The Case for Energy Investment
The United States needs power investment. Power generation and power transmission, as well as upfitting the existing grid infrastructure, are often grouped alongside the AI investment thesis. However, with AI potentially stalling out due to DeepSeek, how vulnerable is the utility investment theme to a similar stall out?
Over the past 2 years, AI investment has led to the surge in data center construction all across the United States, but specifically in the Southeast and Northern Virginia regions. Since data centers have an insatiable demand for electricity, big tech companies have been scrambling to find solutions to power these new data center investments. Microsoft, which their CEO has said the company is power constrained in the AI build out, has reached deals to restart the nuclear reactor on 3 Mile Island. Other companies such as Google, Meta and Amazon have been in talks to invest in Small Modular Nuclear Reactors (SMRs) as a way to solve their AI power needs.
AI is forecast to increase electricity demand in data centers by 165% by the end of the decade. Utility providers usually expect a .5% increase in power demand per year, therefore the surge of data center power demand has created problems for utilities to meet this unexpected rush of demand.
However, AI is in a similar stage to the internet build out in the early 2000s and there are real risks that AI investment is getting out in front of its skis. Aside from Meta and Walmart, few companies have been able to justify the AI investment with efficiency or ROI gains. Microsoft has also notably been cutting back on data center leases in a sign that they may be oversupplied. Microsoft has also been quiet in terms of their next fiscal year AI investment; the company repeatedly has stated they plan to invest $80 billion dollars in AI for FY 2025, however their FY ends in June 2025 so much of this spending has already occurred. Lastly, the Chinese DeepSeek AI model threatens to turn the entire AI investment thesis on its head and even boasts a 10 to 40 times reduction in power needs.
Even with a potential over exuberance in the AI build out, power infrastructure investment still stands to have major gains from the second major growth theme in the United States: Onshoring.
President Trump has been very clear in his intention to bring American Manufacturing back online. Many have countered this stance stating that it will take years to build out manufacturing facilities. What these individuals fail to realize is that the covid supply chain disruptions kick started the onshoring process and manufacturing spending. Now in 2025, the manufacturing construction spending is starting to taper off and the production facilities are starting to open up. Take for example the Steel Dynamics aluminum plant in Mississippi or the TSCM chip foundry in Arizona, both set to start production in 2025.
The chart below highlights the YOY percentage change in construction spending for manufacturing (Blue) and power (Green) in the United States. As discussed, manufacturing construction went parabolic in response to the covid supply chain disruptions and now the increased spending is slowing. Notably, the change in power spending has been less dramatic, suggesting that there have not been equal investments in power to support these new manufacturing facilities.
Utility spending and construction has been difficult mainly due to regulation, permitting and rate increase caps. With Trump in Office, some of these hurdles may become less significant as the need to supply the American Economy with cheap, abundant power is one of the administration's goals and a key to economic growth.
Utility spending and construction will also benefit greatly from breaking out of a 20 year trend. For the past 20 years, power demand has essentially been flat. Although the population has increased, more efficient technologies have caused energy demand to stay relatively unchanged. By 2029 energy demand is forecasted to increase by 15%. This sudden upward increase in power demand will need to be addressed through additional investment and construction. With power construction currently accounting for roughly $150 billion per year, the 15% represents a roughly $20 billion increase in spend.
Looking further into the future, power infrastructure investment will also help to support increased electric vehicles. As well as robotic workforces that are currently being deployed with Amazon. Others, such as Tesla’s Optimus robot, promise future workforce productivity. In the short term, increased natural gas and oil production may be able to support the economy as it starts to build out the power infrastructure needed to support longer term growth and trends. It is time that utility infrastructure investment is talked about independently of adjacent themes such as AI or Onshoring.