The AI buildout is a monumental undertaking, and the scale of spending and corporate investment is unprecedented. The author has been writing extensively about this phenomenon all year, highlighting the staggering amounts being poured into artificial intelligence. However, the real question remains: where is all this money going?
The AI buildout differs significantly from traditional infrastructure projects like railroads. While railroads require substantial resources like steel mills and a large workforce, the AI industry's focus is on data centers. Interestingly, estimates suggest that a staggering 80% of data center costs will go directly to Nvidia, with the remaining 20% allocated to rack components and other related expenses. The actual construction and economic impact of these data centers might be a mere 5% or less of the total spending.
This disparity raises doubts about the applicability of historical multiples. The author expresses skepticism, questioning whether the traditional metrics used to evaluate such projects are still relevant. The AI industry's energy demands are immense, and the construction of power generation facilities is a classic example of a broadly dispersed expense. The numbers surrounding power generation are truly astonishing, and the author's concern is valid: there may not be enough capacity to meet the growing demand for data centers.
A striking example is the situation in Texas, where interconnection requests from data centers alone amount to 474GW (ERCOT). This is a staggering 90% of the total requests, and it's no wonder that Governor Abbott froze the rollout of new data centers in the state. To put this into perspective, the all-time peak demand record for ERCOT is 91.1GW, achieved on July 22, 2026, with normal demand ranging from 40 to 80GW. Morgan Stanley's projections indicate a potential deficit of about 100 nuclear power plants across the entire United States by 2030.
This energy crisis raises a critical question: if these data centers are built on time, will they be allowed to connect to the grid for years? The author's skepticism is understandable, given the potential for a significant energy shortage. The AI buildout's impact on the energy sector and the broader economy is a complex and intriguing topic that warrants further exploration and discussion.
