U.S. Data Center Power Consumption Projected to Quadruple by 2035
A BloombergNEF report projects US data center electricity consumption will quadruple by 2035, reaching 20% of total national generation. The rapid surge in demand for AI compute infrastructure is causing significant strain on regional power grids, potentially increasing costs for high-compute operations like video encoding and delivery.
Key Takeaways
- U.S. data center capacity is forecast to reach 200 gigawatts by 2035, with nearly 50% dedicated to AI training and inference.
- Electricity demand projections for 2035 were revised upward by 83% since December, reflecting an accelerated development pace.
- The PJM Interconnection grid is expected to allocate 34% of its total electricity to data centers, while ERCOT in Texas will hit 22%.
- PJM network power prices surged 76% year-over-year as the grid struggled to manage connection requests and supply-demand imbalances.
Why It Matters
The massive expansion of energy-intensive AI infrastructure will likely inflate operational costs for streaming platforms reliant on cloud-based video encoding and delivery. As regional grids like PJM hit capacity limits, providers may face higher transit fees and localized surges in computing costs. The centralization of AI hardware in the U.S. creates a bottleneck for global scaling, forcing engineers to prioritize energy-efficient compression and edge-caching strategies to mitigate grid-driven price volatility. Watch for American Electric Power’s potential exit from the PJM Interconnection as a signal of localized grid fracturing and its impact on regional cloud availability.
Additional Context
The strain on the U.S. power grid has forced hyperscalers to seek alternative energy sources to sustain expansion. Per The Wall Street Journal in September 2024, Microsoft signed a 20-year power purchase agreement with Constellation Energy to restart a reactor at the Three Mile Island nuclear plant specifically to power its AI data centers. This trend toward 'behind-the-meter' nuclear energy highlights the desperation of major cloud providers to bypass public grid congestion and secure stable pricing. Similarly, Amazon Web Services acquired a 960-megawatt data center campus from Talen Energy in March 2024, which is connected directly to a nuclear facility in Pennsylvania, insulating it from the price volatility seen in the broader PJM market. Simultaneously, cooling requirements for high-density AI chips are exacerbating resource scarcity beyond just electricity. According to a July 2024 report by the Financial Times, Google’s water consumption rose 17% in a single year, primarily driven by the cooling needs of its data centers. This environmental footprint is becoming a regulatory hurdle; in June 2024, local authorities in Ireland and parts of the Netherlands tightened planning permissions for new builds due to concerns over infrastructure exhaustion. For streaming and CDN operators, these constraints suggest a shift toward more efficient liquid cooling technologies and ARM-based processors, which offer better performance-per-watt than traditional x86 architectures in a power-constrained environment.
Read full article at techcrunch.com
Get this in your inbox → Subscribe
Enjoy our coverage?
Add StreamingMeme as a preferred source on Google to see more of our streaming news at the top of your Search results.
Add as preferred source