U.S. prosperity is increasingly interlinked with AI, accounting for nearly half of the S&P 500’s market capitalization. Sustaining America’s hard-won competitive AI edge hinges on uninterrupted access to natural gas, which now comprises approximately 40 percent of the power mix in U.S. AI data centers.
To date, AI data centers have flourished largely because of natural gas-fired power generation, the least-capital-intensive option to power these data centers. Given current technological limits, data centers consuming some natural gas is unavoidable; although adopting them might reduce consumer utility costs, renewables, whose generative capabilities are intermittent and require either substantial storage capacity, cannot fully match gas generation’s speed and reliability.
At first blush, America’s natural gas bounty—trillions of cubic meters, enough to power the country for at least another 80 years—suggests advancing the American AI colossus is inexorable. Yet this bounty belies future downward pressure on production. Through 2034, output is expected to fall below two percent per year on average while demand surges. Output must also remain consistent with rising AI data center and consumer needs, and economic growth.
Meanwhile, utility bills for many struggling Americans are climbing as AI data centers, already a major source of public frustration, increase their electric loads; average consumer electricity prices rose by five percent over the previous year despite natural gas generation increasing by four percent last year. AI data center power consumption is slated to inflict more pain on consumers, projected to more than double between 2025 and 2027 to 66 gigawatts. Also, roughly one in five data center campuses is expected to exceed one GW—roughly one-fifth of New York City’s entire electrical load—in scale by 2030, and one in three by 2035. Big Tech is pouring big dollars into gas to sate these eye-popping demands. Last year, Google committed $40 billion to Texas gas infrastructure through 2027. Amazon is building a sprawling data center campus in Mississippi, powered by a $1.2 billion, 754-megawatt combined-cycle combustion-turbine gas plant.
Concerningly, U.S. adversaries could exploit Big Tech’s burgeoning dependence on gas. Between March and June, Iran intercepted most liquefied natural gas (LNG) shipments transiting the once-bustling Strait of Hormuz (SOH), through which one-fifth of the world’s seaborne LNG passed annually during the pre-war era. America’s vast reserves insulated U.S.-based AI data centers from SOH-related LNG price shocks. But even if that insulation lasts, many U.S.-affiliated AI data centers are now dispersed globally, often in places sensitive to market agitations. For example, like their stateside companions, most upcoming AI data centers in Europe—where LNG prices soared by 35 percent in March after the SOH closure—will concentrate on natural gas; in the UK, more than 100 new data centers plan to burn it.
China’s willingness to disrupt LNG supplies to slow America’s AI advancement remains a wildcard. Beijing—officially planning to become AI’s global leader by 2030—recognizes that AI data centers are key variables in strategic competition with the United States. So, envisioning future Chinese interventions against AI data centers via their LNG supplies in the South China Sea is not inordinately difficult. Like Iran, China could target LNG shipments in or near the Strait of Malacca, the world’s busiest LNG thoroughfare, to secure strategic AI advantages over the United States, or especially during a war, say, over Taiwan. Indeed, as Iran did to spike oil prices in March and April in the SOH, China merely threatening LNG shipping could punish U.S. consumers.
A course correction is due. So long as AI’s reliance on natural gas predominates, policymakers face a Gordian Knot of soaring consumer energy prices, environmental degradation, climate pressures, and adversary coercion. Decisive policy action could optimize Big Tech investments in emerging technologies, such as multifunctional fuel cells, and, to the extent possible, integration with renewables. Steadily reducing this reliance would safeguard America’s technological edge and economy from external shocks. And it would deliver much-needed relief to consumers.
