Energy volatility is turning AI’s power demand into a near term US challenge

The combined effects of the rapid expansion of artificial intelligence (AI) and renewed volatility in global energy markets, amid geopolitical tensions centered on the Iran crisis, are bringing U.S. energy security strategy into a new phase. AI-driven electricity demand was previously framed as a medium-term challenge, with projections suggesting that data center energy demand could nearly double by around 2030. However, recent disruptions in global energy markets are pulling this issue into the realm of near-term strategic concern, as energy shocks could transmit into electricity markets and lead to sharp price volatility over a very short period of time. As electricity demand rises alongside growing instability in global energy markets, the key question is whether the U.S. power system can continue to expand in a stable, predictable, and scalable manner to support the AI industry, for which a stable electricity supply is effectively a lifeline.

There is already broad consensus that AI will be one of the main drivers of electricity demand growth in the coming years, but projections of electricity supply expansion are not keeping pace. According to the U.S. Energy Information Administration (EIA), total U.S. electricity generation is expected to grow only modestly in the near term, at around 2% annually. By contrast, electricity consumption from data centers is projected to rise much more rapidly, as noted above. This divergence raises concerns not only about overall supply adequacy, but also about system reliability. Even if total generation expansion appears sufficient on paper, power systems must maintain adequate reserve margins to ensure stability under peak demand and unexpected disruptions. At the same time, electricity demand is also increasing in other sectors, including electrification of transport and industry, further tightening system capacity. Industry signals are already reflecting this imbalance. Experts and industry leaders have warned that U.S. energy infrastructure is not being built quickly enough to support AI-driven demand.

Meanwhile, recent geopolitical tensions involving Iran further amplify uncertainty in the energy outlook. Concerns over potential disruptions to key transit routes such as the Strait of Hormuz, as well as broader instability in a region central to global oil and gas supply, have contributed to rising and increased volatility in oil and natural gas prices. Given that natural gas accounts for roughly 43% of U.S. electricity generation, fluctuations in energy price inevitably affect the electricity market. While the current energy crisis has not directly translated into a domestic electricity crisis—partly due to the U.S.’s position as a major natural gas producer—it nevertheless exposes the system’s sensitivity to external shocks. What had previously been seen as a longer-term concern—whether electricity supply can keep pace with AI-driven demand—now appears more immediate, as a system already operating with limited spare capacity becomes more vulnerable to external shocks such as geopolitical crisis.

The recent tension affected the natural gas production of Iran and Qatar—both major players in global natural gas supply—have already contributed to price surge in global gas markets. In Europe, benchmark natural gas prices have increased by roughly 50% to 60% since the onset of the Iran crisis, with even sharper short-term spikes observed during periods of escalation. Although less directly exposed to immediate supply disruptions, the U.S. is still closely connected to global energy markets through liquefied natural gas (LNG) trade. When prices rise in higher-demand regions such as Europe, U.S. exporters have strong incentives to redirect supply to the region—which is also aligned with U.S. interests. This export response can tighten domestic supply conditions and place upward pressure on U.S. natural gas prices and eventually cause electricity prices to increase. 

Moreover, the current stage of AI development in the U.S. further amplifies the potential consequences of any disruption in electricity supply. The sector has already entered a phase of large-scale, capital-intensive investment, with over $1.6 trillion committed since 2013. These investments are long-term and largely irreversible, creating strong path dependency that constrains future pathways. Once built, such infrastructure must operate and iterate continuously to justify costs, making it difficult to slow or pause expansion without incurring significant economic losses.

At the same time, AI development follows a scale-driven trajectory, in which performance improvements are closely tied to increases in data, computing power, and model size. Insufficient investment in these areas risks slowing technological progress and undermining the returns needed to justify existing capital commitments. Conversely, maintaining competitive performance often requires continued expansion of computing capacity, which further increases electricity demand. In other words, without significant breakthroughs in efficiency, electricity demand will continue to rise alongside AI development.

Faced with this dilemma in AI development, the U.S. must not only expand electricity supply to support continued growth, but also ensure that future power systems remain resilient in the face of a “Black Swan Event” such as the Iran crisis. Energy sources that are heavily exposed to global fuel markets may struggle to meet these requirements, as their costs and availability can fluctuate with geopolitical developments.

In this context, electricity sources that are less dependent on fuel inputs and more insulated from international market volatility take on greater strategic significance. Renewable energy, in particular, offers a comparatively stable cost structure once deployed and is not directly subject to the same geopolitical risks that affect globally traded fuels.

In addition, renewable energy and AI infrastructure share an important spatial characteristic: both require large amounts of land but are relatively flexible in location. This creates opportunities for co-location in regions with abundant and low-cost land, such as parts of the American Southwest. Co-developing power generation and data center infrastructure in these areas can help reduce transmission constraints and operating costs, while also enhancing system resilience in times of potential disruption.

At the same time, it is important to recognize the limits of this approach. Expanding renewable energy capacity is neither immediate nor without constraints. Deployment takes time, and renewable development in the U.S. continues to face challenges related to supply chains, critical materials, and grid integration. Moreover, energy is also only one of several constraints on AI development, alongside factors such as computing hardware, data availability, and regulatory uncertainty. Addressing the energy dimension alone will not guarantee the smooth expansion of AI.

Yet these limitations do not diminish the relative importance of the shift. Under current conditions—where electricity demand is rising rapidly, and exposure to global energy volatility remains a structural risk—reducing reliance on fuel-based uncertainty becomes a practical priority. Renewable energy stands out as a comparatively more reliable pathway for supporting continued expansion of AI.