US AI Data Centers Spur Power Demand, Diesel Backup, and Emissions Surge
The rapid expansion of artificial‑intelligence workloads has pushed data‑center electricity use up 17% and is expected to more than double overall demand by 2030, with AI tasks alone projected to require about 175 TWh in 2026. Power consumption for AI‑focused facilities is slated to triple, prompting operators to seek reliable backup sources.
Diesel generators remain the dominant standby power for large data centres because they can start within seconds and sustain high loads, especially as many U.S. regions grapple with grid‑connection delays and reliability challenges.
The surge in AI‑driven infrastructure is also driving a natural‑gas generation boom, raising emissions at major cloud providers. Climate Power’s emissions tracker gives Amazon, Microsoft and Google mid‑range scores (68, 64, 61) while Elon Musk’s xAI scores a 6, reflecting its reliance on unregulated gas turbines.
Entities: Amazon.com, Inc. · Climate Power · Google LLC · Microsoft Corporation · xAI, Inc.
Claims
What the coverage asserts, and how well corroborated each claim is across sources.
- [○ 1 SOURCE] Energy consumption for AI‑focused data center facilities is set to triple.
- [○ 1 SOURCE] Climate Power’s tracker gave xAI a score of 6, the lowest among eight companies, due to its use of gas‑turbine generators.
- [○ 1 SOURCE] AI tools and services have led to a 17% increase in electricity consumption for data centers.
- [○ 1 SOURCE] AI‑driven data centre build‑out is spurring a natural‑gas generation boom and double‑digit emissions increases at major cloud providers.
- [○ 1 SOURCE] Average cost of unplanned downtime for data centers has risen to $9,000 per minute ($540,000 per hour).
- [○ 1 SOURCE] 60% of small and medium businesses that experience catastrophic data loss close within six months.
- [○ 1 SOURCE] Diesel generators remain the most common backup power source for large data centers because they can start within seconds.
- [○ 1 SOURCE] Data center electricity demand is expected to more than double by 2030, with AI workloads accounting for about 175 TWh in 2026.