Quick Thinking, Slow Building: The Energy Cost of America's AI Boom

12 May 2026 Content Page 3-4Executive Summary Page 5-10Can power markets handle the AI surge? Page 11-12Beyond the baseline: Agentic AI, adoption velocity andthe rebound effect Page 13-17The hidden bill?: AI’s impact on power prices Page 18-19Preparing AI for energy and energy for AI: policyrecommendations SummaryExecutive •Artificial intelligence is about to impose the largest sustained demand shock onUS electricity infrastructure in decades.By 2030, data-center power consumptionis expected to nearly double, lifting the sector’s share of total US electricity demandfrom roughly 5% to around 9%. Although planned generation additions looksufficient on paper, data centers may absorb nearly half of projected new capacity,leaving thin margins if electric-vehicle adoption or industrial electrification Patrick HoffmannEconomist, ESG & AIpatrick.hoffmann@allianz.com Katharina UtermoehlHead of Thematic and Policy Research Data centers can be built in under two years, but grid connections can take upto seven years in congested markets such as Northern Virginia. Nationwideinterconnection requests now total 1.84 terawatts, exceeding total installed USgenerating capacity. Supply-chain shortages have pushed lead times for criticalgrid equipment to several years, while projected demand would require buildingroughly 8,000 km of high-voltage transmission lines annually – around ten times •Aggregate electricity prices have not yet fully reflected these pressures, buta growing „data-center premium“ is emerging.States hosting the highestconcentration of data-center activity have so far seen price inflation below thenational average, reflecting favorable grid conditions, economies of scale andthe lagged structure of utility rate-setting. Beneath the surface, however, theimpact has become increasingly visible since 2023. US residential customers arealready paying USD1.4bn more per year on their electricity bills as a direct result ofdata-center demand, with just five utilities serving 4.4mn households in Northern gap that has already narrowed from 5% to 3.7% since 2020 and is set to close further.Data-center investment grew 32% in 2025 and is set to rise a further 75% in 2026alone, pointing to an additional electricity price effect of close to 14pps for the mostexposed utilities over 2025–2026, nearly doubling the cumulative four-year effectin just two years. Meanwhile investor-owned utilities filed USD18bn in rate-increase •Preparing energy infrastructure for AI and addressing community concernsis as urgent as building the AI infrastructure itself.The immediate priority isinterconnection reform: binding timelines, penalties for speculative queue filings andpriority treatment for shovel-ready projects with firm power commitments wouldhelp relieve the most acute bottlenecks. Cost allocation is equally important. Unlessdata centers bear a proportionate share of the infrastructure costs they create, Can power markets handle the AIsurge? The AI revolution is reshaping global economies,starting with power markets.In the IEA’s baselinescenario, AI-fueled growth in data-center powerdemand is expected to increase between 58% and 137%across major world regions between 2025 and 2030,supported by annual investment that now rivals thescale of the entire global oil and gas industry (Figure1).¹ The shift is particularly pronounced in the US, wheredata-center power demand is projected to reach 426 to AI remains difficult to isolate, bottom-up modelingsuggests AI inference already accounts for 12–14% oftotal consumption, with total AI data-center capacityreaching 29.6 GW by end-2025.² The implications forpower systems are already visible in the US, where thedata center share of total electricity consumption could ¹ Key Questions on Energy and AI (IEA)² The 2026 AI Index Report | Stanford HAI³ 2024 United States Data Center Energy Usage Report | Energy Technologies Area This significant demand growth is fueling a massiveexpansion in generation capacity.According tothe Energy Information Administration’s (EIA) Short-Term Energy Outlook, power generation in the USis expected to grow by 441 TWh between 2023 and2027, an increase of 10.5%.⁴ This would mean the fiveyears since the breakthrough of generative AI see alarger increase than the preceding two decades, overwhich power generation grew by only 7.7%. 76% of theincrease in power generation is expected to be driven multiple directions simultaneously. Beyond datacenters, electrified transport alone could materiallytighten the balance. Under the IEA’s baseline scenario,US electricity consumption from electric vehicles isprojected to rise from 25.2TWh in 2024 to 91.4TWhby 2030. Near-term signals are mixed, but the risksremain tilted to the upside: Secondary-market EV salescontinue to rise despite the expiry of the federal taxcredit, while gasoline prices elevated by the escalationin the Middle East are reinforcing the economic case On paper, the capacity additions currently incl