with high construction certainty has always remained at about one-third.
The report pointed out that the biggest factor causing the gap is the progress of project development. At present, about 497GW of capacity is still in the announced planning stage, accounting for about 85% of the capacity to be built, but the credible capacity is only 123GW, and the credibility is about 25%. In comparison, there are about 87GW of projects that have entered the construction stage, with a credible capacity of 75GW, and a credibility of 86%.
In addition to construction progress, customer credit and developer type are also important factors. The credibility of data centers built by large cloud computing companies is about 75%, and that of professional data center operators is about 62%. The credibility of emerging AI cloud service providers is only 29%, and projects transformed from cryptocurrency mining facilities are as low as 24%.
Among them, the four major technology giants Amazon, Alphabet, Microsoft and Meta contributed a total of approximately 78GW of trusted construction capacity, accounting for approximately 39% of the total. Amazon leads the way with 33GW, followed by Alphabet with 20GW, Meta with 15GW and Microsoft with 10GW.
On the other hand, local governments in the United States have become increasingly strict in regulating data centers. Texas is the largest data center planning market in the United States, with about 150GW of capacity to be built, but Bernstein estimates that only 43 to 44GW has a high degree of confidence. Grid access and approval restrictions may delay projects that have not yet started construction, but the approximately 21GW of capacity that has already entered the construction phase will be relatively less affected by the new restrictions.
Bernstein believes tighter regulations are more likely to result in project delays or higher construction costs than widespread cancellations. Data center construction usually takes 2 to 3 years, and even longer in areas with tight power supply. Short-term market fluctuations will not immediately change the construction pattern of the entire industry.
Overall, Bernstein believes that the AI data center construction boom is still continuing, but as financing capabilities, customer credit, and regulatory conditions become the key to project progress, companies with actual construction experience, stable funding sources, and high-quality customers will have a better chance of standing out in this wave of AI infrastructure expansion.
