The emergence of Artificial Intelligence (“AI”) has triggered one of the largest growth spurts in US electricity demand in decades. Early indicators show some impacts of this already as grid electricity prices are increasing, signaling that electricity demand is growing faster than supply. While the underlying technologies powering AI systems are expected to become significantly more energy-efficient over time, the sheer scale and pace of AI adoption suggest that overall electricity generation will need to increase. New reports suggest that traditional forms of electricity, such as fossil fuels and renewables, may play a larger role in meeting this demand than frontier sources of electricity such as nuclear.
Electricity Prices Rise
My colleague Sonu Varghese, VP, Global Macro Strategy, has highlighted the persistent inflation in electricity prices. Shown below, the annual rate of electricity price increases has recently hovered between 2% and 5%—a marked rise from the pre-pandemic range of flat to 2%. This sustained elevation is one of the clearest indicators that electricity demand is growing faster than supply.
The emergence of AI is certainly a part of the growth of energy demand. Offsetting this though is the fact that AI’s main systems have shown a history of becoming more efficient over time. Although the introduction of AI caused a large step-function change in energy demand, more efficient systems may provide downward pressure on electricity demand growth over time.
Nvidia’s Exponential Efficiency Gains
Technological progress, particularly in semiconductor development, continues to drive improvements in energy efficiency. Nvidia’s latest generation of chips, known as ‘Blackwell,’ deliver nearly 40x more token output than their previous generation, referred to as ‘Data Center Productivity’ in the table below. I wrote about this breakthrough recently and it reflects a broader phenomenon: as AI usage increases, the output-per-watt from data centers is also rising sharply. Put simply, today’s AI is likely to be the least energy efficient version we’ll ever use.
Still, More Energy Will Be Needed
Even when accounting for these efficiency gains, electricity demand is expected to grow. Estimates of the growth rate vary widely, but a recent report from RMI1 estimates the US will need to grow its electricity generation by a total of 18% over the next decade, or about 1.6% annually. In my view, that’s both a rationally sound estimate and a feasible undertaking.
Stay on Top of Market Trends
The Carson Investment Research newsletter offers up-to-date market news, analysis and insights. Subscribe today!
"*" indicates required fields
Initial signs point to combined cycle gas-turbines as early beneficiaries of this demand, and there’s a potential for ‘renewables plus storage’ systems to gain share. Lazard’s 2025 Levelized Cost of Energy (“LCOE”)2 report shows gas turbines offering the lowest overall cost of generation. Turbine manufacturers, such as GE Vernova, have been early winners of this demand for electricity generation. As shown in the chart below, GE Vernova’s equipment revenue backlog has grown nearly 29% over the past two years and speaks to the influx of orders the company has received to meet this demand.
Solar and wind when paired with storage are a close second choice as the cost of these systems decline. Lazard’s report went on to show that the per-watt cost for utility-scale solar deployments decreased 4% year over year, which was the steepest decline in costs amongst the generation sources researched2. Solar panel manufacturers, such as First Solar, have iterated panel designs and technology to better harness this source of energy. These cost reductions, paired with the greater need for electricity, have allowed First Solar to increase sales by nearly 45% over the last three years. Analysts polled by FactSet expect even stronger growth with current estimates showing First Solar may grow revenue another 60% over the next three years (data as of 6/26/2025).
The growth in electricity demand is largely being met by more traditional forms of energy generation as well as renewables plus storage where viable. More frontier forms of energy generation, such as nuclear, remain a questionable opportunity.
Recent U.S. nuclear project histories raise serious concerns amidst cost and time overruns. NuScale in 2015 attempted to develop a 720 megawatt small module reactor (“SMR”) and estimated costs to be roughly $3.1 billion3. In 2021, the company re-estimated that costs would be closer to $9.3 billion (a 3x overrun) and the project was scrapped3. Separately, the Municipal Electric Authority of Georgia did successfully bring online their Plant Vogtle recently. However, total costs were nearly $34 billion as compared to $14 billion initially expected and the units entered service seven years later than expected4.
These anecdotes aren’t to say nuclear generation is impossible given the current landscape but should illustrate the conundrum facing electricity seekers. Utility owners face a tradeoff between nuclear projects potentially incurring unexpected deviations that lead to higher costs over the lifetime of the plant or sticking with more proven sources that deliver modest outcomes. There’s unlikely to be a perfect solution, but more conventional forms of electricity have seen the most success so far.
AI services continue to proliferate in our lives. The impact of these services is already being felt as grid electricity prices rise, at least due in part to this. While today’s AI is likely to be the most power-hungry form we ever use, the net load growth of roughly 1.6% annually will need to be met. Proven forms of electricity, such as gas and renewables, are currently the most viable paths forward. Meanwhile, nuclear remains full of potential but burdened by uncertainty and delays. Perhaps AI itself can deliver breakthroughs in this area. For now, AI looks like it will be powered by more stable sources.
- https://rmi.org/gas-turbine-supply-constraints-threaten-grid-reliability-more-affordable-near-term-solutions-can-help
- https://www.lazard.com/media/eijnqja3/lazards-lcoeplus-june-2025.pdf
- https://www.nuclearcosts.org/timeline
- https://www.powermag.com/vogtle-nuclear-expansion-price-tag-tops-30-billion/
For more content by Blake Anderson, CFA®, Associate Portfolio Manager click here
8119165.1.-06.27.25A
