. The question I have for you is how long you think society will continue to pay for an AI-powered future that may not benefit them
Energy demand from AI data centers is projected to grow thirtyfold within the next decade:
The law of supply and demand will take care of the energy needs, but costs per kWh are going to increase. The question becomes: Who pays for the additional cost?
In most markets, the answer is likely consumers. Not consumers of AI, mind you, but everyone.
Expect to hear more about this as the impact of AI becomes politically charged. It is not without reason or merit. AI will both disintermediate a subset of the population and raise their cost of living due to higher energy costs.
Our energy grid is already feeling the strain. Electricity costs in the US are up 5.8% over the past year. Utility companies are passing higher energy costs on to consumers. New Jersey’s Board of Public Utilities notified customers earlier this year that their bills will go up between 17% and 20% specifically due to data centers.
Data centers will account for up to 12% of our overall electricity consumption as early as 2028.
This wave of AI innovation could turn our default utility model—spreading costs across all ratepayers—into an Industrial Age relic, built for widely shared infrastructure like water mains or rural electrification. When a handful of companies account for most new demand—and their impact will likely cost jobs, not add more jobs in the local market—asking the public to subsidize data center power feels like a fight waiting to happen.
Dominion Energy Virginia proposed a separate “rate class” for data centers, making them pay closer to the true marginal cost of the power they use.
Meanwhile, data center developers are cutting deals for dedicated power delivery, elbowing their way to the front of the grid. Microsoft alone reportedly demands up to 5 gigawatts of power for its expansion plans. Building their own direct sources of electricity, while complex and time-consuming, could avoid much of the tension that is sure to arise around this issue.
AI’s growth is not self-funding. It needs infrastructure. Consumers are generally willing to support progress, but only to the point that it doesn’t undermine their own security.
Picks and Shovels
One thing is clear: We will build more electricity generation. That has led to an anticipatory boom in the stock price of utilities, certain construction and engineering firms, and midstream energy providers.
This is a big, diverse industry, and many smaller-cap companies are providing essential goods and services to the sector. Compared to the average company in the S&P 500, the stocks of many of these companies are cheap.
This is one pocket of the market where we are looking for—and finding—great opportunities to invest.
Finally, there’s an irony here many ignore: AI has the potential to improve energy efficiency and optimize the very grid it now threatens to overload.
I’m impressed by how AI’s capabilities grow seemingly overnight. I can see AI helping engineers design and test better energy infrastructure, from the individual capacitors to large-scale distribution lines. And I can also imagine software engineers using AI to optimize their systems for energy efficiency.
Smarter demand response, AI-enabled forecasting, and predictive maintenance could help lessen the power burden of automation on society.
For all its potential harms, people with new ideas have less of a barrier to market than ever before. For investors, there’s tremendous upside in innovations that can replace our aging power infrastructure or help AI companies run their data centers more efficiently. The question I have for you is how long you think society will continue to pay for an AI-powered future that may not benefit them.
from Global Macro Update
